xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001681128
007 cr mnu|||uuuuu
008 060106s2003 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000617
Fernandez, Daniel Celestino.
Fourier-transform infrared spectroscopic imaging of prostate histopathology
h [electronic resource] /
by Daniel Celestino Fernandez.
[Tampa, Fla.] :
b University of South Florida,
Thesis (Ph.D.)--University of South Florida, 2003.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 126 pages.
ABSTRACT: Vibrational spectroscopic imaging techniques have emerged as powerful methods of obtaining sensitive spatially resolved molecular information from microscopic samples. The data obtained from such techniques reflect the intrinsic molecular chemistry of the sample and in particular yield a wealth of information regarding functional groups which comprise the majority of important molecules found in cells and tissue. These spectroscopic imaging techniques also have the advantage of acquisition of large numbers of spectral measurements which allow statistical analysis of spectral features which are characteristic of the normal histological state as well as different pathologic disease states. Databases of large numbers of samples can be acquired and used to build model systems that can be used to predict spatial properties of unknown samples.The successful construction and application of such a model system relies on the ability to compile high-quality spectral database information on a large number of samples with minimal sample-to-sample preparation artifact. Tissue microarrays provide a consistent sample preparation for high-throughput infrared spectroscopic profiling of histologic specimens. Tissue arrays consisting of representative normal healthy prostate tissue as well as pathologic entities including prostatitis, benign prostatic hypertrophy, and prostatic adenocarcinoma were constructed and used as sample populations for infrared spectroscopic imaging at high spatial and spectral resolutions. Histological and pathological features of the imaged tissue were correlated with consecutive tissue sections stained with standard histologic stains and visualized via traditional optical microscopy and reviewed with a trained pathologist.
Co-adviser: Levin, Ira W.
Co-adviser: M.D., Santo V. Nicosia
x Pathology and Laboratory Medicine
t USF Electronic Theses and Dissertations.
Fourier-Transform Infrared Spectroscopi c Imaging of Prostate Histopathology by Daniel Celestino Fernandez A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Pathology and Laboratory Medicine College of Medicine University of South Florida Co-Major Professor: Santo V. Nicosia, M.D. Co-Major Professor: Ira W. Levin, Ph.D. Wenlong Bai, Ph.D. Luis H. Garcia-Rubio, Ph.D. Maria Kallergi, Ph.D. Patricia A. Kruk, Ph.D. Date of Approval: May 20, 2003 Keywords: FT-IR, adenocarcinoma, vibr ational, spectroscopy, classification Copyright 2004 Daniel Celestino Fernandez
Dedication To my parents, for their love and support, and my wife, for always believing in me.
Acknowledgements Howard Hughes Medical Institute National Institutes of Health Research Scholars Program National Institutes of Health Graduate Partnership Program University of South Florida College of Medicine Department of Pathology and Laboratory Medicine National Institute of Diabetes, Digestive and Kidney Diseases Ira W. Levin, Ph.D. Santo V. Nicosia, M.D. Stephen M. Hewitt, M.D., Ph.D. Rohit Bhargava, Ph.D. Michael D. Schaeberle, Ph.D. Scott W. Huffman, Ph.D. Patricia McCarthy, Ph.D. Jamie Winderbaum Fernandez, M.D.
i Table of Contents List of Tables .....................................................................................................................iv List of Figures ......................................................................................................................v Abstract .............................................................................................................................vi i Chapter One Introduction ..................................................................................................1 1.1 Electromagnetic Spectrum .........................................................................................1 1.1.1 Interactions of Electromagne tic Radiation with Matter ..............................4 1.2 Basis of Infrared Absorption ......................................................................................5 1.2.1 Requirements for IR Absorption .................................................................6 1.2.2 Number of Vibrational Modes ....................................................................8 1.2.3 Group Frequencies ......................................................................................9 1.3 IR Spectral Feature of Tissues .................................................................................10 1.3.1 Proteins .....................................................................................................10 1.3.2 Carbohydrates ...........................................................................................15 1.3.3 Lipids ........................................................................................................15 1.3.4 Nucleic Acids ............................................................................................17 1.4 FTIR Spectroscopy Background ..............................................................................17 1.4.1 FTIR Spectrometers ..................................................................................18 1.4.2 Infrared Microscopy ..................................................................................20 1.4.3 Mapping with Single-Point Detectors .......................................................22 1.4.4 Raster-scan Imaging Using Multichannel Detectors ................................24 1.4.5 Global FTIR Spectroscopic Imaging ........................................................25 1.5 Spectroscopic Imaging: Data Structure and Applications ......................................27 1.5.1 Image Classification Methods ...................................................................30 1.6 Prostate Background ................................................................................................31 1.6.1 Anatomy and Histology ............................................................................31 1.6.2 Prostate Pathology ....................................................................................33 Chapter Two Methods .....................................................................................................44 2.1 Tissue microarrays ...................................................................................................44 2.1.1 Construction of Prostate Tissue Microarrays ............................................44 2.1.2 Array P-16 Design ....................................................................................45 2.1.3 Array P-40 Design ....................................................................................45 2.1.4 Array P-80 Design ....................................................................................46
ii 2.2 Tissue Array Section preparation .............................................................................47 2.2.1 Optical Substrates for Tissue Array Sections ...........................................47 2.2.2 Deparaffinization ......................................................................................48 2.2.3 Optical imaging of H&E sections .............................................................48 2.3 Spectroscopic Imaging Instrumentation ..................................................................49 2.3.1 Tissue Array FT-IR Data Collection Parameters ......................................51 2.3.2 Modifications and Environmental Considerations ....................................52 2.4 Data Handling and Computational Considerations ..................................................53 2.4.1 Data Pre-Processing ..................................................................................53 2.4.2 Spectral Baseline Correction .....................................................................54 Chapter Three Infrared Spectroscopic Histology of Prostate ..........................................56 3.1 Visualization of Spectral Images a nd Verification of Histologic Features ..............56 3.2 Creation of Ground Truth Data Regions of Interest ................................................58 3.3 Spectral analysis of histologic features and metric selection ...................................62 3.4 Construction of a Supervised Classification Model for Prostate Histology ............64 3.4.1 Spectral Data Reduction ...........................................................................64 3.4.2 Image Classification .................................................................................66 3.4.3 Array P-16, 20-metric, GML Self-classification results ...........................68 3.4.4 Leave-one-out metric evaluation ..............................................................70 3.4.5 Array P-16, 18-metric GML Classification Results .................................72 3.5 Validation of Prostate Hi stology Classification Model ...........................................75 3.5.1 Cross-Array Validation .............................................................................76 3.6 Conclusions and Further Directions .........................................................................79 Chapter Four Infrared Spectroscopic Histopathology of Prostate ...................................81 4.1 Classification strategy ..............................................................................................81 4.2 Array P-80 H&E Stained Section Pathology Analysis ............................................82 4.3 Array P-80 Histology Classification Results ...........................................................83 4.3.1 Spatial Filtering of Hist ology Classification Results ................................84 4.4 Construction of a Supervised Classification Model for Prostate Pathology ............88 4.4.1 Creation of pathology ground truth ROIs .................................................88 4.4.2 Pathology Spectral Data Reduction ..........................................................89 4.4.3 Histogram analysis of Spectral Metric Data .............................................91 4.4.4 Mean-centering of epithelial metric data. .................................................92 4.4.5 Metric Statistical Analysis ........................................................................92 4.4.6 GML Pathology Classification of Array P-80 ..........................................94 4.5 Individual Patient Evaluation of P-80 Pathology Classification ..............................96 4.6 Cross-Array Validation ............................................................................................97
iii 4.7 Conclusions and Further Directions .........................................................................98 References ........................................................................................................................101 About the Author...................................................................................................End Page
iv List of Tables Table 1.1 Spectroscopic techniques utilizing different regions of the electromagnetic spectrum ..............................................................................5 Table 1.2 Staging of primary tumor (T) .........................................................................42 Table 1.3 Staging of regional lymph node involvement (N) ..........................................43 Table 2.1 Spectral frequencies used for spectroscopic baseline correction ...................55 Table 3.1 Histologic class population data .....................................................................61 Table 3.2 Histology Spectral Metric Definitions ............................................................66 Table 3.3 Error Matrix of supervised GML Classification results using 20 spectroscopic metrics ..................................................................................69 Table 3.4 Confusion matrix of supervised GML Classification attempt using 18 spectroscopic metrics ..................................................................................72 Table 3.5 Revised 6-class histology gr ound truth ROIs for Array P-16 and Array P-40 .............................................................................................................77 Table 3.6 Error Matrix for 6-Cl ass, GML Classification Results ..................................78 Table 4.1 Pathology spectral metric parameters .............................................................90 Table 4.2 Results of t-test on mean adenocarcinoma metric values from population of 25 patients on array P-80 for 54 candidate pathology metrics .........................................................................................................93 Table 4.3 Error matrix for 20-metric pa thology GML classification of epithelial tissue on array P-80 .....................................................................................96
v List of Figures Figure 1.1 The electromagnetic spectrum ........................................................................2 Figure 1.2 The infrared region of the electromagnetic spectrum .....................................4 Figure 1.3 Vibrational modes and IR activity of water vapor (A) and carbon dioxide (B) molecules ...................................................................................8 Figure 1.4 Vibrational m odes of methylene group ...........................................................9 Figure 1.5 Structure of a typical amino acid ..................................................................11 Figure 1.6 Basic polypeptide structure ...........................................................................12 Figure 1.7 Common Protein Secondary Structures: -helix and sheet .......................13 Figure 1.8 Michelson Interferometer ..............................................................................19 Figure 1.9 Three Instrumental Approaches for collection of spatially resolved FTIR spectroscopic data ..............................................................................22 Figure 1.10 Schematic repres entation of the image cube ...............................................28 Figure 1.11 Zonal Anatomy of the Prostate ...................................................................32 Figure 2.1 Array P-80 Layout. ........................................................................................47 Figure 2.2 Spectrum Spotlight 300 Microscope Optical Configuration .........................51 Figure 3.1A) Baseline-corrected N-H stretching (3290cm-1) absorbance intensity image of four tissue a rray spots from a single patient on Array P-16 B) Optical images of corresponding H&E stained section. .........................................................................................................57 Figure 3.2 Absorbance Band Ratio Images of tissue array spot s from Array P-16 .......60 Figure 3.3 Histologic class mean spectra .......................................................................63 Figure 3.4Histograms of metric value cl ass frequency distribution for the three most populated classes (epithelium, mixed stroma, & fibrous stroma) for: A) Metric 02 (band ratio 1080/1544cm-1), and B) Metric 11 (band ratio 1400/1390 cm-1) ......................................................67
vi Figure 3.5 Graphical Representation of results of the leave-one-out analysis ...............71 Figure 3.6 Classification results for 2 ti ssue array spots from the same patient ............73 Figure 4.1Array P-80 histol ogy classification results ....................................................83 Figure 4.2 Optical images of H& E stained section of Array P-80 .................................84 Figure 4.3 Spatial filtering techniqu es for classified image results ................................86 Figure 4.4 Sieve operation spatial filtering of histology classfication results for patient 2 from array P-80 ............................................................................88 Figure 4.5 Array P-80 pathology ground truth ROIs ......................................................89 Figure 4.6 Patient-to-patie nt metric variation .................................................................91 Figure 4.7 Array P-80 pathology classification results ..................................................95 Figure 4.8 Individual patient analysis of 20-metric GML pathology classification .......97
vii Abstract Fourier-Transform Infrared Spectroscopi c Imaging of Prostate Histopathology Daniel Celestino Fernandez ABSTRACT Vibrational spectroscopic imaging techni ques have emerged as powerful methods of obtaining sensitive spatia lly resolved molecular information from microscopic samples. The data obtained from such techniques reflect the intrinsic molecular chemistry of the sample and in particular yield a wealth of information regarding functional groups which comprise the majority of important molecules found in cells and tissue. These spectroscopic imaging techniques also have the advant age of acquisition of large numbers of spectral measurements whic h allow statistical analysis of spectral features which are characteristic of the nor mal histological state as well as different pathologic disease states. Databases of large numbers of samples can be acquired and used to build model systems that can be used to predict spa tial properties of unknown samples. The successful construction and applicati on of such a model system relies on the ability to compile high-quality spectral da tabase information on a large number of samples with minimal sample-to-sample prepar ation artifact. Tissue microarrays provide a consistent sample preparation for high-th roughput infrared spectroscopic profiling of histologic specimens. Tissue arrays consisti ng of representative normal healthy prostate tissue as well as pathologic entities incl uding prostatitis, benign prostatic hypertrophy,
viii and prostatic adenocarcinoma were constr ucted and used as sample populations for infrared spectroscopic imaging at hi gh spatial and spectral resolutions. Histological and pathological features of the imaged tissue were correlated with consecutive tissue sections stai ned with standard histologic stains and visualized via traditional optical microscopy and reviewed with a trained pathologist. Spectral analysis of histologic class mean spectra and subseque nt cross-sample statistical validation were used to classify reliable spectral metrics fo r class discrimination. Multivariate Gaussian maximum likelihood classification algorithms were used to reliably classify all pixels in an image scene to one of six different hist ologic subclasses: epithelium, smooth muscular stroma, fibrous stroma, corpora amylacea, lymphocytic infiltration, and blood. The developed database-dependent cl assification methods were used as a tool to investigate subsequent microarrays designed with bot h normal epithelial tissue as well as adenocarcinoma from a large population of patients. Such investigation led to the identification of spectral featur es that proved useful in the preliminary discrimination of benign and malignant prosta tic epithelial tissue.
1 Chapter One Introduction Spectroscopy deals with the interaction of various forms of electromagnetic (EM) radiation with matter. Vi brational spectroscopy provides information regarding the molecular composition and struct ure of a wide range of mate rials including biological tissues. Recent technologi cal advances have led to powerful vibrational imaging approaches involving both near and mid-infr ared, as well as Raman-based platforms providing spatially-resolved chemical information on a microscopic scale. Infrared spectroscopic imaging microscopy, in partic ular, benefits from many decades of instrumentation advances and database compila tions. A brief background into the theory and techniques of infrared spectroscopy follows. 1.1 Electromagnetic Spectrum The wave nature of electromagnetic (EM) ra diation treats the ra diation in terms of oscillating electric and magnetic fields perpen dicular to one another and to the direction of wave propagation traveling with the velocity of light. Certain co ntinuous regions of the EM spectrum have been designated a nd appear in Figure 1.1. Vibrational absorption spectra result from the interacti on of oscillating dipole moments, which occur during molecular vibrations, with the electric field of the radiation, resulting in an energy exchange between the radiati on and the molecular system. Electromagnetic radiation is characterized by its wavelength The specific units typically used to express wa velength vary across the spectru m from angstroms () in the
gamma ray region to meters in the radio wave region or ~10-10 to 102 cm, respectively. The units of m are practical for describing radiati on in the mid-infrared spectral region. In the near-infrared (NIR) regi on the unit nm typically employed just as it is in the visible (VIS) and ultraviolet (U V) spectral regions. le cv Radio a Inre Fr ras m rays.05 10nm 350nm 770nm 2.5m 50m 1mm 300mm Inre Fr r m rays.05 10nm 350nm 770nm 2.5m 50m 1mm 300mm Ultraviolet Microwave Radio Waves Infrared Mid Near Far X-rays -rays & cosmic rays.05 10nm 350nm 770nm 2.5m 50m 1mm 300mm wavelength () frequency( ) energy (E) Ultraviolet Microwave Radio Waves Infrared Mid Near Far X-rays -rays & cosmic rays.05 10nm 350nm 770nm 2.5m 50m 1mm 300mm wavelength () frequency( ) energy (E) Infrared Mid Near Far X-rays -rays & cosmic rays.05 10nm 350nm 770nm 2.5m 50m 1mm 300mm wavelength () frequency( ) energy (E) Visible Figure 1.1 The el ectromagnetic spectrum Electromagnetic radiation can also be characterized by its frequency defined as the number of oscillations of the magnetic or electric field radiation vector per unit of time. The frequency unit is s-1 (oscillations per second), of ten specified in Hertz (Hz). The energy (E) of EM radiation is directly related to its frequency ( ) by the equation hE (1.1) where h is Planks constant with a value h = 6.63 10-34 J s. The frequency and wavelength ( ) of EM radiation are related by the proportionality constant c (the speed of light) according to the equation c (1.2) 2
where c has a value of ~2.99793 10 10 cm s -1 (in a vacuum). Infrared spectroscopists have adopted the convention of expressing frequency in terms of wavenumber with the units of cm -1 . A simple expression for wavenumber is given by 1 (1.3) The units of wavenumber provide a convenient scale for IR spectroscopy, especially the mid-infrared region that spans 200-4000 cm -1 The units of wavenumber are also desirable for IR spectroscopists because they are directly proportional to the energy of radiation, which varies inversely with wavelength as described by equation 1.4. hcE (1.4) The relationships between energy, frequency, and wavelength and the various regions of the electromagnetic spectrum are detailed in figure 1.1. The infrared region of the electromagnetic spectrum is subdivided into three contiguous regions; the near, mid and far infrared regions. The nomenclature of these prefixes refers to the individual sub-regions position relative to the visible region. Figure 1.2 shows these three regions of the infrared spectrum and the ranges they occupy on the wavelength, frequency and wavenumber scales. 3
12900 4000 200 103.9 10141.2 10146.0 10123.0 10110.77 2.50 50 1000 InfraredNearMidFar wavelength (m)frequency (Hz) wavenumber(cm-1)12900 4000 200 103.9 10141.2 10146.0 10123.0 10110.77 2.50 50 1000 InfraredNearMidFar wavelength (m)frequency (Hz) wavenumber(cm-1) Figure 1.2 The infrared region of the electromagnetic spectrum 1.1.1 Interactions of Electromagnetic Radiation with Matter All forms of spectroscopy deal with the interaction of radiation and matter. Numerous possible types of interactions exist and many involve transitions between specific molecular energy states. The monitoring of the absorption and emission of radiation from different regions of electromagnetic spectrum provides information regarding these molecular transitions and consequently gives information regarding the atomic and molecular composition of samples. Quantum mechanical treatments describe both the wave and particle nature of electromagnetic radiation[5, 6]. As seen in figure 1.1, the electromagnetic spectrum spans an extremely wide range of frequencies, and therefore, energies. There are a variety of energy levels that molecules can occupy leading to the possibility of many transitions between states. These energy transitions are of varying magnitudes with corresponding frequencies depending upon the specific regions of the spectrum in which they occur. Radiation from different regions of the electromagnetic spectrum are used as 4
the basis of the many spectroscopic techniques that exist, for which each technique provides molecular information regarding the sample. Table 1.1 contains examples of different types of spectroscopy based on specific regions of the electromagnetic spectrum and the type of chemical information probed. rotational tranisitionsmicrowave spectroscopy1mm to 300 mmMicrowavenuclear spin transitions (in magnetic field)Molecular StructureNMR Spectroscopy> 300 mmRadio Waves50 m to 1mmFar InfraredIR Absorption spectroscopyIR Relection spectroscopyIR emission spectroscopy2.5 m to 50 m Mid Infraredvibrational transitionsthermal emission770 nm to 2.5 mNear Infrared350 nm to 770 nmVisible (VIS)electronic transitionsfluorescence emissionvibrational transitionsUV-VIS spectroscopyfluorescence spectroscopyRaman spectroscopy10 nm to 350 nmUltraviolet (UV)electronic structuremolecular structurex-ray spectroscopyx-ray crystallography0.05 to 10 nmX-raysnuclear decay emission-ray spectroscopy< 0.05 -raysinformationspectroscopywavelength range ()spectral region rotational tranisitionsmicrowave spectroscopy1mm to 300 mmMicrowavenuclear spin transitions (in magnetic field)Molecular StructureNMR Spectroscopy> 300 mmRadio Waves50 m to 1mmFar InfraredIR Absorption spectroscopyIR Relection spectroscopyIR emission spectroscopy2.5 m to 50 m Mid Infraredvibrational transitionsthermal emission770 nm to 2.5 mNear Infrared350 nm to 770 nmVisible (VIS)electronic transitionsfluorescence emissionvibrational transitionsUV-VIS spectroscopyfluorescence spectroscopyRaman spectroscopy10 nm to 350 nmUltraviolet (UV)electronic structuremolecular structurex-ray spectroscopyx-ray crystallography0.05 to 10 nmX-raysnuclear decay emission-ray spectroscopy< 0.05 -raysinformationspectroscopywavelength range ()spectral region Table 1.1 Spectroscopic techniques utilizing different regions of the electromagnetic spectrum 1.2 Basis of Infrared Absorption Photons in the infrared spectral region have energies representative of transitions between molecular vibrational energy levels. While spectroscopic techniques exist which make use of the reflection and emission of infrared radiation, we are most concerned with the absorption of infrared radiation. Nearly all molecules exhibit an infrared spectrum, the noted exceptions being homonuclear diatomics, such as the common gases N 2 O 2 and H 2 . 5
Various interactions can o ccur between radiation and matter that result in the transfer of energy. Quantum mechanical pr inciples require that molecules exist in quantized energy states and thus the absorption of ener gy results in bands that characterize an infrared spectrum. 1.2.1 Requirements for IR Absorption The wave nature of quantum mechanics is most simply represented by the time independent Schrdinger equation E H (1.5) where is the wavefunction of the system, H is the Hamiltonian operator, and E is the energy of a state characterized by . The wavefunction can be used to calculate the transition moment R as shown in the equation j iR (1.6) for a transition between states i and j, where is the electric di pole moment operator ( = er, e is the electronic charge, r is the distance between the charges), and d indicates the integration over all space. For vibrational motions, the electric dipole moment is expressed as ... )( )( 0 2 2 2 2 1 0 0 r rr r rre e (1.7) where 0 is the permanent dipole moment, r is the internuclear distance and re is the equilibrium bond distance. If we consider only the first two terms in equation 1.7 and substitute for in equation 1.6 we obtain 6
)(00*jeirrrR (1.8) which reduces to )(0*jeirrrR (1.9) since 0 is a constant and because of the orthogonality of the wavefunctions. *ji From equation 1.8 it is clear that there must be a change in dipole moment during the vibration in order for a molecule to absorb infrared radiation. The selection rules predict that the fundamental absorption will occur with vibrational quantum number = for a harmonic oscillator, with much weaker overtone absorption corresponding to = etc. for anharmonic conditions. All molecules that are more complex than diatomics have multiple vibrational modes. These vibrational modes each have associated energies that correspond to the particular frequency or wavenumber of infrared radiation. The number, type, and energies of these vibrations are dictated by the molecular structure of the system in terms of the bonds, geometry, atomic masses, and force fields and are thus representative of specific molecules. Vibrational modes that produce a change in dipole moment result in the absorption of IR radiation and are termed infrared-active. Vibrational modes that do not induce in a change in dipole moment are termed infrared-inactive. The requirement for a change in dipole moment during a molecular vibration explains why, for example, homonuclear diatomic molecules do not absorb infrared radiation. 7
1.2.2 Number of Vibrational Modes While diatomic molecules can vibrate only in one dimension or mode, more complicated molecular structures present ot her possible vibrational modes. Linear molecules with N atoms exhibit 3N-5 vibra tional modes, while nonlinear molecules have 3N-6 vibrational modes. Water (a nonlin ear triatomic) and carbon dioxide (a linear triatomic) are illustrative examples. As s een in figure 1.4, the carbon dioxide molecules additional symmetry provides it with four possible vibrat ional modes while the water molecule has only three. Note also that the symmetric stretch of the carbon dioxide molecule produces no net change in dipole mo ment and is thus in frared-inactive. o3 2 position activity o3 2 position activity eBending (in plane) Bending (out-of-plane) t e -1 activity eBending (in plane) Bending (out-of-plane) t e -1 activity AB water carbon dioxide 1 n 3652 cm-1 IR-actived 1 n 3652 cm-1 IR-actived n n 1340 cm IR-inactiver n n 1340 cm IR-inactiver O O n t c c3 band position n t c c3 band position O O O O C C H H O H H O H H O H H O HH O HH OBend Asymmetric Stretch Symmetric StretchVibration321n1596 cm-13756 cm-13652 cm-1band positionIR-active IR-active IR-activeinfrared activity Bend Asymmetric Stretch Symmetric StretchVibration321n1596 cm-13756 cm-13652 cm-1band positionIR-active IR-active IR-activeinfrared activity IR-active 666 cm-1(degenerate)2Bending (in plane) Bending (out-of-plane) Asymmetric Stretch Symmetric StretchVibration31n2350 cm-11340 cm-1band positionIR-active IR-inactiveinfrared activity IR-active 666 cm-1(degenerate)2Bending (in plane) Bending (out-of-plane) Asymmetric Stretch Symmetric StretchVibration31n2350 cm-11340 cm-1band positionIR-active IR-inactiveinfrared activity O O C O O C O O C O O C O O C O O O O C C O O C O O O O C C O O C O O C O O O O C C O O C O O O O C C AB water carbon dioxideAdapted from  Figure 1.3 Vibrational modes and IR activity of water vapor (A) and carbon dioxide (B) molecules As molecular structural complexity incr eases, other types of vibrational modes become possible. The methylene group, fo r example is capable of six different vibrational modes as illu strated in figure 1.4. 8
rocking symmetric stretch wagging asymmetric stretchMethylene Normal Modestwisting scissoring rocking symmetric stretch wagging asymmetric stretchMethylene Normal Modestwisting scissoring C C C C C C C C C C C C C C C C C C C C C C C C HHHHHH H H HHHHrocking symmetric stretch wagging asymmetric stretchMethylene Normal Modestwisting scissoring rocking symmetric stretch wagging asymmetric stretchMethylene Normal Modestwisting scissoring C C C C C C C C C C C C C C C C C C C C C C C C HHHHHH H H HHHHAdapted from  Figure 1.4 Vibrational modes of methylene group 1.2.3 Group Frequencies Various chemical functional groups exhi bit specific infrared frequencies representative of their struct ures. Frequencies such as th ese are known as characteristic or group frequencies. Many of the most common functional groups with characteristic group frequencies are familiar organic groups Functional group frequencies allow the spectroscopist to use IR spectra to qualitativel y identify structural elements in samples. Since vibrational frequency absorption prof iles parallel functional group structure, the spectroscopist investig ating biological material using vibrational tec hniques often depends upon existing database s and extensive compilations of spectral information. 9
10 1.3 IR Spectral Feature of Tissues Modern approaches to histology categorize cells into different types based on their primary physiological function. In such a system cells belong to one or more of the following groups: epithelial cells support cells, contractile cells, nerve cells, germ cells, blood cells, immune cells, or hormone-secreting cells. From a molecular point of view, all of these various types of specialized cells encountered in biological tissue are predom inately comprised of four major types of biomolecules or their subunits: proteins, car bohydrates, lipids, and nucleic acids. Additionally, all four of these types of molecules each have a great deal of structural redundancy. That is, they tend to form polym eric molecules based on subunits that while different, reflect structural sim ilarity. For example, thousands of different proteins exist in a typical cell, and while the individual structure of each prot ein is different, they are all made from the same set of amino acids, a nd share a common backbone structure. 1.3.1 Proteins Protein molecules play many fundamental ro les in the life of every cell in addition to serving various important extracellular f unctions in many tissues. The significance of proteins to biological organism s cannot be understated and thei r utility is evident in the many functions they perform including: enzy matic catalysis, transport and storage, coordinated motion, mechanical support, immune protection, generation and transmission of nerve impulses, and control of growth and differentiation. All proteins are formed as linear chains of amino acid building blocks that can form various secondary and tertiary structures. Eukaryotic proteins are typically assembled
from a set of 20 different -amino acids that share a common template and are distinguished by unique side chain structures. Figure 1.5 shows the molecular structure of a typical amino acid. Amino group Carboxylateion Side chain is distinctive for each amino acidRH+H3NCOOC Amino group Carboxylateion Side chain is distinctive for each amino acidRH+H3NCOOCRH+H3NCOOC Figure 1.5 Structure of a typical amino acid All amino acids share a common structure that includes a central or -carbon atom bonded to a carboxyl group, an amino group and a hydrogen atom. At physiologic pH the amino group is protonated (NH 3 + ) and the carboxyl group exists as the carboxylate ion (COO ), displayed in figure 1.5. Each different amino acid contains a distinctive structure at the side chain position designated as R in figure 1.5. The primary protein or polypeptide structure is formed by linking these amino acid subunits together in a linear chain via a condensation reaction between the amino and carboxyl groups of adjacent amino acids in a linear chain. The linkage that is formed between these amino acid subunits is known as a peptide bond and polypeptide chains that result form a repeating backbone structure that is the same for all proteins. Figure 1.6 shows the basic protein primary structure and the locations of these peptide bonds. 11
12 N RAHNCO C HRBCON CRCHCON CPeptide bonds C Amino Acid A Amino Acid B Amino Acid C H H H H H H H H RAHNCO N C HRBCON CRCHCON CPeptide bonds C Amino Acid A Amino Acid A Amino Acid B Amino Acid B Amino Acid C Amino Acid C H H H H H H H H H Figure 1.6 Basic polypeptide structure The polypeptide backbone structure consists of several functional groups, including a C-N group, a C-H group, an NH 2 group, and a carbonyl group (C=O). Since these functional groups repeat for every amino acid in a protein regardless of the proteins identity or higher-order structure, the absorbance bands resulting from these structures dominate the IR spectra of most proteins. The most prominent of these absorbances include; the Amide I absorption near 1650 cm -1 arising from C=O stretching vibrations (80%) weakly coupled to C-N stretching vibrations (20%), the Amide II absorption near 1545 cm -1 arising from N-H bending vibrations (60%) coupled to C-N stretching vibrations (40%), the Amide III absorption near 1236 cm -1 arising from C-N stretching vibrations, and the Amide A absorbance near 3290cm -1 arising from N-H stretching vibrations. In their native states, most proteins do not exist as simple linear polypeptide structures, but instead form complex secondary and tertiary structures that impart a distinct three-dimensionality to a particular protein. The most common protein
secondary structures are the -helix and -pleated sheet configurations depicted in figure 1.7. C C C C O O O O N C N C N C N C H H H H H H H H H R R R R R C C C C O O O O N C N C C N C N C H H H H H H H H H R R R R C C C C O O O O N C N C N C N C H H H H H H H H R R R R C C C C O O O O N C N C C N C N C H H H H H H H H H R R R R N N N N C C C C C C C C C C C C N N H H H H C H H H H H H H H H C R R H N C R R R N C C H R O O O H O O C O N O R H O R O R N N N N C C C C C C C C C C C C N N H H H H C H H H H H H H H C R R H N C R R R N C C H R O O H O O C O N O R H O R O -helix -sheet (antiparallel) R C C C C O O O O N C N C N C N C H H H H H H H H H R R R R R C C C C O O O O N C N C C N C N C H H H H H H H H H R R R R C C C C O O O O N C N C N C N C H H H H H H H H H R R R R R C C C C O O O O N C N C C N C N C H H H H H H H H H R R R R N N N N C C C C C C C C C C C C N N H H H H C H H H H H H H H H C R R H N C R R R N C C H R O O O H O O C O N O R H O R O R N N N N C C C C C C C C C C C C N N H H H H C H H H H H H H H H C R R H N C R R R N C C H R O O O H O O C O N O R H O R O R N N N N C C C C C C C C C C C N N H H H H C H H H H H H H H H C R R H N C R R R N C C H R O O O H O O C O N O R H O R O R N N N N C C C C C C C C C C C C N N H H H H C H H H H H H H H H C R R H N C R R R N C C H R O O O H O O C O N O R H O R O -helix -sheet (antiparallel) Figure 1.7 Common Protein Secondary Structures: -helix and sheet pleated sheet structures can form between pa rallel polypeptide chains, or between strands with antiparallel orientation, as shown in th e figure. The dotted lin es indicate hydrogen bonds. Both of these recurrent secondary stru ctures involve hydrogen bonding between the oxygen atoms of backbone carbonyl groups and the hydrogen atoms of backbone N-H groups indicated in the figure as dotted lines. These structural arra ngements change bond 13
14 angles and other structural parameters, causing frequency shifts of absorbance bands arising from backbone vibrations. As a resu lt, the relationship betw een IR band positions of protein backbone absorbances, most not ably the Amide I absorbance near 1650 cm -1 and protein structure has been the subject of much work over the past decade[12-16]. For example, several studies have examin ed the amide I bands of polypeptides and proteins whose structures are known to be dominated by one of the common secondary structure motifs, such as -helix, sheet, or unordered structures[17-19]. Such studies have led to the development of some empiri cal rules for the correlation of amide I band features and common secondary structural motifs. On the basis of these empirical rules, IR bands in the 1660-1650 cm -1 spectral region are assigned to -helices, 1640-1620 cm -1 to sheets, 1695-1660 cm -1 to -sheets and -turns, and 1650-1640 cm -1 to unordered structures. Such empirical rules are useful guidelines for obtaining structural information from vibrational spectroscopic information, however, many studies show that such rules are not free from shortcomings. For instance, IR studi es of proteins such as myoglobin and hemoglobin, for which x-ray crystallographic data suggests highly helical-structures with no -sheets, have shown Amide I absorbances in the 1640-1620 cm -1 region[21, 22]. While no conclusive evidence exists to explain the presence of such lower-frequency helix amide I bands, some have sugges ted that strong hydrogen bonding of peptide groups with solvent molecules and distortion of helix structures may contribute to such findings[23, 24].
15 1.3.2 Carbohydrates Carbohydrates are aldehyde or ketone compounds with multiple hydroxyl groups. These important biomolecules play three central roles in all organisms: First, they serve as energy stores and metabolic intermediates. Stored glycogen can be readily broken down into glucose, a preferred metabolic fuel. Glucose is broken down to yield adenosine triphosphate (ATP), a phosphorylated sugar derivative and universal currency of energy in the organism. The second im portant role of carbohydrates is as basic structural components of nucleic acids. Ri bose and deoxyribose suga rs are structural units of all nucleotides and ribonucleotid es whose sequence in nucleic acids is responsible for the storage and expression of genetic information. A third important role of carbohydrates in organisms is that they ar e often linked to proteins and lipids on cell membranes, many playing critical roles in cell signaling and recognition[25, 26]. Common cellular carbohydrates have many vibrational spectral features in the fingerprint region of the mid-IR spectrum due to various vibrational modes of C-O, C-C, and carboxylate groups. Infrared spectroscopy has been used extensively to help characterize biologically important polysacch aride cell-surface components, including glycolipids like diacyl sugars, cerebrosides[28, 29], gangliosides[30, 31], lipopolysaccharides[32-34], a nd mucopolysaccharides. 1.3.3 Lipids Lipids form another import ant class of biomolecules f ound in tissue that play many important roles. Like carbohydrates, lipids provide an important source of energy for metabolism. The hydrophobic nature of lipids contributes significantly to their central
16 role in cellular membrane function, pr oviding barriers whic h partition cells and subcellular organelles. Additionally, lipid s perform a variety of other important functions, from the coenzyme roles of fat-so luble vitamins to the regulatory roles of prostaglandins and steroid hormones to stru ctural and functional roles in the nervous system. Lipids all share the charac teristic of having non-polar, hydrophobic domains. In many cases, long chain fatty acids are responsib le for this hydrophobicity, and such lipids have many vibrational modes associated with C-H groups across the fingerprint region of the mid-IR. The spectral frequency region between 3000-2800 cm -1 also contains four prominent absorbance bands common to many lipids: the methyl antisymmetric stretch ( as CH 3 ) at 2962 cm -1 the methyl symmetric stretch ( s CH 3 ) at 2872 cm -1 the antisymmetric CH 2 stretch ( as CH 2 ) between 2936-2916 cm -1 and the symmetric CH 2 stretch ( s CH 2 ) between 2863-2843 cm -1 . Unfortunately, most standard methods fo r the preparation of sectioned tissue involve the use of one or more nonpolar solvents such as et hanol or xylenes that remove lipids from the tissue section[37, 38]. As a tissue source for FT-IR spectroscopic studies, formalin-fixed paraffin-embedded tissue offe rs some advantages over frozen tissue including higher-quality preservation and acce ss to large libraries of preserved tissue, however, paraffin exhibits many of these comm on lipid absorbances, and therefore must be removed from tissue sections intended for spectroscopic analysis. Effective paraffin removal requires the use of strong nonpolar solven ts such as hexane for several hours at temperatures of 40C further contributing to the extracti on of physiologic lipids from paraffin-embedded tissue.
17 1.3.4 Nucleic Acids Nucleic Acids have been studied extensivel y in both purified state as well via model compounds. The most prominent absorban ces reported are due to vibrations of several functional groups on the repeating backbone structur e of nucleic acids. These include absorbances near 1080cm -1 and 1240 cm -1 attributed respectively to the symmetric and asymmetric stretch of phosphodiester (PO 2 ) moieties. However, the ability of IR spectro scopy to attain vibrational info rmation from quiescent nuclear DNA from cell preparations or tissue sections ha s recently been called in to question and some theoretical analyses of chromatin density and packing used to support the idea that nuclear DNA is too dense to produce apprec iable absorbances in transmission IR spectroscopic experiments. 1.4 FTIR Spectroscopy Background Modern instrumental approaches to the collection of spatially-resolved infrared spectroscopic data share many characteristics and all benefit from the extensive advances made in the field of Fourier transform infr ared (FTIR) spectroscopy over the past three decades. Several excellent books[4, 42, 43] ha ve been written on the subject of FT-IR spectroscopy and contain comprehensive in formation on the technology that has been implemented for years in commercial FT-IR spectroscopy systems. Infrared microspectroscopic imaging systems share many common features. Most consist of a research-grade FT-IR spectro meter that provides an output beam of modulated infrared radiation used as a sour ce for an infrared microscope equipped with infrared detectors. Modern approaches to the collection of spatially-resolved spectral
18 data are best differentiated in terms of the type of infrared detection employed. The following sections discuss instrumental aspects of spectrometers and infrared microscopes, as well as strategies for colle cting FT-IR spectroscopi c imaging data with three different types of infrared detection: single-point mapping, raster scanning with linear multichannel detectors, and global FT-IR imaging with Focal Plane Array (FPA) detectors. 1.4.1 FTIR Spectrometers The majority of commercial research-g rade FTIR spectrometers incorporate a broadband infrared source, Mich elson interferometer, sample compartment, and infrared detection with either deuterated triglycine sulfate (DTGS) or mecury cadmium telluride (MCT) single-point detectors. Many commerc ial FTIR instruments exist for dedicated analyses typically implemented in industrial settings for process assessment and quality control analyses. Such spectrometers are ty pically designed to be lower in cost than research-grade spectrometers, which offer more flexibility in the types of measurements that are possible as well as increased sensitiv ity and higher spectral signal-to-noise ratios (SNRs). Figure 1.8 shows the schematic design of the Michelson interferometer, which is the optical portion of the spectrometer that is used to modulate the radiation. The interferometer is composed of two perpendicu lar beam paths often referred to as separate arms of the interferometer. These beampaths intersect at the beamsplitter, an optical component that when placed at 45-degree angle to the normal both reflects and transmits exactly 50% of incident radiation. In the mid-IR region, beamsplitters are typically
constructed from potassium bromide (KBr) with a thin coating of germanium (Ge) or silicon (Si), and many commercial instruments allow beamsplitters to be changed to other materials for coverage of specific spectral regions. Figure 1.8 Michelson Interferometer As depicted in Figure 1.8, polychromatic radiation from an infrared source, typically a ceramic globar, is passed through an aperture to form a beam. This beam strikes the beamsplitter at a 45 angle, dividing the beam in half. Half of the beam is directed at a fixed mirror, while the other half is diverted to a mirror whose displacement can be varied along the axis of the incident beam. After striking these mirrors, the beams in the two arms of the interferometer are sent back to the beamsplitter, where they recombine and interfere with each other. The beamsplitter divides the recombined beam in half again, sending half back toward the source, while the other half is used for spectroscopy and is directed through sample and subsequently detected. 19
20 When the moving mirror occupies a displa cement where the pathlengths in the two arms of the interferometer are equal, then the recombining beams are precisely in-phase and only interfere constructively. This mi rror position produces the most intense beam for every frequency of radiation. As the mi rror moves from this position, a pathlength difference is created in the two arms of the interferometer that causes specific interference patterns for differe nt mirror displacements. If the mirror is continuously scanned, then the intensity of the recombined beam will vary with respect to time in a frequency or wavelength dependent manner. The function of the spectromet er is to encode a modula tion on the polychromatic IR source radiation such that detection of the in tensity of the encoded radiation with respect to time or in the time domain yields spectra l information in the frequency domain. The Fourier transform part of the techniques name refers to the mathematical operation that is required to transform the raw data collected by the instrument in the time domain, known as the interferogram, into a intensity profile in the frequency domain, otherwise know as an infrared spectrum. 1.4.2 Infrared Microscopy Infrared microspectroscopic imaging systems typically couple the modulated output beam of a FTIR spectrometer to an infrared microscope for use as source radiation for obtaining spectroscopic informa tion from microscopic regions of a sample. Infrared microscopes perform similarly to conventiona l optical microscopes and are typically set up to image with visible light along the same optical path. However, they have many structural differences that stem from some fundamental properties of infrared radiation.
21 One major limitation of infrared spectrosc opy is related to its exceptional molecular sensitivity. As mentioned in section 1.2, all covalently bonded molecules, with the exception of homonuclear diatomics, absorb in frared radiation. Op tical components used in conventional microscopes ar e composed almost exclusivel y of borosilicate glass or quartz, both of which have broad absorbances over much of the infrared spectrum. For this reason, infrared microscopes are designe d to use reflective optics wherever possible, and refractive optics have to be manufactured from alternative materials, such as halide salts, which are transparent over the sp ectral regions of interest. Most Infrared microscopes use Cassegrai n condenser and objec tive lenses and can be operated in either transmi ssion or reflectance modes. In reflectance mode, one side of the Cassegrain objective primary mi rror is typically used to di rect the radiation onto the sample while the opposite portion of the primar y mirror is used to collect the reflected radiation. Infrared microscopes are often outfitted with automated high-precision motorized mapping stages, which permit the sample to be positioned precisely in the plane perpendicular to the op tical path. Most microscope s incorporate a visible light source and detection system, typically a video camera. Adjustable mirrors are used to switch between visible and infrared modes and some models incorporate a beamsplitter to allow for simultaneous imaging in both spectral regions. The different strategies that can be employed to collect spatially-resolved infrared microspectroscopic data depend on the types of infrared detection systems available of the microscope. Panels A-C of Figure 1. 9 depict three different approaches based respectively on single-point, linear-array, and focal plane array (FPA) detection. A discussion of each approach follows.
Sample Aperture Aperture Single Element Infrared DetectorMicroscope CCD Visible Detector Turning Mirror Visible Light Source Precision Stage Rapid-Scan Interferometer Sample Mulichannel Infrared DetectorMicroscope CCD Visible Detector Turning Mirror Visible Light Source Precision Stage Rapid-Scan Interferometer Sample Multichannel Infrared DetectorMicroscope CCD Visible Detector Turning Mirror Visible Light Source Microscope Stage Rapid-or Step-Scan Interferometer Focal Plane Array Detector AB C Sample Aperture Aperture Single Element Infrared DetectorMicroscope CCD Visible Detector Turning Mirror Visible Light Source Precision Stage Rapid-Scan Interferometer Sample Mulichannel Infrared DetectorMicroscope CCD Visible Detector Turning Mirror Visible Light Source Precision Stage Rapid-Scan Interferometer Sample Multichannel Infrared DetectorMicroscope CCD Visible Detector Turning Mirror Visible Light Source Microscope Stage Rapid-or Step-Scan Interferometer Focal Plane Array Detector AB C Figure 1.9 Three Instrumental Approaches for collection of spatially resolved FTIR spectroscopic data A) Point-mapping using single element dete ction; B) Raster-Scan imaging using linear multichannel detection; and C) Global FT-IR imaging using 2-D focal plane 1.4.3 Mapping with Single-Point Detectors In single element microspectroscopic instru mentation, spectral information from a small, specified area of the sample is obtai ned by restricting the area illuminated by the infrared beam using opaque apertures of cont rolled size. The collected radiation is then diverted to a sensitive detector. To identify the area to be examined, however, a corresponding white light optical image is also required. Clearly, focusing the infrared 22
23 beam for maximal throughput and minimal disp ersion in the sample plane requires the optical and infrared paths be parfocal and collinear. By restricting the infrared beam to a small spatial area of the sample, and sequentially moving to different regularly-spaced sample locations with a high precision microscope stage, spatially-resolved spectroscopic data from large sample areas can be mapped out point by point. This strategy, ofte n referred to as point-mapping, suffers from several limitations. The cross-sectional diameter of the beams used in such infrared microscopes must be large enough to fully illuminate the area pa ssed by the largest apertu re setting that may be employed, for example a 100x100 um square. There is a tradeoff between the spatial resolution of mapping data that can be acqui red and corresponding th roughput due to the need to block out more and more of the avai lable radiation. Aperture use decreases the instrumental throughput due to diffraction when the aperture is of the same dimension as the wavelength of light (~3-14um), thus lim iting the highest achiev able data spatial resolution. Apertures also permit the passage of some diffracted light from outside the apertured region. The use of a second set of ap ertures in tandem to reject stray radiation can improve spatial fidelity, unfortunately at the cost of additiona l throughput loss. Throughput is important because it directly af fects the spectral si gnal to noise ratio (SNR), and losses in throughput require larger acquisition time s for signal recovery. Data acquisition time is the major drawback to single-point mapping approaches. Spectral information is acquired for each spa tial location in the final map one-by-one and there is significant time overhead for moving the sample to each new sampling location.
24 1.4.4 Raster-scan Imaging Using Multichannel Detectors While single element microspectroscopy provi des the capability to obtain spectra from small spatial regions, poor SNR characte ristics, diffraction e ffects and stray light issues resulting from the use of apertures li mit the applicability of this point mapping approach. A multichannel detection approach to circumvent some of these issues has recently been implemented with a linear array detector employed to image an area corresponding to a rectangular spatial area on the sample. The sample stage is moved precisely to sequentially image a selected spa tial area on the sample. This data collection strategy is referred to as push-broom mapping or raster scanning. The process is conceptually similar to point-by-point mapping but takes advantage of the multiple channels of detection. Hence, imaging a large sample area is faster by a factor of n, for a linear array detector containing n elements. The instrument is schematically displayed in Figure 1.9B. Point mapping detectors are typically 100 250 m in size; in contrast, an individual detection element in a linear array detector is of th e order of tens of micrometers. Employing a linear array eliminates the need for apertures, as small detector elements directly image different sample spatial regions. For example, a detector element 25 m in size can be operated at 1:1 magnification or 4:1 magnification to provide a 25 m or a 6.25 m effective pixel size with available, relatively aberrationfree infrared optics. This approach circumvents the debilitating diffraction effects resulting from the use of small apertures in single channel detection systems and provides higher quality data when desired spatial reso lutions approach the wavelengths of light being used. In addition, th e spatial resolution, data quali ty, and time for data acquisition
25 are no longer coupled as in point mapping methods. The data acquisition time depends solely on the size of the image a nd quality of data desired, a nd is correlated less with the spatial resolution, which is determined by the employed optics. A high-precision, motorized stage that re producibly steps in small increments is used and the interferometer is operated in a continuous scan mode. In combination with high performance multichannel detectors, this mode combines high performance multichannel detectors with the most desirable properties of rapid-scan interferometry to yield high quality spectroscopic imaging data. 1.4.5 Global FTIR Spectroscopic Imaging The state of the art in FT IR microspectroscopic imag ing instrumentation is the combination of an infrared microscope equi pped with a focal plane array (FPA) detector and an FTIR spectrometer[47, 48], as s hown in Figure 9C. FPA detectors are constructed of thousands of i ndividual detection elements la id out in a two-dimensional grid pattern. An FPA matched to the characte ristics of the optical system is capable of imaging the entire field of view afforded by the optics and of utilizing a large fraction of the infrared radiation spot size at the plane of the sample. The increase in the number of individual detectors with respect to a lin ear array provides a correspondingly larger multichannel advantage. For example, an FPA with pixel dimensions p x p, provides a p 2 time savings relative to a single element detector and a p 2 /n time savings compared to a linear array detector containing n elements. For a 128 x 128 element FPA detector relative to the single element cas e, the advantage is a factor of 16,384, while compared to a 16-element linear array detector; the multicha nnel advantage is a factor of 2048. FPA
26 detectors are also capable of imaging large spatial areas simultaneously without inherent inefficiencies of moving the sample or re-set ting the interferometer to scan a different area. The considerable reduction in data acq uisition times allows for imaging large areas, as well as the examination of dynamic proc esses in a single field of view. The first and, to date, most popular appro ach to FTIR micro-imaging spectrometers incorporates a step-scan interferometer . While conti nuous or rapid-scan spectrometry involves scanning th e moving mirror at a consta nt velocity, a step-scan interferometer is capable of stepping th e moving mirror to discrete, evenly-spaced intervals and maintaining individual mirror posi tions with very little displacement error. A constant retardation over an extended time period allows suitable time for signal averaging and for data readout and storage. Short time delays prio r to data acquisition are necessary for mirror stabiliza tion at the onset of the step. Detector signal is integrated for only a fraction of the total time required fo r collection of each frame. The integration time, number of frames co-added, and number of interferometer retardation steps (a function of desired spectral resolution) determine the total time required for the experiment. Since the integra tion time determines the data quality, efforts have been made to increase the ratio of the integrati on time to the total data acquisition time. Imaging configurations that utilize a rapid scan interferometer have been proposed for small arrays. Slow data readout and storage rates for many FPA detectors preclude conventional rapid-scan mirror velocities, thus appro aches must make use of so called slow-scan mirror velocities of 0.01 cm/s. A generalized data acquisition scheme that permits true rapid scan data acquisit ion for FPA detectors has been proposed, where the integration time of individual frames collect ed by the FPA detector is
27 negligible with respect to the complete interferogram acquisition. For most FPA detectors available today, the motion of the moving mirror does not allow co-addition of frames at individual retardat ions in the continuous sca nning mode, but successive singleframe acquisitions can be averaged to increase data SNRs. Compared to step-scan data acquisition, rapid scan data collection (mirror velocity > 0.025 cm/s) allows for fast interferogram capture as no time is spent on mi rror stabilization. The error arising from the deviation in mirror position during frame collection is hypothesized to be the next largest contributor of noise compared to th e dominant contribution from random detector noise. At present, the advantages of c ontinuous-scan relative to step-scan approaches are a decreased cost of instrumentation and an increased data collection efficiency. 1.5 Spectroscopic Imaging: Data Structure and Applications Spectroscopic imaging data, regardless of its method of collection, can be conceptualized as an image cube with two dimensions corresponding to the spatial axes of the sample and the third dimension to the spectral frequency or wavelength. Digital image data is represented as a collection of rectangular picture elements or pixels, each with an associated brightness value or magnitude. Spectroscopic image data can be thought of as a collection of super-imposable and spectrally consecutive image planes, whose pixel values consist of the spatia lly independent absorbance at the spectral frequency or wavelength specified by the image plane. Alternatively, the data structure can be conceptualized to consist of individu al spatial locations or pixels each with an associated absorbance spectrum. The concept of the image cube is represented schematically in figure 1.10.
x y Wavelength AxisSpatial Axes Figure 1.10 Schematic representation of the image cube These alternative views of the data structure influence the type of information that can be extracted from the data. For example, we can specify distinct spatial locations in a spectroscopic image, and display the associated spectra for simultaneous comparison of absorption features across the full spectral region collected. Alternatively we can specify a particular absorption feature of interest and display the associated spectral image plane. The brightness values of pixels in such an image will correspond to the samples spatial distribution of the species responsible for the absorption at the associated spectral frequency. FTIR imaging of biological systems has demonstrated a potential to complement other imaging approaches. For biomedical applications, the technique may be used to 28
29 examine chemical changes due to pathological abnormalities and to follow histological alterations with high accuracy. Non-destruct ive morphological visualization of chemical composition rapidly provides structural and spatial information at an unprecedented level. Specifically, thousands of spect ra routinely acquired in an imaging experiment may be employed for statistically mean ingful data analyses, which in the example of biological tissue samples may prove ultimately useful in medical diagnoses. Si nce the visualization contrast is dictated by inherent chemical a nd molecular properties, no sample treatments, such as histopathological staining techniques required for optical microscopy, are necessary. A typical example of the type of tissue information that can be retrieved was demonstrated by examining monkey cerebellu m sections. Distributions of lipid relative to protein allowed eas y differentiation of white and gray matter areas. Purkinje cells in rat cerebella, which strongly in fluence motor coordination and memory processes, were visualized using FTIR imaging techniques[55, 56]. Neuropathologic effects of a genetic lipid storage disease, Niemann-Pick type C (NPC), were distinguishable on the basis of spectral data without the use of external histological staining. Statistical analysis provided a numerical confirmation of these determinations consistent with a significant demyelination wi thin the cerebellum of the NPC mouse. IR spectroscopy has been used for a number of y ears to characterize mineralized structures in living organisms (notably, bone). FTIR im aging spectroscopy[58, 59] of bone allows spatial variations of a number of chemical components to be non-destructively monitored. Correlations in bone between FTIR imaging and optical microscopy involving chemical composition, regional morphologies and the de velopmental processes have been made,
30 and an index of crystallinity/bone maturity could be determined providing structural information in a non-destructive manner. 1.5.1 Image Classification Methods One of the most useful approaches to extr acting data from such data structures is the process of image classifica tion. Image classification algorithms automatically assign each pixel in an image scene to a specific class or group based on its spectral properties or pattern. Unsupervised Classification refers to the automatic part itioning of pixels into classes of spectral similarity without the use of any class training data. Supervised Classification is the process of classifying pixels into specific cla sses based on their spectral similarity to user-supplie d training data for each class. Unsupervised classification methods have the advantage that no extensive prior knowledge of the image scene is necessary a nd the potential for hu man error is far less than with supervised methods. Additionally, they are useful for fi nding natural spectral patterns and groups in spectral images. Howeve r, they are limited in their usefulness by the need to identify the resulti ng classes after the classificatio n is performed. For this reason, such unsupervised methods are of lit tle usefulness for diagnostic implementation. Supervised classification methods have several advantages relative to unsupervised strategies. First, the analyst has control over the specific number and identity of class categories and can tailor them for specific tasks. Supervised classification is tied to areas of known identity, determined through the process of selecting training regions. Additionally, regions of training data can be used during the process of classifier development to evaluate classifier performa nce. While inaccurate classification of
31 training data indicates serious classification problems and/or problems with training data selection, accurate classifica tion of training data does not always assure accurate classification of other image data. Supervised image classification methods ha ve several disadvantages and limitations as well. By creating classes and assigni ng training populations, the analyst imposes a classification structure on the data. If the user-defined class structure does not match the natural class structure within the data, the classes may not be distin ct or well defined in multidimensional space. Training populations that do not accurately represent the natural distribution of values within a class may resu lt in severe classification error. Finally, classes unknown to the analyst and not incl uded in the training data may also be misclassified and thereby remain undiscovered. 1.6 Prostate Background 1.6.1 Anatomy and Histology In men, the prostate is a retroperitoneal gland located just below the bladder that surrounds the urethra. The gland is divide d into four zones: peripheral, central, transitional, and periurethral as shown in Figure 3.1. Dis tinctions between these zones are important because prolifer ative lesions vary according to the zone in which they occur. For instance, nodular hyperplasia, al so known as benign pr ostatic hypertrophy or hyperplasia (BPH), occurs predominantly in the central zone, whereas most adenocarcinomas occur in the peripheral zone.
adapted from  Figure 1.11 Zonal Anatomy of the Prostate Histologically, the prostate is a compound tubuloalveolar gland in which glandular spaces are lined by epithelium. Specifically, the gland is lined by a layer of low cuboidal epithelium at the basal surface, which is covered by a layer of columnar mucus-secreting cells. The glands contain a discrete basement membrane and are separated by abundant fibromuscular stroma. Some ducts in the gland are lined by tall columnar epithelium, but as they approach the urethra, the epithelium changes to more cuboidal and eventually into the transitional epithelium that lines the urethra and urinary bladder. 32
33 While prostatic epithelial tissue and fibromuscular stroma make up the bulk of the gland, there are several other important hist ological features seen in the prostate. Numerous blood vessels run thr oughout the prostate, as well as peripheral nervous tissue innervating the gland. Prostates from ol der men frequently contain small, spherical corpora amylacea comprosed primarily of c ondensed glycoprotein in the glandular lumina. 1.6.2 Prostate Pathology 22.214.171.124 Incidence Prostatic carcinoma is the most common form of cancer in men and it is estimated that 221,000 new cases will be diagnosed in the United States in 2003. The incidence of newly diagnosed cases of prostate cancer in the US was 100,000 in 1988, and has risen steadily since then to just under 200,000 in 1994. Mortality in the US due to prostate cancer rose from 28,000 to 36,000 during the same time period, however recent evidence suggests that mortality has peaked and may be falling. The estimated mortality for US men in 2003 is 29,000. This decline has been attributed to increased screening efforts and active treatment of localized disease by radiation and radical prostatectomy. 126.96.36.199 Latent Prostate Cancer In 1954, Franks observed an extraordinarily high prevalence of microscopic foci of what he termed latent prostate cancer during autopsy of men who died from other diseases. His observations have been corroborated by several investigators[71, 72]
34 and the occurrence of these incidental can cers has been shown to increase with age affecting approximately 20% of men in their 20s, 30% of men in their 50s, and 70% of men in their 80s. The lifetime chance th at a man will develop clinically apparent prostate cancer is less than 10%, thus the majority of these tiny cancers detected at autopsy are clinically in significant. While it is clear that early diagnosis and treatment of prostate adenocarcinoma leads to an impr oved mortality and morbidity, these findings point out the importance of being able to diffe rentiate potentially dangerous cancers from the very small, well-differentiated, slow-g rowing lesions which are unlikely to present clinically during the patie nts natural lifespan. 188.8.131.52 Etiology and Risk Factors It has become clear that genetics play a significant role in the pathogenesis of prostate adenocarcinoma. Male relatives of men who have died from prostate cancer have a greater-than-expected incidence of the disease. An early study by Woolf of 228 men dying of prostate cancer found the relative n early 3-fold increase in the relative risk of first-degree relatives compared to a c ontrol group. Subsequent studies have confirmed this familial association[76-78], and demonstrated the importance of screening PSA values in asymptomatic men from families with 3 or more members affected by prostate cancer[74, 79]. Recent evidence supports the existence of a genuinely hereditary form of early onset prostate cancer exhibiting Mendelian autosomal dominant inheritance. The exact gene defects have not been elucidated for these families but possible locations have been mapped to chromosome 1q24-25 as well as the X chromosome suggesting the
35 possibility of X-linked inheritance. R ecent evidence suggest that mutations in the tumor suppressor genes BRCA-1[ 83] and BRCA-2[84, 85] confer increased risk of developing prostatic adenocarci noma, and attempts to screen for those at risk are currently being studied. The most influential factor conferring risk of developing prostate cancer besides familial inheritance is age. African-American men have roughly twice the lifetime risk of their wh ite counterparts and higher PSA and tumor volume in a study adjusted for age, stage, pa thologic stage, Gleason score, and volume of benign disease. Other predisposing factors for clinical pr ostate cancer include the presence of testosterone and dihydrotestosterone (DHT), sexual history positive for early first sexual experience and multiple sexual partners, a diet high in saturated animal fat and low in yellow and green vegetables, and environmental or occupational exposure to several pollutants including ca dmium and the radioactive agents 51 Cr, 59 Fe, 60 Co, and 65 Zn. Vasectomy has been suggested as a possible risk conferring event[92-94] though some studies failed to demo nstrate a conclusive link[95, 96]. 184.108.40.206 Diagnosis 220.127.116.11.1 Clinical Presentation With the recent widespread increase of PSA testing in men at risk for prostate cancer, a large proportion of patients presenting with the disease are asymptomatic. Clinically apparent prostate cancer presents with a spectrum of symp toms related to the extent of disease progression. Urinary symp toms occur in localized as well as advanced disease states as well as in extremely co mmon condition of benign prostatic hyperplasia
36 (BPH). Symptoms related to bladder outflow obstruction, such as hesitancy, poor stream, and a sensation of incomplete voiding arise from urethral occlusion by the tumor or nodular mass. Urinary frequency and urgency are irritative symptoms th at develop due to detrusor muscle instability s econdary to outflow obstruction or directly by tumor invasion of the trigone of the bladde r and pelvic nerves. Invasi ve cancer can produce other symptoms both locally and at distant sites. Local extension of prostate cancer can present with hematuria and/or hemospermia due to i nvasion of the prostatic urethra or seminal vesicles. Direct inva sion of the distal urinary sphincter can cause urinary symptoms unrelated to outflow obstruction, while si milar invasion of the neurovascular bundles posteriorly can lead to erectile dysfunction and pain. Significant posterior invasion of prostate cancer can produce lower bowel symptoms including rectal bleeding and constipation due to large intestine obstruction near the rectum. Symptoms that indicate local metastatic disease include bone pain, paraplegia due to cord compression, lymph node enlargement, lower limb lymphedema, an d loin pain while lethargy, cachexia, and hemorrhage may indicate significant systemic metastases. 18.104.22.168.2 Digital Rectal Examination (DRE) Digital rectal examination (DRE) is an inexpensive method of prostate cancer detection which has been the focus of many clinical studies[98-103]. One problem with the test is that it is subjective and cons equently depends on the experience of the examiner. Another is that several other conditions can lead to a false-positive DRE finding, including BPH, prostatitis, prostatic calculi, ejaculatory duct anomaly, seminal vesicle anomaly, and rectal wall phlebolith or polyp/tumor. Early stages of prostate
37 cancer (T2a) are characterized by a firm peripheral nodule that does not distort the capsule, while more advanced cancers feel hard and more diffuse. T3 stage tumors often present an altered prostate contour while re taining movement of the gland as a whole contrasted with the fixed, immobile presentation of T4 stage tumors. 22.214.171.124.3 Prostate Specific Antigen (PSA) Prostate-specific antigen is a 34 kD gl ycoprotein specifically found in prostate epithelium. It is a neutral serine protease designed to lyse semina l-vesicle protein. A small percentage of PSA normally escapes th e prostatic ducts and enters the bloodstream where it exists bound mainly to the proteins alpha-1-antichymotrypsin (ACT) and alpha macroglobulin (MG), leaving a small proportion of free PSA in the serum. Prostatespecific antigen has established utility fo r the immunohistochemical identification of metastatic disease of prostatic origin, for monitoring of biochemical recurrence after therapy and for assessment of disease status in men who are at high risk for biopsy complications. Screening measures for serum PSA levels ha ve increased the detection rate of earlystage prostate cancer and are thought to be in part responsible for the downward stage migration trend seen in the disease. Consider able variability exists in the world of PSA testing. The cutoff for normal total PSA is accepted to be 4.0 ng/mL though some evidence suggests lowering this cu toff in at risk populations. While most clinical assays measure total PSA (bound + free) a significant ad vantage is afforded when an additional test for free PSA is performed. Strong evid ence exists that PSA complexed with ACT increases in prostatic carcinoma[104, 105] and the lack of availab ility of a test to
38 specifically measure serum ACT-complexed PSA led to the use of percent free-to-total PSA ratio to approximate complexed PSA. Such ratios proved to be especially useful in the population of men with total PSA values in the gray zone of 2.5 to 10 ng/ml. Recent development of a reli able assay for ACT-PSA complex looks promising and may outperform both total PSA and free-to-total PSA ratio as a more specific analyte for cancer. Other methods to improve PSA performance that have been studied include PSA density[110, 111], transitional zone density, PSA velocity[113, 114], and ag e-specific PSA. 126.96.36.199.4 Diagnostic Imaging Transrectal ultrasound imaging (TRUS) produces high-resolution images of the prostate which are useful for assessing extent of tumor involvement and extension as well for guiding needle biopsies to sample areas su spected of harboring tumor foci. Prostate cancers are frequently hypoechoic on TRUS, but can also be isoechoic and more rarely hyperechoic. Characteristics of prostate cancer that can be evaluated by TRUS include asymmetry of prostate size, shape, indefinite differentiation between the central and peripheral zones, and bul ging or disruption of the cap sule. Advances in color Doppler TRUS allowing analysis of a bnormal blood flow look promising for the identification of hypervascular regions in the peripheral zone . Computed tomography (CT) scanning is useful in metast atic disease to identify the presence of lymphadenopathy in the pelvis and is suggested only when ot her factors identify risk of tumor spread (i.e. PSA>20ng/mL and Gleason gr ade > 7). Advances in Magnetic resonance (MR) imaging endorectal coil de sign have allowed the acquisition of
39 high-resolution differentially weighted MR images of prostatic disease that are probably the most accurate technique currently avai lable for assessing the extent of tumor involvement. Additionally, dynamic contrast enhanced MR imaging may provide tumor angiogenesis information. 188.8.131.52 Biopsy Interpretation and Grading of Prostatic Adenocarcinoma The definitive diagnosis of prostatic ad enocarcinoma involves the cytological and histological confirmation of the established criteria of malignancy. The diagnostic criteria for carcinomas in biopsies of the pros tate involve both arch itectural and cytologic findings. Low to medium power analysis of the arrangement of the glandular acini is useful and is the basis of the Gleason scale for grading prostatic adenocarcinoma, the predominant scoring system used in the Unite d States. Malignant acini are typically scattered haphazardly in the stroma either sing ly or in clusters. The acini in cancer are typically small to medium sized with contour s that are less smooth than adjacent normal and hyperplastic acini. Cytologic abnormalitie s in adenocarcinoma include nuclear and nucleolar enlargement present in a majority of malignant cells. Nucleolar size greater than 1.5 mm suggests malignancy while identifica tion of two or more nucleoli in a single cell is virtually diagnostic of malignancy. 184.108.40.206.1 Gleason Grading System The Gleason Grading system is the most widely used system for grading prostatic adenocarcinoma. It relies heavily on the exam ination of low power architectural features of the arrangement of prosta tic acini. The Gleason scal e rates glandular patterns of proliferation on a scale of 1 (most differentiated) to 5 (least differentiated). Most prostate
40 cancers contain more than one of these patter ns and thus the Gleason score for a biopsy interpretation is reported as the combination of the two most prominent patterns. Scores range from 2-10 and should be reported as the composite scor e and its component patterns with the most prevalent pattern list ed first. For example a biopsy sample with a predominant pattern of 3 and a sec ondary pattern of two would be reported as 3+2=5. In practice most cancers have at least one score of 3, and the score of 1 is rarely used. Gleason grade 1 architecture is describe d as very well differentiated and is minimally distorted. Neoplastic glands ar e round, closely packed, single, separate, uniform in shape and diameter, and are shar ply delineated from fi brovascular stroma. Hyperplastic glands also fulfill these crit eria, therefore a classification as grade 1 adenocarcinoma also requires occasional en larged nucleoli > 1mm in diameter. In practice a Gleason score of 1 is rarely used. Gleason grade 2 pattern (well differentiated) consists of glands which still exhibit a m ild but definite stromal separation between glands with more variation in the shape and size of glands than is s een in grade 1, but less than that of grade 3. Grade 2 tumors remain circumscribed, and definite separation of the malignant glands exists at the tumor peri phery suggesting ability to spread to the surrounding stroma. Tumor gland separation is usually less than one average gland diameter. Gleason grade 3 cancers exhibit more extreme variation in size, shape, and separation than grade 2 and are typically sp aced more than one average gland diameter apart. The cytoplasm of grade 3 tumor cells tends to be more basophilic than lower grade cancers and nuclei are variable but still larg er than lower grades and almost always contain prominent nucleoli. Gleason grade 4 cancers may exhibit any of 4 different
41 morphologic patterns. Glands with a cribiform pa ttern have large masses of tumor cells punctuated by sieve-like spaces. Such a patt ern was classified as grade 3 by Gleason, however, subsequent reclassification to grade 4 was based on the conclusion that most, if not all examples of cribiform carcinoma are equivalent to grade 4 carcinoma growing within preexisting lumina. The distinctive feature of grade 4 tumors is ragged and invading edges in contrast to the smooth edges of grade 3. Other arch itectural variants of grade 4 adenocarcinoma include solid, micro acinar, and papillary. Gleason grade 5 tumors completely lack glandular differentiati on. Such tumors can be arranged in solid masses, cords, trabeculae, sheets, or may app ear as single cells infiltrating the stroma. 220.127.116.11.2 Importance of Histologic Grading Cancer grade at time of diagnosis has b een investigated extensively for correlations with other tumor characteristics and clinical behavior. Every measure of survival and recurrence is strongly correlated with cancer grade. These measures include crude survival, tumor-free survival after treatment, metastasis-free survival, and cause-specific survival. Such correlation has been desc ribed and validated in numerous studies. Age-adjusted, fifteen-year, cancer-speci fic mortality rates for men with Gleason scores of 2 -4, 5, 6, 7 and 8-10 ar e 4-7%, 6-11%, 18-30% 42-70%, and 60-87% respectively. Tumor volume has been co rrelated with histol ogic grade in both transurethral and radical pr ostatectomy specimens. A study by McNeal showed that in Gleason grade 4 and 5 tumors, 22 of 38 tumors >3.2 cm 3 had tumor-positive nodes while positive nodes were present in only 1 out of 171 tumors <3.2 cm 3 Two studies
independently confirmed that the strongest predictor of progression of poorly differentiated cancer is tumor volume[129, 131]. Other studies have found correlations between Gleason grade and PSA levels. Gleason grade is also one of the strongest and most useful predictors of pathologic stage in many studies including the progression of capsular perforation, seminal vesicle invasion, and lymph node and bone metastases and can be correlated with expression levels of MIB-1 (Ki-57), a tissue marker for proliferation[133-136]. 18.104.22.168 Staging of Prostatic Adenocarcinoma Accurate assessment of the clinical stage of prostatic adenocarcinoma is important for the estimation of prognosis, selection of treatment, and evaluation of therapeutic results. The Tumor Node Metastasis (TNM) staging system is used to stage prostatic adenocarcinoma. The current TNM clinical staging is shown below in tables 1.2 and 1.3. TX Primary tumor cannot be assessed T0 No evidence of primary tumor T1 Clinically inapparent tumor not palpable or visible by imaging T1a Tumor incidental histological finding in 5% or less of tissue resected T1b Tumor incidental histological finding in more than 5% of tissue resected T1c Tumor identified by needle biopsy. Nonpalpable, not visible in imaging. T2 Tumor confined within the prostate T2a Tumor involves one lobe T2b Tumor involves both lobes T3 Tumor extends through the prostate capsules T3a Unilateral extracapsular extension T3b Bilateral extracapsular extension T3c Tumor invades the seminal vesicle(s) T4 Tumor invades any of bladder neck, external sphincter, or rectum T4a Tumor invades any of bladder neck, external sphincter, or rectum T4b Tumor invades levator muscles and/or the pelvic wall adapted from  Table 1.2 Staging of primary tumor (T) 42
NX Regional lymph nodes cannot be assessed N0 No regional lymph node metastasis N1 Metastasis in regional node(s) adapted from  Table 1.3 Staging of regional lymph node involvement (N) 43
44 Chapter Two Methods 2.1 Tissue microarrays Tissue microarray technology provides a plat form for the high throughput analysis of tissue speciemens in research. They are used for the target verification of cDNA microarray results, expression profiling of tumors and tissues, as well as epidemiology based investigations. Well-design ed tissue arrays reduc e the variability of experiments performed in a repetitive fashion on large populations, and provide consistent sample-to-sample preparation. There are currently no reported studies a pplying vibrational sp ectroscopic imaging techniques to the analysis of tissue microa rray specimens. The tissue microarray is an attractive sample platform for pathological spectroscopic imaging approaches for several reasons. First, tissue arrays can be construc ted from archival material, allowing for large sample populations representative of normal tissu e and disease processes to be examined. Second, tissue microarrays provide consistent sample preparation across a large sample population, minimizing sample-to-sample data vari ation. Finally, serial sections of tissue microarrays can be analyzed with othe r techniques to provide complementary information invaluable to the interpretation of spectroscopic imaging results. 2.1.1 Construction of Prostate Tissue Microarrays Sections from three prostate tissue micr oarrays constructed in the Tissue Array Research Program Laboratory, Laboratory of Pathology, Center for Cancer Research, of
45 the National Cancer Institute by Dr. Stephen M. Hewitt were used as samples for the experiments in this study. The tissue arra y donor material was obtained from formalinfixed paraffin-embedded blocks from radical prostatectomy specimens taken from cases of confirmed prostate ade nocarcinoma from specimens obtained from the Cooperative Human Tissue Network (CHTN) with approval of the appropriate Institutional Review Boards or Office of Human Res earch Subjects. The tissue arra ys were constructed with a 0.6 mm needles. The arrays were constructed using a Beecher Instruments (Silver Spring, MD) ATA-27 Automated Tissue Arrayer. For sake of clarity, the arrays will be referred to by the respective patient populations used for their construction. Speci fic details regarding th e layout of Array P16, Array P-40 and Array P-80 appe ar in the sections below. 2.1.2 Array P-16 Design Array P-16 was constructed using donor tissue from a population of 16 patients with confirmed prostate adenocarcinoma. Eight unmapped 0.6 mm cores from each patient were used for a maximum spot number of 128 spots/section. Donor core locations were determined by examination of H&E stained sections of the donor blocks and were chosen to provide a representative sampling of both normal prostate histology and pathology from each patient. 2.1.3 Array P-40 Design Array P-40 was constructed from donor tis sue from a population of 40 patients that included the set 16 patients used in the cons truction of Array P-16. Five unmapped 0.6 mm cores from each of the forty patients were used for a maximum spot number of 200
46 spots/section. Donor core locations were c hosen from locations representative of both adenocarcinoma and benign epithelium. 2.1.4 Array P-80 Design Array P-80 was constructed of donor tissu e from a population of 79 patients with confirmed adenocarcinoma. Two mapped 0.6 mm cores were used from each patient for a maximum spot number of 160 spots/section. H&E-stained secti ons of the donor tissue blocks were used as a guide to carefully se lect tissue from a region of adenocarcinoma for one core and benign epithelium for the co rresponding core. Figure 2.1 below contains an image of an H&E stained section of Array P-80 and a corresponding schematic representation of the core layout.
47 1917 15 1413121 20 20 19 1 18 1 17 1 16 16 15 1 14 14 13 1 12 1 11 11 4039383736 353433321 3029282726 2524232221 409736 4 1 2922726 242222 6059587 56 55545352 51 0494847 46 45444342 41 687 5 5 50987 6 44442 41 8079 78 77 76 7574 73 2 71 7069 8 67 66 6564 63 62 61 07 7 7 75 74 7 72 1 70 6 7 6 564 3 61 4039 38 3736 3534 33 3231 3029 28 2726 2524 23 2221 403 8 7 421 8 224 2 222 059585756 55 5453251 5049484746 454443424 5 4521 5094474 4444441 8079 78 7776 7574 3 7271 7069 68 676 656 63 6261 8079 7 776 64 6 6 a h An n il 2019181716 1514131211 10 9876 54321 2019181716 1514131211 10 9876 54321 20 20 19 19 18 18 17 17 16 16 15 15 14 14 13 13 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 20 20 19 19 18 18 17 17 16 16 15 15 14 14 13 13 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 4039383736 3534333231 3029282726 2524232221 4039383736 3534333231 3029282726 2524232221 60595857 56 55545352 51 50494847 46 45444342 41 60595857 56 55545352 51 50494847 46 45444342 41 8079 78 77 76 7574 73 72 71 7069 68 67 66 6564 63 62 61 8079 78 77 76 7574 73 72 71 7069 68 67 66 6564 63 62 61 4039 38 3736 3534 33 3231 3029 28 2726 2524 23 2221 4039 38 3736 3534 33 3231 3029 28 2726 2524 23 2221 6059585756 5554535251 5049484746 4544434241 6059585756 5554535251 5049484746 4544434241 8079 78 7776 7574 73 7271 7069 68 6766 6564 63 6261 8079 78 7776 7574 73 7271 7069 68 6766 6564 63 6261 Adenocarcinoma Cores Benign Epithelium Cores Adenocarcinoma Cores Benign Epithelium Cores 600 m Figure 2.1 Array P-80 Layout. The right panel contains a visible optical image of an H&E stained section of array P-80. A schematic representation of the core layout appears on the left with patient numbers. 2.2 Tissue Array Section preparation 2.2.1 Optical Substrates for Tissue Array Sections Standard optical materials, such as thos e found in microscope slides, are generally composed of glass, quartz or fused silica. These materials all absorb radiation in the infrared region at wavelengths longer than 2 m. For this reason, transmission experiments in the mid-IR require the use of alternative optical materials. Several different halide salts are commonly used as optical materials for IR spectroscopy and each possess different optical and physical properties.
48 Tissue array sections intended for IR im aging experiments were mounted on 3 mmthick, polished, barium fluoride (BaF 2 ) optical windows. Barium fluoride is transparent from 0.15-12.5 m, which covers the visible and the entire spectral range of the FT-IR instrument. Additionally, BaF 2 optical elements have the lowest solubility in water (0. 17gm/100gm water at 23 C) of materials with similar optical characteristics. 2.2.2 Deparaffinization Histology grade, low melt (58-62 C) pa raffin was removed from tissue array sections by covering the tissue surface with hexane for 5 minutes. The samples were rinsed with hexane several times and depa raffinization was continued by immersion in hexane at 40C with continuous stirri ng for 48 hours. Every 3-4 hours during the deparaffinization process, the immersion vessel was emptied, rinsed thoroughly with acetone followed by hexane. Once dry, the vessel was filled with fresh neat hexane to promote flow of embedded paraffin from the tissue. Thorough deparaffinization was assured by monitoring the disappear ance of the paraffin band at 1462 cm -1 at several sites on the tissue arrays. 2.2.3 Optical imaging of H&E sections Tissue array sections contiguous with thos e used for IR imaging analysis were mounted on glass slides and stained with hematoxylin and eosin for traditional histopathological analysis. H&E stained tissue microarra y sections were optically imaged using an Olympus BH-2 microsc ope equipped with a high resolution (3 megapixel), Peltier cooled, 10-bit, Q-imag ing micropublisher digital camera. Tissue array spots were imaged individually through a 4x Olympus -corrected microscope
49 objective and 10x camera eyepiece objective. The H&E sections were reviewed with a pathologist for diagnostic features on a two-headed teaching microscope. 2.3 Spectroscopic Imaging Instrumentation FT-IR spectroscopic imaging was perfor med on a Perkin-Elmer (Shelton, CT) Spectrum Spotlight 300 imaging microspectrometer equipped with dual mode detection system. The imaging system is comprise d of two main optical components: the spectrometer and microscope. The spectrometer houses a ceramic globar broadband infrared source, a continuous-scanning Michel son interferometer, and a macro sampling area for non-microscopic single point FT-IR measurements. The modulated infrared output beam of the spectrometer is coupled to an infrared microscope and focused onto the sample using Cassegrain optics. In transmission mode, the infrared beam is focused onto the sample through a Cassegrain condenser. The c ondenser position can be varied along the beam axis to correct for optical effects caused by substrate thickness and refractive index. Transmitted radiation is collected by a Cassegrain objective and focuse d onto one of two mercury cadmium telluride (MCT) detectors. Figur e 2.1 shows a detail ed diagram of the microscope and the optical path in transmission mode. A traditional single mercury cadmium tellu ride (MCT) detector is used in the instruments point mode to take single-poi nt spectroscopic measurements. Seamless software control of variable-width, rota ting knife-edge apertures, and a motorized mapping stage with a precision error < 1 m allow flexible collection of high quality infrared spectroscopic mapping data in point detection mode. The microscope portion of
50 the instrument features a visible LED light so urce and video camera that are linked with control software for automated collection of vi sible-light images as shown in figure 2.1. Self-referenced stage position is dynamically linked with both captured visible images and spectroscopic imaging results. This feat ure allows the operator to choose sample areas for spectroscopic imaging experime nts via a simple interface by selecting a rectangular area on the displa yed optical visible image of the sample, and provides interactive registration of infrared spect roscopic imaging results with corresponding optical visible images. The spotlight 300s image mode utilizes a 16 -element MCT linear array detector to build infrared spectroscopic images of any designated rectangular sample area in a linemapping fashion. A fixed optical zoom allows the instrument to collect image data at two different spatial resolutions. The effectiv e pixel size of these two resolution are 25 x 25 m in low resolution and 6.25 x 6.25 m at high resolution.
beam from spectrometer 1 2 3 4 5 6 4 7 Perkin-Elmer Spectrum Spotlight 300 -Microscope Assembly Features 1)Dual-mode Mercury Cadmium Telluride (MCT) detector.2)Visible CCD camera for optical image acquisition3)Knife-edge apertures for single point measurements4)Dichroic mirrors allow for a common infrared and visible beam path.5)Z-fold allows variable pixel resolution at the sample.6)High-precision sample stage linked directly to the interferometer allows for synchronized scanning and flexible image acquisition 7)LED visible illumination source beam from spectrometer 1 1 2 2 3 3 4 4 5 5 6 6 4 4 7 Perkin-Elmer Spectrum Spotlight 300 -Microscope Assembly Features 1)Dual-mode Mercury Cadmium Telluride (MCT) detector.2)Visible CCD camera for optical image acquisition3)Knife-edge apertures for single point measurements4)Dichroic mirrors allow for a common infrared and visible beam path.5)Z-fold allows variable pixel resolution at the sample.6)High-precision sample stage linked directly to the interferometer allows for synchronized scanning and flexible image acquisition 7)LED visible illumination source adapted from  Figure 2.2 Spectrum Spotlight 300 Microscope Optical Configuration The 1 GB of RAM in the controlling computer limits the size of a single line-mapping image cube acquisition in the imaging mode. The maximum sample area size that can be collected is thus a function of several collection parameters including spatial resolution (high or low), spectral resolution, and spectral wavelength range. Practical considerations such as the liquid nitrogen dewar hold time of 7 hr can also limit the maximum size of image data collection in practice. 2.3.1 Tissue Array FT-IR Data Collection Parameters IR Spectroscopic images of the tissue array spots were collected in transmission configuration in image mode at the high-resolution zoom setting (pixel size of 6.25m). 1641 data points were collected across the spectral region from 4000-720 cm -1 yielding 51
52 spectra with a resolution of 4 cm -1 (2 cm -1 data point interval). Four interferograms were co-added for each individual measurement to increase data signal-to-noise ratios (SNRs). Background spectra consisting of 190 coadded interferograms were collected from nearby locations on the BaF 2 flats between the tissue spots. Data collection with these parameters for a typical 600 m tissue array spot results in a spectroscopic imaging data set with spatial dimensions of ~115 x 115 pixels and a file size of approximately 85 MB. Acquisition time for a typical tissue array spot was approximately 35-40 min. The average SNR fo r a single pixel abso rbance spectrum of tissue was >500:1. 2.3.2 Modifications and Environmental Considerations The microscope and spectrometer assemblie s were enclosed in a Plexiglas housing to enable efficient purging with dry nitrogen gas to remove water vapor and to eliminate air currents. The computer controlling the system was s ituated outside the housing and the exhaust streams from the cooling fans of the spectrometer (source) and microscope (detector electronics) were vented out of the housing to maintain a stable room temperature atmosphere within the housing during data collection. Once the sample was placed on the stage, all positioning, focusi ng, and experimental control could be performed remotely by computer control wit hout opening the housing to the atmosphere. After opening the housing for any reason, 20 minutes were allowed for atmospheric equilibration before spectroscopic measurements were resumed.
53 2.4 Data Handling and Computational Considerations 2.4.1 Data Pre-Processing In its imaging mode, the Spectrum Spotlight 300 makes use of the dead time while the microscope stage is stepped to a new positio n to perform several computational tasks. The functions include in terferogram apodization, fast Fourier transform of collected data to single beam spectra, and ratioing of samp le spectra to background spectra to provide absorbance spectra. Spectroscopic imaging data of tissue array spots were collected individually or in small contiguous groups, checked for spectral qua lity (SNR, baseline fluctuations, etc.), and corre cted for atmospheric water vapor and carbon dioxide using Perkin Elmer proprietary software. The resulting, atmosphere-corrected, spectroscopic images were imported into ENVI (RSI inc., Boulder, CO) using software written in IDL by Dr. Rohit Bhargava; all subsequent image processing was performed in this software environment. Some downstream statistical analyses and chart pl otting were performed using Microsoft Excel and Origin. All processing was carried out computers equipped with 1.7 GHz Intel Pentium 4 processors and a mi nimum of 1 GB of RAM. Individual tissue array spots were mosa icked into one large spectroscopic image dataset for each individual array section for further processi ng. For Array P-16, the final size of the whole-array spectroscopic imag e was ~ 500 x 3680 pixels (or ~1.8 million individual spectra) producing a file size of ~14 GB. Spectro scopic image datasets of the two sections of Array P-40 were ~ 4370 x 550 pixels or (or ~2.4 million individual
54 spectra) with a file size of ~17 GB. A rray P-80 had a final size of 2160x1250 pixels (or ~2.7 million spectra) with a file size of ~18.5 GB. 2.4.2 Spectral Baseline Correction Every infrared absorbance spectrum in th e image scene was individually baseline corrected using custom-designed routines written in IDL by Dr. Rohit Bhargava. Regression is used to calculate the values that lie on the line-segment intersecting each pair of points. These values are subsequently subtracted from the spectral absorbance at the corresponding frequency, and the process is repeated for each spectrum in the image scene. Several hundred average spectra fr om different tissue regions on multiple spots of Array P-16 were compared and frequency pos itions observed to be consistent local minima were chosen as baseline points. A list of the frequency positions used as spectral baseline points appears in Table 2.1.
9821184114412969481328135214781764198422822392254226442708300036823774spectral baseline points(cm-1) 9821184114412969481328135214781764198422822392254226442708300036823774spectral baseline points(cm-1) Table 2.1 Spectral frequencies used for spectroscopic baseline correction The baseline-corrected absorbance intensity of the N-H stretching protein backbone vibration (or Amide A) at 3290 cm -1 was used to differentiate tissue from empty space on the array. All pixels with an absorbance less than 0.08 at 3290 cm -1 were masked to zero for all spectral data points and disregarded during any subsequent processing. 55
56 Chapter Three Infrared Spectroscopic Histology of Prostate 3.1 Visualization of Spectral Images and Verification of Histologic Features Infrared spectroscopic imaging datasets of prostate tissue microarray sections were initially visualized by plotti ng images of the baseline-corrected absorbance at 3290 cm -1 This wavenumber position corresponds to th e N-H stretching absorbance band or Amide A absorbance, a backbone vibration found in all proteins. Since proteins are basic structural elements of all prostate tissue, Amide A absorbance images are useful for verifying the presence of spots and structural correlation of features with visible optical images of the corresponding H&E stained section. The baseline corrected Amide A absorbance images for 4 tissue array spots fr om a single patient are shown in fig 3.1A along with a corresponding H&E stained consecutive section in Fig 3.1B.
57 AB 0.250.200.150.100.05ABSORBANCE INTENSITY AB 0.250.200.150.100.05ABSORBANCE INTENSITY Figure 3.1A) Baseline-corrected N-H stretching (3290cm-1) absorbance intensity image of four tissue array spots from a single patient on Array P-16 B) Optical images of corresponding H&E stained section. The tissue microarray sections used for IR spectroscopic imaging experiments are subject to harsh deparaffinization conditions of immersion in hexane at 40C for 4 hours. These conditions caused artifactual damage to a handful of spots in each array sections. Typical artifactual problems included partial or complete absence of spots, spots that folded over onto themselves, and spots which were partially detached from the surface of
58 the optical flat. N-H stretching absorbance images such as those seen in figure 3.1A were extremely useful for discovering spots that we re subject to such damage so that they could be eliminated from further analysis. 3.2 Creation of Ground Truth Data Regions of Interest In order to analyze spectra and to train and test classification models, ground truth data for different histological features or classes needed to be established. The name ground truth stems from remote sensing applications where field data from various sources on the ground are acquired and registered with image data to enable class training and/or evaluation of classifi cation performance. A pathologist examined the matching H&E stained tissue array sections microscopically and different histological fe atures present in each spot were marked on optical images of the corresponding H&E staine d sections. The region of interest (ROI) tool in ENVI allows the user to designate a co llection of pixels as belonging to a set, or ROI. ROIs can be manually generated by selecting geometric areas on the spectroscopic images with drawing tools such as rectangles ellipses, or polygons. Pixels may be added to or deleted from ROIs individually, allowi ng the user to carefully edit such groups. ROIs can also be generated from parameters of the data itself, which can be particularly useful. Once created, these ROIs can be used in a variety of image analysis operations from image subsetting and masking to statistic al analyses and image classification. In analyzing the spectroscopic datasets, specific images derived from various absorbance band ratios provided high contrast for discerning different histologic features in the tissue. Fig 3.2A shows the 1080 cm -1 /1544 cm -1 absorbance band ratio image of
59 four tissue array spots from a single patient on Array P16. The 1080 cm -1 band is attributed to a C-O stre tching vibration of glycogen and the band at 1544 cm -1 to the Amide II vibration of the protein backbone. The 1080 cm -1 /1544 cm -1 image provides high contrast between prostate epithelium and stroma. Areas of higher ratio intensity in Fig 3.2A correspond to the basophilic-staining epithelial regions in the optical image of the corresponding H&E stained section in pane l B. The eosinophilic stromal regions of the tissue correspond to lower intensity regions of the 1080 cm -1 /1544 cm -1 ratio image suggesting that glycogen/protein levels are hi gher in epithelia l tissue than in stroma. Another absorbance band ratio that produced useful images was 1206 cm -1 /1544 cm -1 At the spectral resolution used of 4 cm -1 the absorbance feature at 1206 cm -1 typically appears as a shoul der off the higher intensity combination band at 1236 cm -1 attributed to both Amide III vibrational mode of proteins and the asymmetric stretch of phosphodiester (PO 2 ) groups in phospholipids and nucleic acids. Fig 3.2C shows the 1206 cm -1 /1544 cm -1 absorbance band ratio image of 4 tissue array spots taken from a different patient on Array P-16. Compar ison with the image of the matching H&E stained section (Fig 3.2D) reveals poor cont rast between epithelial and stromal tissues, however, excellent contrast is seen between an area of lymphocytic infiltration, indicated by the highest intensity area in the upper spot, and the surrounding stromal and epithelial components.
D AB C 100 mH&E Absorbance Ratio 1210/1544 cm-1H&E 100 m Absorbance Ratio 1080/1544 cm-1 0.08 0.06 0.04 0.02 0.00 INTENSITY INTENSITY 0.300 0.225 0.150 0.075 0.000D AB C 100 mH&E Absorbance Ratio 1210/1544 cm-1H&E 100 m Absorbance Ratio 1080/1544 cm-1 0.08 0.06 0.04 0.02 0.00 0.08 0.06 0.04 0.02 0.00 INTENSITY INTENSITY 0.300 0.225 0.150 0.075 0.000 Figure 3.2 Absorbance Band Ratio Images of tissue array spots from Array P-16 Various absorbance band images and band ra tio images were interactively overlaid and used to assist the ROI creation process. Using the pathologi st-reviewed, marked optical images of the H&E stained sections as a guide, collections of pixels in the spectroscopic image of each tissue spot were assigned to one of the ten histological class ROIs listed in table 3.1. The epithelial class includes pixels from different histopathological states, incl uding normal benign epithelium, benign prostatic hyperplasia (BPH), prostatic intraepithelial neoplasia (PIN ), and prostatic ade nocarcinoma (CaP). 60
Stromal histological features were separated into 3 subclasses: fibrous stroma, smooth muscular stroma, and mixed stroma based on the H&E section images and spectral differences noted between these three subclasses. Remaining classes included sites of lymphocytic infiltration, vessel endothelium and muscular coat, peripheral nerve tissue, ganglion cells, blood cells, and corpora amylacea. In making the component analysis, much care was taken to include only those pixels that were definitively representative of a particular class, and therefore pixels near edges or class borders were eliminated to insure that class spectral statistics remain uncontaminated. number of spectra in class ROI162956103962843823623591976275174609114448029316 patient array1134lymphocytes153554Total828corpora amylacea767blood cells0ganglion cells0peripheral nerve54endothelium560smooth musclestroma30144mixedstroma19092fibrous stroma1134epithelial tissue40 patient arrayHistologicClass number of spectra in class ROI162956103962843823623591976275174609114448029316 patient array1134lymphocytes153554Total828corpora amylacea767blood cells0ganglion cells0peripheral nerve54endothelium560smooth musclestroma30144mixedstroma19092fibrous stroma1134epithelial tissue40 patient arrayHistologicClass Table 3.1 Histologic class population data Class data were stored separately for each spot and histologic class as individual regions of interest (ROI) in ENVI and could be operated on individually at the spot level or merged to patient level or into a single ROI at the class level. This flexibility enables downstream comparisons to be made at the spot-spot and patient-patient level for each class and across classes. 61
62 3.3 Spectral analysis of histologic features and metric selection The individual ROIs from each spot were merged together to form a single large ROI for each of the ten histologic classes fo r each array. The total number of pixels, where each pixel represents an individual spect rum, is shown for each histologic class in table 3.1. The spectra from each ROI were averaged to create a mean spectrum for each class, displayed in figure 3.3.
Normalized AbsorbanceWavenumber (cm-1)-1) 14001300120011001000D Normalized AbsorbanceWavenumber (cm-1)-1) 3600 3400 3200 3000 2800Normaized AbsorbanceWavenmber (cm-1)-1) 175017001650160015501500B C 350030002500200015001000 Wavenumber (cm-1)Normalized AbsorbanceA-1)AEPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERV GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM B C D Normalized AbsorbanceWavenumber (cm-1)-1) 14001300120011001000D Normalized AbsorbanceWavenumber (cm-1)-1) 14001300120011001000D Normalized AbsorbanceWavenumber (cm-1)-1) 3600 3400 3200 3000 2800Normalized AbsorbanceWavenumber (cm-1)-1) 175017001650160015501500B C 350030002500200015001000 Wavenumber (cm-1)Normalized AbsorbanceA-1)A 350030002500200015001000 Wavenumber (cm-1)Normalized AbsorbanceA-1)AEPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM EPITHELIUM FIBROUS STROMA MIXED STROMA SMOOTH MUSCLE NERVE GANGLION CELLS BLOOD LYMPHOCYTES CORPORA AMYLACEA ENDOTHELIUM B C D Figure 3.3 Histologic class mean spectra The spectra were calculated from baseline corrected spectra and were normalized to amide II absorbance at 1544cm-1. Panel A contains the full spectral window collected 720-4000 cm -1 Panels B, C, and D contain enlargements of the corresponding boxes in panel A. 63
64 3.4 Construction of a Supervised Classi fication Model for Prostate Histology 3.4.1 Spectral Data Reduction The mean spectra for each histologic class were compared and spectral features, frequencies, and band ratios could be identified for dist inguishing the various classes from one another. A set of metrics was de veloped involving absorb ance band ratios and peak centers of gravity for features across the entire spectral region. Metric values were computed using software routines written in the statistical lang uage IDL by Dr. Rohit Bhargava and implemented in the remote sens ing software environment ENVI (RSI, inc., Boulder, CO). Histograms of each training class populat ion were plotted and compared for each metric. Most distributions approximate d a normal distribution and showed some variation in mean and standard deviation between classes. Metrics which did not approximate a normal distribution for most clas ses were discarded, since such data can lead to poor performance with parametric classification methods, particularly with Gaussian Maximum Likelihood cl assification algorithms di scussed below in section 3.4.2 Metrics that showed no significant va riation between classes were also discarded, since their inclusion would likel y add only noise to the classi fication. The spectroscopic imaging dataset was reduced from 1641 spect ral bands (wavenumber positions) to a 20band set of candidate spectral metrics, reduc ing the tissue array im aging dataset from 14 GB to a manageable 160 MB. The construction of successful classifica tion model is by nature an interactive, process. Information is gained in small b its as individual problems are identified and
65 strategies are altered to ad just. A common problem encount ered is the existence of classes which possess bimodal di stributions in several spectral bands. Such observations typically indicate that the class is composed of two or more spectrally distinct subclasses. In such cases, classification accuracy can of ten be dramatically improved by splitting the training data for the suspect class into separa te classes. Similar histogram analysis performed on several absorbance band ratio im ages from early FT-IR imaging studies of non-array prostate tissue indica ted that stromal tissue in th e prostate was composed of spectrally distinct subclasses. These prelim inary results formed the basis for splitting stroma into three separate subclasses: fi brous stroma, smooth muscular stroma, and mixed fibromuscular stroma. A listing of the parameters for each of the 20 candidate spectral metrics appears below in Table 3.2.
MaxMinDenominator BandNumerator Band1062154412121080 13901034120615441544154415441544154415441544(cm-1)1034145012361164 1400105012101236165432929661034106210801114(cm-1)Band Ratio ParametersCenter-of-GravitySpectral Region Band Ratio20 Band Ratio19 Band Ratio18 Band Ratio1715601478Center of Gravity1617641572Center of Gravity1536823000Center of Gravity1412961184Center of Gravity1314201372Center of Gravity12 Band Ratio11 Band Ratio10 Band Ratio9 Band Ratio8 Band Ratio7 Band Ratio6 Band Ratio5 Band Ratio4 Band Ratio3 Band Ratio2 Band Ratio1(cm-1)(cm-1)Type of MetricMetric # MaxMinDenominator BandNumerator Band1062154412121080 13901034120615441544154415441544154415441544(cm-1)1034145012361164 1400105012101236165432929661034106210801114(cm-1)Band Ratio ParametersCenter-of-GravitySpectral Region Band Ratio20 Band Ratio19 Band Ratio18 Band Ratio1715601478Center of Gravity1617641572Center of Gravity1536823000Center of Gravity1412961184Center of Gravity1314201372Center of Gravity12 Band Ratio11 Band Ratio10 Band Ratio9 Band Ratio8 Band Ratio7 Band Ratio6 Band Ratio5 Band Ratio4 Band Ratio3 Band Ratio2 Band Ratio1(cm-1)(cm-1)Type of MetricMetric # Table 3.2 Histology Spectral Metric Definitions 3.4.2 Image Classification Several different algorithms exist for the supervised classification of multispectral image data. Some of the more simplistic classification algorithms such as parallelpiped or minimum-distance approaches do not consider variation that may be present within spectral classes and do not perform well when frequency distributions from separate classes overlap. Histogram analysis of individual metric value class distributions indicated that both significant intraclass variation in spectral metric values exist and that significant overlap between metric value frequency distributions of different classes was common. As examples, individual class histograms for the three most common or populated training 66
classes (epithelium, mixed stroma, and fibrous stroma) are displayed for metric 02 values (Fig. 3.4A) and for metric 11 values (Fig 3.4B). 0.8091.01.11.21.31.41.5 FREQUENCY (normalized to class size)METRC 11 VALUE (band ratio 1400/1390cm-1) Epithelium Mixed Stroma Fibrous Stroma 0.060.080.100.22.214.171.124.200.220.24 FREQUENCY (normalized to class size)METRIC 02 VALUE (band ratio 1080/1544cm-1) Epithelium Mixed Stroma Fibrous StromaA B 0.8091.01.11.21.31.41.5 FREQUENCY (normalized to class size)METRC 11 VALUE (band ratio 1400/1390cm-1) Epithelium Mixed Stroma Fibrous Stroma 0.060.080.100.126.96.36.199.200.220.24 FREQUENCY (normalized to class size)METRIC 02 VALUE (band ratio 1080/1544cm-1) Epithelium Mixed Stroma Fibrous StromaA B Figure 3.4Histograms of metric value class frequency distribution for the three most populated classes (epithelium, mixed st roma, & fibrous stroma) for: A) Metric 02 (band ratio 1080/1544cm-1), and B) Metric 11 (band ratio 1400/1390 cm-1) A parametric approach to supervised classification that is particularly well suited to deal with such natural intraclass spectral variation and intercla ss overlap of metric frequency distributions is the Gaussian Maximum Likelihood (GML) Classifier. An 67
68 n-dimensional probability surface for each class is generated from both class mean and variance statistics for training data consisting of n spectral bands. As classification ensues, each pixels discre te spectrum can be used to calculate the corresponding conditional probability or likelihood that the pixel belongs to each class separately from the individual class n-dimensional probability surfaces. The pixel is then assigned to the class with the highest conditional proba bility. Classification sensitivity can be adjusted by imposing minimum probability thre sholds that cause pixels below a usersupplied minimum conditional probability to be relabeled as unclassified. A supervised Gaussian Maximum Likelihood (GML) algorithm implemented in ENVI was used to classify the 20 metric dataset of the entire tissue array. The 10 different histologic class ROIs were used as input to trai n the classifier. No thresholding was imposed during the classification forcing each pixel in the tissue array image scene to be classified as one of the ten histologi c subtypes. The 10 hist ologic training ROIs were used next as a preliminary validation set to evaluate the performance of the classification. 3.4.3 Array P-16, 20-metric, GML Self-classification results An extremely useful tool for the evaluati on of image classification results is the expression of classification accuracy in terms of an error matrix. Such error matrices are also commonly referred to (appropriately) as confusion matrices. Error matrices compare, on a class-by-class basis, the re lationship between known reference data (ground truth class data) and the corresponding results of a classification attempt.
The same 10 histology class ROIs from Array P-16 that were used as training input for the 20-metric, Gaussian Maximum Likelihood (GML) classification of all tissue on Array P-16 were used as ground truth to calculate an error matrix for the same classification. These data appear below as Table 3.3. 69 0.000.000.560.000.000.250.000.140.03BLOOD0.001.690.280.050.000.000.070.055.68GANGLION0.001.370.560.050.000.003.920.030.79NERVE11.460.460.420.000.002.290.851.160.12ENDOTHELIUM0.000.000.000.000.000.000.000.003.48LYMPHOCYTES0.000.000.000.000.000.000.000.010.01CORPORA AMYLACEA0.640.000.043.900.000.000.7323.180.00SMOOTH MUSCLE0.000.003.560.000.000.000.694.820.25FIBROUS STROMA0.160.000.000.280.000.000.400.130.00MIXED STROMA0.000.460.170.282.070.190.000.110.18EPITHELIUMBLOODGANGLIONNERVEENDOTHELIUMLYMPHOCYTESCORPORA AMYLACEASMOOTH MUSCLEFIBROUS STROMAMIXED STROMAEPITHELIUMGround Truth ClassResult of Classification 87.7497.7294.1294.1597.8299.8196.3694.1970.4289.64 87.7497.7294.1294.1597.8299.8196.3694.1970.4289.64 0.000.000.560.000.000.250.000.140.03BLOOD0.001.690.280.050.000.000.070.055.68GANGLION0.001.370.560.050.000.003.920.030.79NERVE11.460.460.420.000.002.290.851.160.12ENDOTHELIUM0.000.000.000.000.000.000.000.003.48LYMPHOCYTES0.000.000.000.000.000.000.000.010.01CORPORA AMYLACEA0.640.000.043.900.000.000.7323.180.00SMOOTH MUSCLE0.000.003.560.000.000.000.694.820.25FIBROUS STROMA0.160.000.000.280.000.000.400.130.00MIXED STROMA0.000.460.170.282.070.190.000.110.18EPITHELIUMBLOODGANGLIONNERVEENDOTHELIUMLYMPHOCYTESCORPORA AMYLACEASMOOTH MUSCLEFIBROUS STROMAMIXED STROMAEPITHELIUMGround Truth ClassResult of Classification Table 3.3 Error Matrix of supervised GML Classification results using 20 spectroscopic metrics The classifier was implemented in ENVI and was trained on sets of reference spectra assigned to one of ten histologic classes. All matrix values are given in units of percent of ground truth class pixels. The columns represent the ground truth or correct class designation and the rows represent the result class as assigned by the GML classifier. The numbers at each position are the percent of the number of total pixels in the column (ground truth class) that were classified as the class of the row. For example, if we examine the epithelium column, we see that 89.64% of epithelial pixels were correctly classified, 0.25 % of epithelial pixels were misclassified as fibrous stroma, 3.48% of epithelial pixels were
70 misclassified as lymphocytes, etc. The valu es that occupy the dia gonal of the confusion matrix (shown in red in Table 3.3) are the classification accuracy for a given class. These values show that this initial classification attempt performs above 94% for all classes except for epithelium (89.6%), mixed stroma (70.4%), and blood (87.7%). 3.4.4 Leave-one-out metric evaluation It was clear from the histogram analysis of the individual metrics in the original set of 20 that certain metrics were better for discriminating cert ain classes than others. In light of the significant fre quency distribution overlaps s een in many cases, a given metrics inclusion in the classification attemp t might provide little to modest increase in classification accuracy for single class or small number of classes while causing a significant decrease in accuracy in the remaining classes. To test for the presence of such contaminating metrics in the original set of 20, a leave-one-out anal ysis was performed. The image scene was reclassified 20 separate times using a total of 19 spectral metrics per attempt, leaving out a different metric fo r each successive trial. The accuracy change for the 3 classes with the worst 20-metric classification accuracy (epithelium, mixed stroma, and blood) with respect to the 20 me tric classification was recorded for each successive trial and is shown below in Figure 3.5.
-5.5-3.5-188.8.131.52.56.58.510.512.51234567891011121314151617181920Metric Left OutPercent change in accuracy Epithelium Mixed Stroma Blood Figure 3.5 Graphical Representation of results of the leave-one-out analysis The tissue array data was reclassified 20 separate times with a total of 19 metrics, sequentially leaving out a different metric. The percent change in classification accuracy for the three histologic classes which performed poorly in the 20 metric classification attempt (epithelium, mixed stroma, and blood) are plotted with the metric number left out varying along the x-axis. While the results of the leave one out analysis were analyzed for every class, for the sake of clarity, Figure 3.5 contains only data from the three classes (epithelium, mixed stroma, and blood) which had the most classification error in the original 20-metric classification. These three classes stand to benefit most from the removal of a possible contaminating metric, and from the results in Fig 3.5 we see that two metrics clearly stood out as detrimental to classification accuracy. All three of the poorly classified classes (epithelium, mixed stroma, and blood) show a significant increase in accuracy when metric 9 and metric 18 are left out individually. 71
3.4.5 Array P-16, 18-metric GML Classification Results Reclassification with the GML algorithm in the absence of both metric 9 and metric 18 produced the most promising self-evaluated results, which are represented as a error matrix in Table 3.4. The training ROIs were used as ground truth input to generate the error matrix results. 72 1.740.000.000.000.000.030.022.77GANGLION0.001.140.560.000.000.004.100.010.48NERVE1.270.680.550.000.002.070.900.470.07ENDOTHELIUM0.000.000.000.000.000.000.000.000.92LYMPHOCYTES0.000.000.000.000.000.000.000.020.01CORPORA AMYLACEA0.640.000.044.180.000.000.945.530.00SMOOTH MUSCLE0.000.003.680.000.000.000.691.150.19FIBROUS STROMA0.480.000.000.280.000.002.760.790.00MIXED STROMA0.000.910.300.283.850.190.000.210.16EPITHELIUMBLOODGANGLIONNERVEENDOTHELIUMLYMPHOCYTESCORPORA AMYLACEASMOOTH MUSCLEFIBROUS STROMAMIXED STROMAEPITHELIUMGround Truth ClassResult of Classification 97.6197.2693.6992.7696.1599.8194.0493.0492.5195.55 0.000.001.950.000.000.440.000.130.00BLOOD0.00 97.6197.2693.6992.7696.1599.8194.0493.0492.5195.55 0.000.001.950.000.000.440.000.130.00BLOOD0.001.740.000.000.000.000.030.022.77GANGLION0.001.140.560.000.000.004.100.010.48NERVE1.270.680.550.000.002.070.900.470.07ENDOTHELIUM0.000.000.000.000.000.000.000.000.92LYMPHOCYTES0.000.000.000.000.000.000.000.020.01CORPORA AMYLACEA0.640.000.044.180.000.000.945.530.00SMOOTH MUSCLE0.000.003.680.000.000.000.691.150.19FIBROUS STROMA0.480.000.000.280.000.002.760.790.00MIXED STROMA0.000.910.300.283.850.190.000.210.16EPITHELIUMBLOODGANGLIONNERVEENDOTHELIUMLYMPHOCYTESCORPORA AMYLACEASMOOTH MUSCLEFIBROUS STROMAMIXED STROMAEPITHELIUMGround Truth ClassResult of Classification Table 3.4 Confusion matrix of supervised GML Classification attempt using 18 spectroscopic metrics Metric 9 and metric 18 were left out of the original set of 20. All 10 classes are classified at an accuracy above 92.5%. A color-coded classified result image of four tissue array spots from a single patient are shown in figure 3.6 with the corresponding H&E section in panel B for comparison. Classification correspondence with the histological features observed in the H&E section is outstanding.
EPITHELIUMFIBROUS STROMAMIXED STROMASMOOTH MUSCLENERVEGANGLION CELLSBLOODLYMPHOCYTESCORPORA AMYLACEAENDOTHELIUM EPITHELIUMFIBROUS STROMAMIXED STROMASMOOTH MUSCLENERVEGANGLION CELLSBLOODLYMPHOCYTESCORPORA AMYLACEAENDOTHELIUM Figure 3.6 Classification results for 2 tissue array spots from the same patient GML Classification was performed with a total of 18 metrics selected from the results of the leave-one-out analysis as shown in figure 3.5. Epithelial pixels were classified correctly 95.6% of the time with the majority of misclassification as ganglion (2.8%) and lymphocytes (0.9%). Mixed stroma pixels were classified correctly 92.5% of the time with the majority of misclassification not surprisingly, as smooth muscle stroma (5.5%) and fibrous stroma (1.15%). Fibrous stroma pixels were classified correctly 93% of the time with the major misclassification predominately occurring as nerve (4%). Interestingly, nerve pixels were correctly classified with an accuracy of 93.7% with the majority of misclassification as fibrous stroma (3.7%). Upon close inspection, the mean spectra of the fibrous stroma and nerve training ROIs proved to have many similarities, as seen in figure 3.3. Spectral similarities between nerve and fibrous stroma include absorbance peaks at 1034 cm -1 and 73
74 1206 cm -1 and a shoulder at 1280cm -1 As a result of this spectral similarity, a substantial number of pixels at the stromal-epithelial interface were observed to be misclassified as nerve when they probably belong to the fibrous stroma or mixed stroma class. Smooth muscle stroma pixels were classified with an accuracy of 94% with the majority of misclassification as mixed stroma (2.8%) and endothelium (2%). Again of note is that while endothelium was correctly cla ssified 92.8% of the time, the majority of misclassification occurred as smooth muscle stroma. While fibrous stroma and nerve represent a pair of classes whose similarity seems likely based on a compositional similarity, the connection between endotheli um and smooth muscle stroma is probably due to impurity in the endotheli al training class. The endot helial training class by far had the fewest number of training spectra at 359. This reflects both the paucity of discernible endothelial tissue visible in prostate sections on H&E st aining and the difficulty in correctly identifying it in co rresponding IR spectroscopic im ages. Endothelial cells are typically very hard to identify as they are single-layered, and are contiguous with the smooth muscular media which is more pronounced in arterial vessels. With a single pixel in the IR spectroscopic images repres enting 6.25 m of ti ssue per edge, it seems highly likely some of the endot helial training pixe ls are contaminated with signal from smooth muscle tissue of the vessel media. Sim ilarly, blood pixels were classified with an accuracy of 97.6% with th e majority of misclassification as endothelium. Lymphocyte pixels were clas sified with an accuracy of 96.2% with all of the misclassification as epithelial pixels (3.8%). A large proportion of pixels which were incorrectly classified as lyphocytes probably re present true spectral mixtures of different
75 class types, since lymphocytic infiltration necessarily overlays regions of stroma and epithelial tissue. Ganglion pixels were cla ssified to an impressive 97.3% with the majority of misclassification as nerve. Co rpora amylacea were classified to an accuracy of 99.8%. While this accuracy value seem s aberrantly high compared with the other classes, examination of the class mean sp ectrum of corpora amyl acea compared with the other class mean spectra (figure 3.3) reveals that it is quite extreme compared with every other spectrum which probably accounts fo r the high self-classi fication accuracy. 3.5 Validation of Prostate His tology Classification Model These impressive results with a simple set of 20 metrics hint at the promise of this approach. One can be certain that many mo re metrics exist that if included would improve classification accuracy. One of the many advantages of this approach is that we can design our metrics to highlight the prope rty of a spectral feature that is changing across classes, whether it be band height relative to another ba nd or band center of gravity irrespective of height. Metrics whic h measure other spectral properties such as absorbance band widths are othe r obvious choices to be tested in the future, while data collection at higher spectral resolution a nd with higher single-pixel SNRs will uncover newly resolvable spectral features which can be harnessed as metrics to improve classification accuracy. An important caveat mentioned prominently most remote sensing references [6163, 141] is that accuracy estimates made usi ng training data regions as ground truth do not necessarily indicate that similar results will be seen when classifying other regions of the image scene. The pixels in the ROI sets used for classifier training and evaluation
76 make up only a tiny fraction of the total number of tissue pixels in the full spectroscopic image of Array P-16. Several spots from Array P-16 were purposely avoided during the training ROI selection process so that they could be used for qualitative validation of promising classification results. Examina tion of these spots with respect to their matching H&E stained sections gave a qualitati ve sense that the 18-metric classification was performing quite well on tissue that was not included in the trai ning sets. As an example, the lower spot in Figure 3.6 contai ns no pixels used in any of the 10 training ROIs, any the classification results agree well with the image of the matching H&Estained section. 3.5.1 Cross-Array Validation As noted in table 3.2, a se t of histology ground truth ROIs was constructed for the spectroscopic imaging dataset of Array P-40 in the same manner as described in section 3.2 in reference to Array P-16. In light of the observed cl assification trends seen in the 18 metric, P-16 training data error matrix in Table 3.3 and discussed in section 3.5, adjustments were made to the classification model class st ructure. The endothelial cl ass was discarded due to insufficient ground truth ROI pi xel populations on both Array P-16 and Array P-40. The extremely thin nature of this tissue structure on cross-section further adds to the difficulty in both establishing ground tr uth information for this pot ential class and evaluating results since pixels in the spectroscopic images have a size of 6.25 m of tissue per pixel edge. Visual analysis of the H&E-staine d section of Array P-40 revealed almost no contiguous areas of pure smooth muscle as seen frequently in Array P-16. Furthermore,
the 18-Metric self-classification results indicated that most of the misclassified mixed stroma pixels were incorrectly classified as smooth muscle stroma and vice versa. Consequently, the ground truth data smooth muscle stroma class and mixed stroma classes were merged into a single mixed stroma class separately for both Array P-16 and Array P-40. The spectral similarity and commission errors seen between the fibrous stroma and nerve classes suggested they might also be better off combined as a single fibrous-stroma class. However, no appreciable nerve or ganglion tissue was found in any of the Array P-40 spots so both the P-16 nerve and ganglion training data were excluded from the cross-array classification attempt. These adjustments to the histology class structure result in a total of 6 classes. Table 3.5 contains the revised, 6-class, histology ground truth class ROI set population data for both Array P-16 and Array P-40. number of spectra in class ROI162956628359103911444773608029316 patient array828corpora amylacea153554Total767blood cells1134lymphocytes19092fibrous stroma30704mixedstroma1134epithelial tissue25 patient arrayHistologicClass number of spectra in class ROI162956628359103911444773608029316 patient array828corpora amylacea153554Total767blood cells1134lymphocytes19092fibrous stroma30704mixedstroma1134epithelial tissue25 patient arrayHistologicClass Table 3.5 Revised 6-class histology ground truth ROIs for Array P-16 and Array P-40 The 6 ground truth ROIs for Array P-16 listed in Table 3.5 were used as training data for supervised classification of all tissue from Array P-40. The same 18 metrics used for classification of Array P-16 in section 3.5.5 were used for both the training data from Array P-16 and for P-40 image data to be classified by the GML algorithm. 77
All pixels in the P-40 image scene were classified and the 6 ground truth class ROIs from Array P-40 were used to construct an error matrix for the cross-array classification result which appears below in Table 3.6. 78 0.000.000.000.040.01BLOOD0.000.000.000.002.30LYMPHOCYTES0.000.000.000.000.10CORPORA AMYLACEA0.260.000.001.851.26FIBROUS STROMA4.950.000.008.100.58MIXED STROMA0.005.395.390.380.58EPITHELIUMBLOODLYMPHOCYTESCORPORA AMYLACEAFIBROUS STROMAMIXED STROMAEPITHELIUMGround Truth ClassResult of Classification 94.7894.7194.6191.5297.5395.74 94.7894.7194.6191.5297.5395.74 0.000.000.000.040.01BLOOD0.000.000.000.002.30LYMPHOCYTES0.000.000.000.000.10CORPORA AMYLACEA0.260.000.001.851.26FIBROUS STROMA4.950.000.008.100.58MIXED STROMA0.005.395.390.380.58EPITHELIUMBLOODLYMPHOCYTESCORPORA AMYLACEAFIBROUS STROMAMIXED STROMAEPITHELIUMGround Truth ClassResult of Classification Table 3.6 Error Matrix for 6-Class, GML Classification Results The classifier was trained on 6-class ground truth data from Array P-16 and applied to classify all tissue pixels in the image data from Array P-40. The same set of 18 spectral metrics used in section 3.5.5 were used for this classification. The error matrix results indicate that classification accuracy in 5 out of 6 classes exceeds 94.5%. Fibrous stroma was the class with the lowest classification accuracy at 91.5%, however, nearly all of such misclassified pixels were incorrectly classified as mixed stroma. This result likely speaks more to the heterogeneity of stroma in general than to any serious problems with the classification itself.
79 3.6 Conclusions and Further Directions These results indicate that such a 6-cla ss, supervised GML classification model can be used to successfully segm ent spectroscopic images of unstained sections of prostate tissue into useful histologic classes based on their spectral properties with respect to spectral class information from a database of previously imaged tissue from a number of patients. Histological class information obtai ned from such images is useful for image display, however standard staining procedures are far cheaper and provide similar information. FT-IR spectroscopic imaging da ta analyzed in this fashion can provide histological image information from unstained specimens. Standard staining techniques can interfere with other anal ytical techniques, such as immunohistochemistry and in situ hybridization, as well as, nucleic acid rec overy from laser capture microdissected material. The histological class information obt ained could also be used to study morphological relationships, such as epithelial/s tromal density ratios in various different states of normal prostate tissue, nodular hyperplasia (BPH), and varying grades of prostatic adenocarcinoma[143, 144]. The supervised classi fication methods for providing histological class information from IR spectroscopic imaging data developed in the above sections are well-suited for automation, provi ding a means for rapid evaluation necessary for high throughput analyses. Furthermore, such histological classifications can be used as a tool for downstream analysis of spectral information from epithe lial tissue in an effort to further study the infrared spectroscopic properties of beni gn prostate epithelial tissue and prostatic adenocarcinoma in many patients. If reliabl e spectral indicators of disease presence and
80 progression can be found, then FTIR microspe ctroscopic imaging techniques can be used as an objective tool to aid in the detecti on and diagnosis of prostatic adenocarcinoma. The next section continues with some prelim inary experiments using a third tissue array, P-80, designed to investig ate some of these issues.
81 Chapter Four Infrared Spectros copic Histopathology of Prostate 4.1 Classification strategy Array P-80 is the most logical choice as a starting point for the analysis of spectral features of populations of benign and malignant prostate epith elial tissue. Array P-80 was constructed from formalin-fixed, paraffin-embedded tissue blocks cut from radical prostatectomy specimens from populatio n of 80 patients with confirmed prostatic adenocarcinoma. The array was constructed with 2 cores from each patient, one from a region of representative ade nocarcinoma, and one from a region with only normal benign epithelium. The intention of the array design was to provide a large patient population and relatively even sampling of benign and malignant tissue for every patient. The first step of the analysis will ap ply the histology classi fication developed in section 3, using the class statistics from the P-16-Array training popul ations to train the classifier. The histology classification result s will be used along with the pathologists interpretation of the matching H&E-stained se ction to designate sepa rate ROIs for benign and malignant epithelium for each patient. Mean spectra will be used to develop a large set of candidate spectral metrics for di stinguishing between benign epithelium and adenocarcinoma. Spectral metrics that show a statistically signifi cant difference between the benign and adenocarcinoma popul ations will then be used in attempts to self-classify Array-P-80 and cross-validate by classifying other arrays with training data from ArrayP-80.
82 4.2 Array P-80 H&E Stained Section Pathology Analysis The H&E stained, matching section of A rray P-80 was carefully reviewed with a pathologist and each spot was evaluated for se veral histopathological parameters. Before the review process, the visible optical image of each spot was printed on a separate sheet of paper and used to record the pathologists comments during the review process. Each spot was assessed initially for tissue preservati on and preparation. Spots that contained significant preparation artifact or no epithe lium were removed from analysis. The pathologist carefully characterized the remaini ng spots and detailed records were kept for subsequent ROI creation and an alysis. The pathological status of epithelial tissue in each spot was considered individually and a ny epithelial tissue for which the pathological or preparation status was at all questionable was marked on the optical images so that it would not be considered in later analyses. All regions of confirmed prostatic ade nocarcinoma were in dividually assigned a Gleason Grade of 1-5 indicating the predominan t Gleason pattern seen in the spot. Once the pathology analysis was complete, the re sults were tabulated and it was found that a total 38 patients contained usable benign ep ithelial tissue 51 patients contained usable regions of prostatic adenocarcinoma. A total of 25 patients from array P-80 contained regions of both benign epithelium and conf irmed prostatic carcinoma. Without the corresponding benign tissue from the same patient as a control, any analysis of spectral features of adenocarcinoma tissue would be questionable. For this reason, the fullspectrum spectroscopic imaging datasets for th e tissue array spots fo r these 25 patients were mosaicked into a single spectrosc opic image for faster processing during downstream analyses.
4.3 Array P-80 Histology Classification Results The 18 histology metrics used in sections 3.5.5 and 3.6.1 were calculated for the new 25 patient image of array P-80 from the baseline-corrected full spectrum image data. The same histology classification performed in section 3.6.1 as cr oss-array validation, was applied instead to the 25-patient, 18-metr ic image of array P-80. The 6 histologic class ROIs from array P-16 listed in table 3. 4 were used as class training data for the GML classification of the 18-metric image. The histology classifica tion results for this 25-patient image are displayed in Figure 4. 1. Figure 4.2 contains the corresponding optical images from the matching H&E stained section. EPITHELIUM FIBROUS STROMA MIXED STROMA EPITHELIUM FIBROUS STROMA MIXED STROMA 700 m 700 mBenign Spot Cancer Spot Benign Spot Cancer Spot BLOOD LYMPHOCYTES CORPORA AMYLACEA BLOOD LYMPHOCYTES CORPORA AMYLACEA Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 Patient 10 Patient 09 Patient 06 Patient 15 Patient 14 Patient 11 Patient 16 Patient 01 Patient 20 Patient 05 Patient 19 Patient 04 Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 Patient 10 Patient 09 Patient 06 Patient 15 Patient 14 Patient 11 Patient 16 Patient 01 Patient 20 Patient 05 Patient 19 Patient 04 Example Figure 4.1Array P-80 histology classification results 83
Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 Patient 10 Patient 09 Patient 06 Patient 15 Patient 14 Patient 11 Patient 16 Patient 01 Patient 20 Patient 05 Patient 19 Patient 04 Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 Patient 10 Patient 09 Patient 06 Patient 15 Patient 14 Patient 11 Patient 16 Patient 01 Patient 20 Patient 05 Patient 19 Patient 04 Figure 4.2 Optical images of H&E stained section of Array P-80 4.3.1 Spatial Filtering of Histology Classification Results The collected datasets have an effective pixel size of 6.25 m x 6.25 m. The spectral data are collected over the wavelength range from 4000-720 cm -1 or 2.5-13.8 m. A given pixel will therefore contain some spectral information from tissue locations represented in the image data by neighboring pixels. Since most cells also have a size within or near the spectral wavelength range of radiation, some misclassification can be attributed to spectral bleeding from neighboring pixels that contain a different class of tissue. As expected, this phenomenon is most prevalent along borders between different histologic classes. Additionally, though the histological GML classification performs quite accurately, it is after all, a model and like all models has an inherent error rate. 84
85 While many types of spatial filtering techniques exist for digital image processing, the nature of the classification results suggest a particular method is most applicable for removing randomly distributed misclassified pixels. Spectroscopic image data and spectral metric data both span continuous ranges of data values within a single image plane. Most commonly applied spatial im age filtering techniques work well with such data and involve some type of spatially-de pendent averaging of pi xel values within a defined local neighborhood of pixels. The GML classifier assigns each pixel in the image scene to one of 6 discrete classes. Each class is represented in the imag e results data by a unique integer values. In such a case, the data values associated with each pixel do not form any sort of continuous scale, thus any spatial filtering techni ques that rely on averaging would produce meaningless results. Several useful spatial filtering techniques have been developed for image classification results. Some comple x methods utilize the conditional probability statistics developed during the application of the GML classification algorithm to analyze individual pixels with re spect to a defined local ne ighborhood of pixels. Two more simple operations, which also produce satisfactory spatial filtering results, are the sieve operation and majority analysis. A sieve operation considers the neighborhood of pixels around a center pixel of class X, and applies a group minimum threshold. If the number of pixels in the neighborhood classified as X is less than the group minimum, then the pixel is relabeled as unclassified. The pro cess is repeated for every pixel in the image. An alternative spatial filtering technique that can be applied to improve the appearance of image classificati on results is a majority analysis. This technique also considers a kern el, or set of neighboring pixe ls, which is rastered across
the image pixel-by-pixel. As the kernel moves, the center pixel is changed to the class that occupies the majority of the kernel positions that do not contain unclassified pixels. The weight of the center pixel can be changed in integer increments to alter the amount of filtering applied. This technique provides effective filtering of randomly misclassified pixels and changes them to the class that dominates the neighborhood. For this reason, majority-filtered images appear smoother than sieve-filtered results, which contain more unclassified pixels. Figure 4.2 contains results of these two different filtering strategies on a small example region of a classified prostate histology image representing a typical border between epithelium and mixed stroma. 1111 11111 1111 11111 Sieve Operation 8 Nearest Neighbor Group Min = 5 Majority Analysis 3x3 Kernel Even weighting Raw Classification Result EpitheliumMixed Stroma EpitheliumMixed Stroma Figure 4.3 Spatial filtering techniques for classified image results The majority analysis produces results that are extremely smooth and preferable for general classification image display, however, it is important to point out that the 86
87 majority analysis changes the class designation of pixels based solely on spatial information without any spectral information whatsoever. At this stage in the data analysis, the sieve method is much more appr opriate precisely because it is subtractive. Since the histology classification results will be used to c onstruct epithelial ROIs for downstream spectroscopic analyses it is important that such populations be as spectrally pure as possible. The histology classification result imag e displayed in figure 4.1 was spatially filtered using a sieve operation implemente d in ENVI using a neighborhood of eight pixels, and a group minimum thre shold of five. The results were qualitatively compared with the matching H&E section and found to provide satisfactory removal of randomly misclassified pixels, while also removing que stionable pixels near class boundaries. Figure 4.3 contains images of the raw hi stology classification a nd post-sieve operation classification image for patient 2 from array P-80.
Raw Histology Classification Sieve Results 8 Nearest Neighbor Group Min 5 EPITHELIUM FIBROUS STROMA MIXED STROMA EPITHELIUM FIBROUS STROMA MIXED STROMA BLOOD LYMPHOCYTES CORPORA AMYLACEA BLOOD LYMPHOCYTES CORPORA AMYLACEA 200 m 200 m Figure 4.4 Sieve operation spatial filtering of histology classfication results for patient 2 from array P-80 4.4 Construction of a Supervised Classi fication Model for Prostate Pathology 4.4.1 Creation of pathology ground truth ROIs The sieved histology classifi cation results produced in section 4.3 were used as the starting point for the designation of pathology ground truth ROIs for array P-80. First, the sieved histology classification result for a given spot was compared with the annotated optical image of the matching H&E stained section. The epithelial classification result pixels that corresponded to epithelial tissue sele cted for use in the 88
marked optical H&E stained imaged were grouped into separate ROIs for benign epithelium and prostatic adenocarcinoma for each patient for a total of 50 ROIs. An image of the pathology ground truth ROIs and the corresponding number of pixels in each ROI is shown in Figure 4.4. AdenocarcinomaTotal = 42,239 Pixels AdenocarcinomaTotal = 42,239 Pixels Benign Epithelium Total = 19,492 Pixels 700 m 700 mArray P80 Pathology Regions-of-Interest (ROIs) Individual ROI size in pixels 885 698 510 3126 96 3883 662 1143 2037 2001 1073 222 404 803 1219 2248 302 3038 2464 690 289 985 350 518 532 400 1549 833 1774 1113 137 378 623 1476 1479 2934 2604 1434 394 1464 Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 1812 406 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 19 Patient 10 Patient 09 Patient 06 2751 1832 Patient 15 Patient 14 Patient 11 2276 Patient 16 Patient 01 849 Patient 20 Patient 05 Patient 19 Patient 04 583 1099 1334 885 698 510 3126 96 3883 662 1143 2037 2001 1073 222 404 803 1219 2248 302 3038 2464 690 289 985 350 518 532 400 1549 833 1774 1113 137 378 623 1476 1479 2934 2604 1434 394 1464 Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 1812 406 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 19 Patient 10 Patient 09 Patient 06 2751 1832 Patient 15 Patient 14 Patient 11 2276 Patient 16 Patient 01 849 Patient 20 Patient 05 Patient 19 Patient 04 583 1099 1334 Figure 4.5 Array P-80 pathology ground truth ROIs 4.4.2 Pathology Spectral Data Reduction The mean infrared absorbance spectrum for each ROI was created and normalized to Amide II protein backbone absorbance at 1544 cm -1 These mean spectra were compared and spectral features, frequencies, and band ratios could be identified for distinguishing benign epithelial tissue from prostatic adenocarcinoma. A set of 54 candidate metrics was developed involving absorbance band ratios and peak centers of gravity for features across the entire spectral region. A listing of the parameters for each of the 54 candidate spectral metrics appears below in table 4.1. 89
32903060Band Ratio38 32903064Band Ratio39 32903078Band Ratio40 32903084Band Ratio41 32903180Band Ratio42 32903192Band Ratio43 32903202Band Ratio44 32903214Band Ratio45 32903226Band Ratio46 32903232Band Ratio479821144 Center of Gravity4811441182 Center of Gravity4911821296 Center of Gravity5013521426 Center of Gravity5114781578 Center of Gravity5215851718 Center of Gravity53 14501390Band Ratio24 14501400Band Ratio25 10801032Band Ratio26 10801016Band Ratio27 12361208Band Ratio28 12361278Band Ratio30 12361262Band Ratio29 14501080Band Ratio33 16521080Band Ratio34 32901080Band Ratio35 32903044Band Ratio36 32903050Band Ratio37 12361516Band Ratio31 12361080Band Ratio32 15443082Band Ratio20 15443290Band Ratio21 15443450Band Ratio22 14501426Band Ratio23MaxMinDenominator BandNumerator Band 1544154415441544154415441544154415441544154415441544154415441544154415441544(cm-1) 165215881562153615161502145013121278123612061170115811161080106210401012966(cm-1)Band Ratio ParametersCenter-of-GravitySpectral Region 30003682Center of Gravity54 Band Ratio19 Band Ratio18 Band Ratio17 Band Ratio16 Band Ratio15 Band Ratio14 Band Ratio13 Band Ratio12 Band Ratio11 Band Ratio10 Band Ratio9 Band Ratio8 Band Ratio7 Band Ratio6 Band Ratio5 Band Ratio4 Band Ratio3 Band Ratio2 Band Ratio1(cm-1)(cm-1)Type of MetricMetric # 32903060Band Ratio38 32903064Band Ratio39 32903078Band Ratio40 32903084Band Ratio41 32903180Band Ratio42 32903192Band Ratio43 32903202Band Ratio44 32903214Band Ratio45 32903226Band Ratio46 32903232Band Ratio479821144 Center of Gravity4811441182 Center of Gravity4911821296 Center of Gravity5013521426 Center of Gravity5114781578 Center of Gravity5215851718 Center of Gravity53 14501390Band Ratio24 14501400Band Ratio25 10801032Band Ratio26 10801016Band Ratio27 12361208Band Ratio28 12361278Band Ratio30 12361262Band Ratio29 14501080Band Ratio33 16521080Band Ratio34 32901080Band Ratio35 32903044Band Ratio36 32903050Band Ratio37 12361516Band Ratio31 12361080Band Ratio32 15443082Band Ratio20 15443290Band Ratio21 15443450Band Ratio22 14501426Band Ratio23MaxMinDenominator BandNumerator Band 1544154415441544154415441544154415441544154415441544154415441544154415441544(cm-1) 165215881562153615161502145013121278123612061170115811161080106210401012966(cm-1)Band Ratio ParametersCenter-of-GravitySpectral Region 30003682Center of Gravity54 Band Ratio19 Band Ratio18 Band Ratio17 Band Ratio16 Band Ratio15 Band Ratio14 Band Ratio13 Band Ratio12 Band Ratio11 Band Ratio10 Band Ratio9 Band Ratio8 Band Ratio7 Band Ratio6 Band Ratio5 Band Ratio4 Band Ratio3 Band Ratio2 Band Ratio1(cm-1)(cm-1)Type of MetricMetric # Table 4.1 Pathology spectral metric parameters 90
4.4.3 Histogram analysis of Spectral Metric Data Initial metric evaluation was conducted pl otting histograms of different pathology ground truth ROIs for individual metrics. Hi stograms analyzed on a patient-to-patient basis revealed that for many metrics, a similar directional shift in the means of frequency distributions between benign and adenocarcinom a populations was present. For many of these metrics, while the direction of the shift was consistent from patient-to-patient, the absolute values of the respective distributions varied quite significantly among patients. This situation is depicted schematically in figure 4.5. Benign Epithelium Adenocarcinoma Benign Epithelium Adenocarcinoma 0020.040.060.080.100.120.14 Frequency (Normalized) Metric Value 0020.040.060.080.100.120.14 Frequency (Normalized) Patient 1 Patient 2 Patient 3 Benign Epithelium Adenocarcinoma Benign Epithelium Adenocarcinoma Benign Epithelium Adenocarcinoma Benign Epithelium Adenocarcinoma 0020.040.060.080.100.120.14 Frequency (Normalized) Metric Value0020.040.060.080.100.120.14 Frequency (Normalized) 0020.040.060.080.100.120.14 Frequency (Normalized) Metric Value 0020.040.060.080.100.120.14 Frequency (Normalized) 0020.040.060.080.100.120.14 Frequency (Normalized) Patient 1 Patient 2 Patient 3 Figure 4.6 Patient-to-patient metric variation It was clear that many of these metrics were providing information regarding real spectral differences between benign and cancerous prostate tissue, however, the significant patient-to-patient variation rendered these metrics ineffective for use in parametric classification attempts. 91
92 4.4.4 Mean-centering of epithelial metric data. The data were mean-centered in order to make use of the spectroscopic information contained in the metrics affected by signi ficant patient-to-patient variation and to simplify the process of metric evaluation. The mean metric spectrum for each patients benign ground truth ROI was calculated by av eraging the individual metric-spectra within each ROI. The discrete 54-metric sp ectrum of each individua l epithelial pixel was divided by the mean benign metric spectru m from the corresponding patient. This calculation has the effect of normalizing the benign population distributions for all patients individually for each metric. Thus, all patient-patient variation among benign metric distributions is effectively collapsed such that recalculation of the benign 54metric spectrum for any patient would yi eld a value of one at every metric. 4.4.5 Metric Statistical Analysis A major advantage of mean-centering the metric data is that it simplifies the task of identifying which metrics provide statistica lly significant discrimi nation between benign and adenocarcinoma patient populations. The mean 54-metric spectrum for each patients adenocarcinoma ground truth ROI population was recalculated from the benign mean-centered metric data. A one population ttest was applied to each of 54 sets of 25 patient-mean metric values to determine if the 25 patient population was significantly different from the constant 1.0 at the 0.05 leve l. The results of the t-test for each metric and the associated p-values are listed in table 4.2.
94 The t-test results indicate that 20 metr ics from the candidate set of 54 pathology metrics show statistically significant deviat ion between their respective populations of adenocarcinoma and patient-matche d benign epithelial pixels. 4.4.6 GML Pathology Classification of Array P-80 The set of 20 pathology metrics identifie d in section 4.2.8 for discriminating between benign and malignant prostate tissue we re used as data for GML classification of all epithelial pixels from P-80 ground truth epithelial ROIs. The benign ground truth ROI sets for all 25 patients were merged into one large benign training ROI comprised of 19,492 total pixels. Likewise, the adeno carcinoma ground truth ROI sets for all 25 patients were merged into one large adeno carcinoma training RO I comprised of 42,239 total pixels. A supervised 2-class (beni gn epithelium & adenocarcinoma) classification was implemented in ENVI using the 20 metrics identified in section 4.2.8. The classification image results for all 25 patients appear below in Figure 4.6.
Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 Patient 10 Patient 09 Patient 06 Patient 15 Patient 14 Patient 11 Patient 16 Patient 01 Patient 20 Patient 05 Patient 19 Patient 04 Patient 25 Patient 24 Patient 23 Patient 22 Patient 21 Patient 17 Patient 12 Patient 07 Patient 02 Patient 18 Patient 13 Patient 08 Patient 03 Patient 10 Patient 09 Patient 06 Patient 15 Patient 14 Patient 11 Patient 16 Patient 01 Patient 20 Patient 05 Patient 19 Patient 04 Epithelium 700 mAdenocarcinoma Figure 4.7 Array P-80 pathology classification results The ground truth training ROIs were used to construct an error matrix to evaluate the classification results on a whole-array basis. The error matrix appears below in Figure 4.3. 95
96 BENIGN EPITHELIUMADENOCARCINOMABENIGN EPITHELIUMGround Truth ClassResult of Classification 74.50 89.59 10.41ADENOCARCINOMA25.50 74.50 89.59 10.41ADENOCARCINOMA25.50BENIGN EPITHELIUMADENOCARCINOMABENIGN EPITHELIUMGround Truth ClassResult of Classification Table 4.3 Error matrix for 20-metric pathology GML classification of epithelial tissue on array P-80 These classification results give a sense that in general, the classifier is performing adequately for distinguishing benign epithelium from regions of adenocarcinoma. While very little misclassification of ground truth benign pixels is seen, there are a handful of adenocarcinoma spots in Figure 4.8 which seem to be classified with less certainty than the remainder of the patients. 4.5 Individual Patient Evaluation of P-80 Pathology Classification The 20-metric pathology classification results were analyzed next on an individual patient basis. For each of the fifty pathology ground truth ROIs (25 benign + 25 adenocarcinoma), the percentage of ROI pixels classified as adenocarcinoma for each ROI was plotted as a bar chart in Figure 4.7 below.
Individual patient analysis of Pathology Classification Results010203040506070809010001234567891011121314151617181920212223242526Patient NumberPercent of ROI pixels classified as Adenocarcinoma Benign Epithelium Adenocarcinoma Figure 4.8 Individual patient analysis of 20-metric GML pathology classification The data reveals that the pixels from the benign ROIs of all 25 patients were classified with an accuracy > 80%. Imposing a minimum threshold for adenocarcinoma classification of 20% on the data in figure 4.7 provides 100% discrimination between foci benign and malignant epithelial tissue across the entire population of 25 patients. 4.6 Cross-Array Validation Again it must be noted that such self-evaluation of training data ROIs represents the best possible scenario for producing accurate supervised results. To examine the cross-array performance of the pathology classification model, training data from array p-80 was used to classify mean-centered 20-metric data from other arrays. Upon the pathologists review of the H&E stained section, arrays P-16 and P-40 were each found to contain 5 patients with usuable regions of both benign epithelium. Initial cross-array classification attempts did not yield consistent results. While some 97
98 individual patients yielded results similar to those seen with Array P-80, the limited population sizes of five patient s on each array made it impossi ble to draw any substantive conclusions regarding the cross-array performance of the developed pathology classification model. 4.7 Conclusions and Further Directions These results indicate that spectral feat ures from FTIR spectroscopic imaging data can be used to differentiate between regions of h ealthly benign prostate epithelial tissue and regions harboring prostati c adenocarcinoma. The results presented in this section represent an initial attempt to probe the infrared spectroscopi c characteristics of prostate histopathology and serve to highlight the pr omise that such vibrational spectroscopic imaging techniques hold for the objective analysis of sectioned tissue. Such methods provide simple, readily inte rpretable image-based results that convey histological and pathological information provided by refe rencing spectral database information. Some of the most useful of these preliminary results are those from section 4.8 from the t-test analyses of individual metrics. The t-test results displayed in table 4.2 show the 20 metrics for which the patient populations of patient-mean adenocarcinoma me tric values differed significantly from corresponding benign metric valu es. Close examination of the spectral parameters of each of these 20 metrics listed in table 4.1 re veals that most successful metrics involve spectral information from a handful of spectral regions corresponding to specific vibrational modes. Several me trics involve spectral inform ation from the spectral region between 1200-1000 cm -1 a region with prominent absorbances due to vibrational modes
99 of glycogen, as well as symmetri c stretching of phosphodiester (PO 2 ) groups of nucleic acids. The region between 1300-1200 cm -1 also contributed to several metrics; this region contains spectral abso rbances due to protein Amide III modes and antisymmetric stretching modes of nucleic acid PO 2 groups. Finally, many metrics involved features from the spectral region between 1590-1500 cm -1 a region whose main absorbance is the Amide II mode of proteins arising from N-H bending modes coupled to C-N stretching on the protein backbone. Future spectroscopic imaging of prostate tissue at higher spectral resolutions will allow more information to be extracted from these spectral regions. Alternative classification methods, such as spectral-angle mapping and hierarchical cluster analysis more readily make use of continuous spectra l information, and can be employed using data from these isolated regions of the spect rum. If performed on substantially larger patient populations, it is likely that such approaches will lead to more specific information regarding, spectrally similar s ubgroups of related cancers and correlations with histologic grade an d/or disease progression Clearly, studies conducted w ith larger tissue microarrays and patient populations will advance our understanding of the spectrosc opic properties of prostate pathology. The technology utilized to collect vibrational spectroscopic imaging data is advancing at a rapid pace. Faster collection times, bette r SNRs, and higher data collection at higher spatial and spectral resolutions will all add to the power of this technique in the future. Among the most promising future analytical approaches will be to create techniques to register spectroscopic image data with results from other analytical techniques, such as immunohistochemical staining and in-situ hybridization conducte d after IR data
100 collection, or performed on serial tissue arra y section. Such combinatorial approaches should enable calibrations to be constructe d that could tentativ ely predict staining patterns for multiple panels of antibodies or other probes via spectral pattern recognition from spectroscopic image data of unstained tissue.
101 References 1. Salzer, R., et al., Infrared and Raman imaging of biol ogical and biomimetic samples. Fresenius Journal of Analytical Chemistry, 2000. 366(6-7): p. 712-726. 2. Ingle, J.D. and S.R. Crouch, Spectrochemical analysis. 1988, Englewood Cliffs, N.J.: Prentice Hall. v, 590. 3. Meloan, C.E., Elementary infrared spectroscopy. 1963, New York,: Macmillan. vii, 193 p. 4. G nzler, H. and H.M. Heise, IR spectroscopy : an introduction. 2000, Weinheim: Wiley-VCH. xiii, 361 p. 5. Struve, W.S., Fundamentals of molecular spectroscopy. 1989, New York: Wiley. xii, 379 p. 6. Levine, I.N., Quantum chemistry. 4th ed. 1991, Englewood Cliffs, N.J.: Prentice Hall. x, 629 p. 7. Stevens, A. and J.S. Lowe, Histology. 1992, London,New York, Philadelphia: Gower Medical Pub.(Distributed in the USA a nd Canada by J.B. Lippincott Co.). 378 p. 8. Stryer, L., Biochemistry. 4th ed. 1995, New York: W.H. Freeman. xxxiv, 1064 p. 9. Lehninger, A.L., D.L. Nelson, and M.M. Cox, Principles of biochemistry. 2nd ed. 1993, New York, NY: Worth Publis hers. xli, 1013,  p. 10. Solomons, T.W.G., Organic chemistry. 5th ed. 1992, New York: Wiley. 1 v. (various pagings). 11. Jackson, M. and H.H. Mantsch, Biomedical Infrared Spectroscopy, in Infrared spectroscopy of biomolecules, D. Chapman, Editor. 1996, Wiley-Liss: New York. p. 311-340.
102 12. Jackson, M. and H.H. Mantsch, Valinomycin and Its Interaction with Ions in Organic-Solvents, Detergents, and Lipids Studied by Fourier-Transform Ir Spectroscopy. Biopolymers, 1991. 31(10): p. 1205-1212. 13. Kubelka, J. and T.A. Keiderling, Differentiation of beta-she et-forming structures: Ab initiobased simulations of IR absorption and vibrational CD for model peptide and protein beta-sheets. Journal of the American Chemical Society, 2001. 123(48): p. 12048-12058. 14. Haris, P.I. and D. Chapman, The Conformational-Analysis of Peptides Using FourierTransform Ir Spectroscopy. Biopolymers, 1995. 37(4): p. 251-263. 15. Silva, R., et al., Discriminating 3(10)from alpha he lices: Vibrational and electronic CD and IR absorption study of rela ted Aib-containing oligopeptides. Biopolymers, 2002. 65(4): p. 229-243. 16. Barth, A. and C. Zscherp, What vibrations tell us about proteins. Quarterly Reviews of Biophysics, 2002. 35(4): p. 369-430. 17. Jackson, M., P.I. Haris, and D. Chapman, Fourier-Transform Infrared Spectroscopic Studies of Lipids, Polypeptides and Proteins. Journal of Molecular Structure, 1989. 214: p. 329-355. 18. Prestrelski, S.J., D.M. Byler, and M.N. Liebman, Comparison of Various MolecularForms of Bovine Trypsin Correlation of Infrared-Spectra with X-Ray CrystalStructures. Biochemistry, 1991. 30(1): p. 133-143. 19. Surewicz, W.K., H.H. Ma ntsch, and D. Chapman, Determination of Protein Secondary Structure by FourierTransform Infrared-Spectroscopy a CriticalAssessment. Biochemistry, 1993. 32(2): p. 389-394. 20. Torii, H. and M. Tasumi, Theoretical Analyses of the Amide I Infrared Bands of Globular Proteins, in Infrared spectroscopy of biomolecules, D. Chapman, Editor. 1996, Wiley-Liss: New York. p. 1-18. 21. Byler, D.M. and H. Susi, Examination of the Secondary Structure of Proteins by Deconvolved Ftir Spectra. Biopolymers, 1986. 25(3): p. 469-487.
103 22. Dong, A., P. Huang, and W.S. Caughey, Protein Secondary Structures in Water from 2nd-Derivative Amide-I Infrared-Spectra. Biochemistry, 1990. 29(13): p. 33033308. 23. Jackson, M., P.I. Haris, and D. Chapman, Fourier-Transform Infrared Spectroscopic Studies of Ca2+Binding Proteins. Biochemistry, 1991. 30(40): p. 9681-9686. 24. Trewhella, J., et al., Calmodulin and Troponin-C Structures Studied by FourierTransform Infrared-Spectroscopy Effects of Ca-2+ and Mg-2+ Binding. Biochemistry, 1989. 28(3): p. 1294-1301. 25. Lis, H. and N. Sharon, Lectins: Carbohydrate-specific proteins that mediate cellular recognition. Chemical Reviews, 1998. 98(2): p. 637-674. 26. Schnaar, R.L., et al., Adhesion of Eukaryotic Cells to Immobilized Carbohydrates. Methods in Enzymology, 1989. 179: p. 542-558. 27. Lewis, R., et al., Physical-Properties of Glycosy ldiacylglycerols an Infrared Spectroscopic Study of the Gel-Phase Polymorphism of 1,2-Di-OAcyl-3-O(Beta-D-Glucopyranosyl)-Sn-Glycerols. Biochemistry, 1990. 29(38): p. 89338943. 28. Jackson, M., D.S. Johnston, and D. Chapman, Differential Scanning Calorimetric and Fourier-Transform Infrared Spectrosc opic Investigations of Cerebroside Polymorphism. Biochimica Et Biophysica Acta, 1988. 944(3): p. 497-506. 29. Lee, D.C., I.R. Miller, and D. Chapman, An Infrared Spectroscopic Study of Metastable and Stable Forms of Hy drated Cerebroside Bilayers. Biochimica Et Biophysica Acta, 1986. 859(2): p. 266-270. 30. Mueller, E., et al., Oriented 1,2-dimyrist oyl-sn-glycero-3phosphorylcholine/ganglioside membranes: A Fourier transform infrared attenuated total reflection sp ectroscopic study. Band assi gnments; Orientational, hydrational, and phase behavior; And effects of Ca2+ binding. Biophysical Journal, 1996. 71(3): p. 1400-1421.
104 31. Mueller, E. and A. Blume, Ftir Spectroscopic Analysis of the Amide and Acid Bands of Ganglioside Gm1, in Pure Form and in Mixtures with Dmpc. Biochimica Et Biophysica Acta, 1993. 1146(1): p. 45-51. 32. Brandenburg, K., S. Kusumoto, and U. Seydel, Conformational studies of synthetic lipid A analogues and partia l structures by infr ared spectroscopy. Biochimica Et Biophysica Acta-Biomembranes, 1997. 1329(1): p. 183-201. 33. Brandenburg, K., Fourier-Transform Infrared-Spectroscopy Characterization of the Lamellar and Nonlamellar Structures of Free Lipid-a and Re Lipopolysaccharides from Salmonella -Minnesota and EscherichiaColi. Biophysical Journal, 199 3. 64(4): p. 1215-1231. 34. Naumann, D., et al., Investigations into the Polymorphism of Lipid-a from Lipopolysaccharides of Escherichia-Co li and SalmonellaMinnesota by FourierTransform Infrared-Spectroscopy. European Journal of Biochemistry, 1987. 164(1): p. 159-169. 35. Barbucci, R., et al., Physicochemical Surface Characte rization of Hyaluronic-Acid Derivatives as a New Cl ass of Biomaterials. Journal of Biomaterials SciencePolymer Edition, 1993. 4(3): p. 245-273. 36. Lewis, R.N.A.H. and R.N. McElhaney, Fourier Transform Infrared Spectroscopy in the Study of Hydrated Lipids and Lipid Bilayer Memebranes, in Infrared spectroscopy of biomolecules, D. Chapman, Editor. 1996, Wiley-Liss: New York. p. 159-202. 37. Clark, G. and Biol ogical Stain Commission., Staining procedures. 4th ed. 1981, Baltimore: Published for the Biological Stain Commission by Williams & Wilkins. xi, 512 p. 38. Presnell, J.K., M.P. Schreibman, and G.L. Humason, Humason's Animal tissue techniques. 5th ed. 1997, Baltimore: Johns Hopkins University Press. xix, 572 p. 39. Parker, F.S., Applications of infrared, raman, and resonance raman spectroscopy in biochemistry. 1983, New York: Plenum Press. xiv, 550 p.
105 40. Liquier, J. and E. Taillandier, Infrared Spectroscopy of Nucleic Acids, in Infrared spectroscopy of biomolecules, D. Chapman, Editor. 1996, Wiley-Liss: New York. p. 131-158. 41. Diem, M., S. Boydston-White, and L. Chiriboga, Infrared spectroscopy of cells and tissues: Shining light onto a novel subject. Applied Spectroscopy, 1999. 53(4): p. 148A-161A. 42. Griffiths, P.R. and J.A. De Haseth, Fourier transform infrared spectrometry. Chemical analysis ; v. 83. 1986, New York: Wiley. xv, 656 p. 43. Christy, A.A., Y. Ozaki, and V.G. Gregoriou, Modern fourier transform infrared spectroscopy. Comprehensive analytical chemistry, v. 35. 2001, Amsterdam ; New York: Elsevier. xx, 356 p. 44. Schaeberle, M.D., I.W. Levin, and E.N. Lewis, Infrared and raman spectroscopy of biological materials, in Practical spectroscopy, B. Yan, Editor. 2001, M. Dekker: New York. p. 231-258. 45. Treado, P.J. and M.D. Morris, Infrared and Raman Spectroscopic Imaging, in Microscopic and spectroscopic im aging of the chemical state, M.D. Morris, Editor. 1993, M. Dekker: New York. p. 71-108. 46. Perkin-Elmer (Shelton, CT) Spectru m Spotlight 300 FTIR Imaging System 47. Lewis, E.N., et al., Fourier-Transform Spectroscopic Imaging Using an Infrared Focal-Plane Array Detector. Analytical Chemistry, 1995. 67(19): p. 3377-3381. 48. Colarusso, P., et al., Infrared spectroscopi c imaging: From planetary to cellular systems. Applied Spectroscopy, 1998. 52(3): p. 106A-120A. 49. Bhargava, R. and I.W. Levin, Noninvasive imaging of molecular dynamics in heterogeneous materials. Macromolecules, 2003. 36(1): p. 92-96. 50. Bhargava, R. and I.W. Levin, Fourier transform infrared imaging: Theory and practice. Analytical Chemistry, 2001. 73(21): p. 5157-5167.
106 51. Bhargava, R., et al., Novel route to faster Fourier transform infrared spectroscopic imaging. Applied Spectroscopy, 2001. 55(8): p. 1079-1084. 52. Snively, C.M., et al., Fourier-transform infrared imaging using a rapid-scan spectrometer. Optics Letters, 1999. 24(24): p. 1841-1843. 53. Huffman, S.W., R. Bhargava, and I.W. Levin, Generalized implementation of rapidscan Fourier transform infrared spectroscopic imaging. Applied Spectroscopy, 2002. 56(8): p. 965-969. 54. Lewis, E.N., et al., High-fidelity Fourier transform infrared spectroscopic imaging of primate brain tissue. Applied Spectroscopy, 1996. 50(2): p. 263-269. 55. Lewis, E.N., et al., Applications of Fourier transform infrared imaging microscopy in neurotoxicity, in Imaging Brain Structure and Function. 1997, NEW YORK ACAD SCIENCES: New York. p. 234-247. 56. Lester, D.S., et al., Infrared microspectroscopic im aging of the cerebellum of normal and cytarabine treated rats. Cellular and Molecular Biology, 1998. 44(1): p. 2938. 57. Kidder, L.H., et al., Infrared spectroscopic imaging of the biochemical modifications induced in the cerebellum of the Niemann-Pick type C mouse. Journal of Biomedical Optics, 1999. 4(1): p. 7-13. 58. Mendelsohn, R., et al., IR microscopic imaging of pathol ogical states and fracture healing of bone. Applied Spectroscopy, 2000. 54(8): p. 1183-1191. 59. Marcott, C., et al., Infrared microspectroscopic im aging of biomineralized tissues using a Mercury-Cadmium-Telluride focal-plane array detector. Phosphorus Sulfur and Silicon and the Related Elements, 1999. 146: p. 417-420. 60. Mendelsohn, R., E.P. Paschalis, and A.L. Boskey, Infrared spectroscopy, microscopy, and microscopic imaging of mineralizing tissues: Spectra-structure correlations from human iliac crest biopsies. Journal of Biomedical Optics, 1999. 4(1): p. 1421.
107 61. Campbell, J.B., Introduction to remote sensing. 3rd ed. 2002, New York: Guilford Press. xxxi, 621 p.,  p. of plates. 62. Richards, J.A. and X. Jia, Remote sensing digital im age analysis : an introduction. 3rd ed. 1999, Berlin ; New Yo rk: Springer. xxi, 363 p. 63. Lillesand, T.M. and R.W. Kiefer, Remote sensing and image interpretation. 4th ed. 2000, New York: John Wiley & Sons. xii, 724 p. 64. Cotran, R.S., et al., Robbins pathologic basis of disease. 6th ed. 1999, Philadelphia: Saunders. xv, 1424 p. 65. McNeal, J.E., Prostate, in Histology for pathologists, S.S. Sternberg, Editor. 1997, Lippincott-Raven: Philadelphia. p. 997-1017. 66. Jemal, A., et al., Cancer statistics, 2003. Ca-a Cancer Journal for Clinicians, 2003. 53(1): p. 5-26. 67. Boring, C.C., T.S. Squires, and T. Tong, Cancer statistics, 1993. CA Cancer J Clin, 1993. 43(1): p. 7-26. 68. Mettlin, C.J. and G.P. Murphy, Why is the prostate cancer death rate declining in the United States? Cancer, 1998. 82(2): p. 249-51. 69. Kirby, R.S., M.K. Brawer, and T.J. Christmas, Prostate cancer. 2nd ed. 2001, London: Mosby. xiii, 230. 70. Franks, L.M., Latent Carcinoma of the Prostate. Journal of Pathology and Bacteriology, 1954. 68(2): p. 603-&. 71. Breslow, N., et al., Latent Carcinoma of Prostate at Autopsy in 7 Areas Collaborative Study Organized by Inte rnational-Agency-forResearch-onCancer, Lyons, France. International Journal of Ca ncer, 1977. 20(5): p. 680-688. 72. Sakr, W.A., et al., The Frequency of Carcinoma and Intraepithelial Neoplasia of the Prostate in Young Male-Patients. Journal of Urology, 1993. 150(2): p. 379-385.
108 73. Sheldon, C.A., R.D. Williams, and E.E. Fraley, Incidental Carcinoma of the Prostate a Review of the Literature and Cr itical Reappraisal of Classification. Journal of Urology, 1980. 124(5): p. 626-631. 74. Silverberg, E. and J.A. Lubera, Cancer statistics, 1989. CA Cancer J Clin, 1989. 39(1): p. 3-20. 75. Woolf, C.M., An Investigation of the Familial Aspects of Carcinoma of the Prostate. Cancer, 1960. 13(4): p. 739-744. 76. Gronberg, H., F. Wiklund, and J.E. Damber, Age specific risks of familial prostate carcinoma: a basis for screening re commendations in high risk populations. Cancer, 1999. 86(3): p. 477-83. 77. Cannon, L., et al., Genetic epidemiology of prosta te cancer in the Utah Mormon genealogy. Cancer Surv, 1982. 1: p. 47-69. 78. Steinberg, G.D., et al., Family History and the Risk of Prostate-Cancer. Prostate, 1990. 17(4): p. 337-347. 79. Matikainen, M.P., et al., Detection of subclinical cance rs by prostate-specific antigen screening in asymptomatic men from high-risk prostate cancer families. Clinical Cancer Research, 1999. 5(6): p. 1275-1279. 80. Carter, B.S., et al., Mendelian Inheritance of Familial Prostate-Cancer. Proceedings of the National Academy of Sciences of the United States of America, 1992. 89(8): p. 3367-3371. 81. Smith, J.R., et al., Major susceptibility locus for prostate cancer on chromosome 1 suggested by a genome-wide search. Science, 1996. 274(5291): p. 1371-1374. 82. Xu, J.F., et al., Evidence for a prostate cancer susceptibility locus on the X chromosome. Nature Genetics, 1998. 20(2): p. 175-179. 83. Rosen, E.M., S. Fan, and I.D. Goldberg, BRCA1 and prostate cancer. Cancer Invest, 2001. 19(4): p. 396-412.
109 84. Gronberg, H., et al., BRCA2 mutation in a family with hereditary prostate cancer. Genes Chromosomes Cancer, 2001. 30(3): p. 299-301. 85. Edwards, S.M., et al., Two percent of men with earlyonset prostate cancer harbor germline mutations in the BRCA2 gene. Am J Hum Genet, 2003. 72(1): p. 1-12. 86. Bonn, D., Prostate-cancer screening targets men with BRCA mutations. Lancet Oncol, 2002. 3(12): p. 714. 87. Pienta, K.J. and P.S. Esper, Risk-Factors for Prostate-Cancer. Annals of Internal Medicine, 1993. 118(10): p. 793-803. 88. Moul, J.W., et al., Racial differences in tumor volume and prostate specific antigen among radical prostatectomy patients. Journal of Urology, 1999. 162(2): p. 394397. 89. Steele, R., et al., Sexual Factors in Epidemiology of Cancer of Prostate. Journal of Chronic Diseases, 1971. 24(1): p. 29-&. 90. Kipling, M.D. and Waterhou.Ja, Cadmium and Prostatic Carcinoma. Lancet, 1967. 1(7492): p. 730-&. 91. Rooney, C., et al., Case-control study of prostatic can cer in employees of the United Kingdom Atomic Energy Authority. Bmj, 1993. 307(6916): p. 1391-7. 92. Rosenberg, L., et al., Vasectomy and the Risk of Prostate-Cancer. American Journal of Epidemiology, 1990. 132(6): p. 1051-1055. 93. Giovannucci, E., et al., A Retrospective Cohort Study of Vasectomy and ProstateCancer in United-States Men. Jama-Journal of the American Medical Association, 1993. 269(7): p. 878-882. 94. Giovannucci, E., et al., A Prospective Cohort Study of Vasectomy and ProstateCancer in United-States Men. Jama-Journal of the American Medical Association, 1993. 269(7): p. 873-877.
110 95. Howards, S.S. and H.B. Peterson, Vasectomy and Prostate-Can cer Chance, Bias, or a Causal Relationship. Jama-Journal of the American Medical Association, 1993. 269(7): p. 913-914. 96. Stanford, J.L., et al., Vasectomy and risk of prostate cancer. Cancer Epidemiology Biomarkers & Prevention, 1999. 8(10): p. 881-886. 97. Scher, H.I., Hyperplastic and Malignant Diseases of the Prostate, in Harrison's principles of internal medicine, A.S. Fauci, Editor. 1998, McGraw-Hill, Health Professions Division: New York. 98. Chodak, G.W. and H.W. Schoenberg, Early Detection of Prostate-Cancer by Routine Screening. Jama-Journal of the American Medical Association, 1984. 252(23): p. 3261-3264. 99. Chodak, G.W., P. Keller, and H.W. Schoenberg, Assessment of Screening for Prostate-Cancer Using the Digital Rectal Examination. Journal of Urology, 1989. 141(5): p. 1136-1138. 100. Wajsman Z., C.T., Detection and Diagnosis of Prostate Cancer, in Prostatic cancer, G.P. Murphy, Editor. 1979, PSG Pub. Co.: Littleton, Mass. p. 94-99. 101. Jacobsen, S.J., et al., Screening digital rectal exami nation and prostate cancer mortality: A population-based case-control study. Urology, 1998. 52(2): p. 173179. 102. Richert-Boe, K.E., et al., Screening digital rectal examination and prostate cancer mortality: a case-control study. Journal of Medical Sc reening, 1998. 5(2): p. 99-103. 103. Friedman, G.D., et al., Case-control study of screeni ng for prostatic cancer by digital rectal examinations. Lancet, 1991. 337(8756): p. 1526-9. 104. Stenman, U.H., et al., A complex between prostate-specific antigen and alpha 1antichymotrypsin is the major form of prostate-specific ant igen in serum of patients with prostatic cancer: assay of th e complex improves clinical sensitivity for cancer. Cancer Res, 1991. 51(1): p. 222-6.
111 105. Christensson, A., et al., Serum Prostate-Specific Antig en Complexed to Alpha-1Antichymotrypsin as an Indicator of Prostate-Cancer. Journal of Urology, 1993. 150(1): p. 100-105. 106. Higashihara, E., et al., Significance of serum free pros tate specific antigen in the screening of prostate cancer. Journal of Urology, 1996. 156(6): p. 1964-1968. 107. Luderer, A.A., et al., Measurement of the Proportion of Free to Total ProstateSpecific Antigen Improves Diagnostic Performance of Prostate Specific Antigen in the Diagnostic Gray Zone of Total ProstateSpecific Antigen. Urology, 1995. 46(2): p. 187-194. 108. Zhang, W.M., et al., Characterization and immunological determination of the complex between prostate-specific antigen and alpha(2)-macroglobulin. Clinical Chemistry, 1998. 44(12): p. 2471-2479. 109. Brawer, M.K., et al., Complexed prostate specific antigen provides significant enhancement of specificity co mpared with total prosta te specific antigen for detecting prostate cancer. Journal of Urology, 2000. 163(5): p. 1476-1480. 110. Benson, M.C., et al., Prostate Specific Antigen Density a Means of Distinguishing Benign Prostatic Hy pertrophy and Prostate-Cancer. Journal of Urology, 1992. 147(3): p. 815-816. 111. Babaian, R.J., et al., Comparative analysis of prosta te specific antigen and its indexes in the detection of prostate cancer. Journal of Urology, 1996. 156(2): p. 432-437. 112. Horninger, W., et al., Improvement of specificity in PSA-based screening by using PSAtransition zone density and percent fr ee PSA in addition to total PSA levels. Prostate, 1998. 37(3): p. 133-137. 113. Carter, H.B., et al., Longitudinal Evaluation of Prosta te-Specific Antigen Levels in Men with and without Prostate Disease. Jama-Journal of the American Medical Association, 1992. 267(16): p. 2215-2220.
112 114. Carter, H.B., et al., Prostate-Specific Antigen Variabi lity in Men without ProstateCancer Effect of Sampling Interval on Prostate-Specific Antigen Velocity. Urology, 1995. 45(4): p. 591-596. 115. Etzioni, R., R. Cha, and M.E. Cowen, Serial prostate speci fic antigen screening for prostate cancer: A computer m odel evaluates competing strategies. Journal of Urology, 1999. 162(3): p. 741-748. 116. Ferguson, J.K., et al., Prostate-specific antigen de tected prostate cancer: pathological characteristics of ultrasound visible versus ultrasound invisible tumors. Eur Urol, 1995. 27(1): p. 8-12. 117. Bree, R.L., The role of color Doppler and stagi ng biopsies in prostate cancer detection. Urology, 1997. 49(3A): p. 31-34. 118. Yu, K.K. and H. Hricak, Imaging prostate cancer. Radiologic Clinics of North America, 2000. 38(1): p. 59-+. 119. Harris, R.D., A.R. Schned, and J.A. Heaney, Cancer with Endorectal Mr-Imaging Lessons from a LearningCurve. Radiographics, 1995. 15(4): p. 813-829. 120. Jager, G.J., et al., Dynamic TurboFLASH subtraction technique for contrastenhanced MR imaging of the prostate: Correlation with hi stopathologic results. Radiology, 1997. 203(3): p. 645-652. 121. Bostwick, D.G. and P.A. Dundore, Biopsy pathology of prostate. 1st ed. Biopsy pathology series ; 20. 1997, London ; New York: Chapman & Hall. xii, 267 p. 122. Gleason, D.F., Histologic grading and clinical staging of prostatic adenocarcinoma, in Urologic pathology : the prostate, M.P. Tannenbaum, Editor. 1977, Lea & Febiger: Philadelphia. p. 171-197. 123. Iczkowski, K.A. and D.G. Bostwick, Prostate biopsy interpretation Current concepts, 1999. Urologic Clinics of North Am erica, 1999. 26(3): p. 435-+. 124. Montironi, R., Prognostic factors in prostate can cer Pathologists glean a wealth of clinical detail from the smallest piece of tissue. British Medical Journal, 2001. 322(7283): p. 378-379.
113 125. McNeal, J.E., et al., Cribriform adenocarci noma of the prostate. Cancer, 1986. 58(8): p. 1714-9. 126. Nielsen, K., et al., Histological grade, DNA ploidy and mean nuclear volume as prognostic factors in prostatic cancer. Apmis, 1993. 101(8): p. 614-20. 127. Epstein, J.I., G. Pizov, and P.C. Walsh, Correlation of pathologic findings with progression after radical retropubic prostatectomy. Cancer, 1993. 71(11): p. 3582-93. 128. Chodak, G.W., et al., Results of Conservative Management of Clinically Localized Prostate-Cancer. New England Journal of Medi cine, 1994. 330(4): p. 242-248. 129. Egawa, S., et al., Long-Term Impact of Conserva tive Management on Localized Prostate-Cancer a 20-Year Experience in Japan. Urology, 1993. 42(5): p. 520526. 130. Albertsen, P.C., et al., Competing risk analysis of men aged 55 to 74 years at diagnosis managed conservatively for clin ically localized prostate cancer. Jama, 1998. 280(11): p. 975-80. 131. Gaffney, E.F., S.N. Osullivan, and A. Obrien, A Major Solid Undifferentiated Carcinoma Pattern Correlates with Tu mor Progression in Locally Advanced Prostatic-Carcinoma. Histopathology, 1992. 21(3): p. 249-255. 132. Blackwell, K.L., et al., Combining Prostate-Specific Antigen with Cancer and Gland Volume to Predict More Reliably Pa thological Stage the Influence of Prostate-Specific Antigen Cancer Density. Journal of Urology, 1994. 151(6): p. 1565-1570. 133. Claudio, P.P., et al., Expression of cell-cycle-regulated proteins pRb2/p130, p107, p27(kip1), p53, mdm-2, and Ki-67 (MIB-1 ) in prostatic gland adenocarcinoma. Clinical Cancer Research, 2002. 8(6): p. 1808-1815. 134. Cowen, D., et al., Ki-67 staining is an independe nt correlate of biochemical failure in prostate cancer treated with radiotherapy. Clinical Cancer Research, 2002. 8(5): p. 1148-1154.
114 135. Sebo, T.J., et al., Perineural invasion and MIB-1 pos itivity in addition to gleason score are significant preoperative pr edictors of progression after radical retropubic prostatectomy for prostate cancer. American Journal of Surgical Pathology, 2002. 26(4): p. 431-439. 136. Bryden, A.A.G., et al., Ki-67 index in metastatic prostate cancer. European Urology, 2001. 40(6): p. 673-676. 137. Bubendorf, L., et al., Tissue microarray (TMA) technology: miniaturized pathology archives for high-th roughput in situ studies. Journal of Pathology, 2001. 195(1): p. 72-79. 138. Bubendorf, L., et al., Hormone therapy failure in human prostate cancer: Analysis by complementary DNA and issue microarrays. Journal of the National Cancer Institute, 1999. 91(20): p. 1758-1764. 139. Kononen, J., et al., Tissue microarrays for high-thr oughput molecular profiling of tumor specimens. Nat Med, 1998. 4(7): p. 844-7. 140. Brochure, Perkin Elmer Spectrum Spotlight 300 FT-IR imaging system. 2001, Perkin-Elmer Instruments, LLC. 141. Schowengerdt, R.A., Remote sensing, models, and methods for image processing. 2nd ed. 1997, San Diego: Academic Press. xlv, 522 p. 142. Fend, F. and M. Raffeld, Laser capture microdissection in pathology. J Clin Pathol, 2000. 53(9): p. 666-72. 143. Marks, L.S., et al., Morphometry of the prostate: I. Di stribution of tissue components in hyperplastic glands. Urology, 1994. 44(4): p. 486-92. 144. Svindland, A., L.M. Eri, and K.J. Tveter, Morphometry of benign prostatic hyperplasia during androgen suppressive th erapy. Relationships among epithelial content, PSA density, and clinical outcome. Scand J Urol Nephrol Suppl, 1996. 179: p. 113-7.
115 About the Author Daniel Celestino Fernandez received his A.B. in Chemistry in 1997 from Amherst College in Amherst, MA. During his senior year he undertook an honors research project where he was exposed to vibrational spectroscopy for the first time and wrote a thesis entitled Spectroscopic Characterization of th e Iron-Sulfur Cluster of Pyruvate FormateLyase Activating Enzyme. After college he returned home to Tampa, FL where he began medical school at the University of S outh Florida. After fi nishing his first two years of study toward an M.D. degree, he accepted a Howard Hughes Medical Institute Research Scholar Fellowship at the National Institutes of Health in Bethesda, MD. In the Section of Molecular Biophysics of the Laborat ory of Chemical Physics of the National Institute of Diabetes, Digestive, and Kidne y Diseases, he joined a team working on biological applications of spect roscopic imaging techniques with Dr. Ira W. Levin. After two years as an HHMI-NIH Research Scholar with the help of the NIH Graduate Partnerships Program, he enrolled in the Medical Sciences Ph.D. program in the Department of Pathology and Laboratory Medicine at the Co llege of Medicine of the University of South Florida and was able to stay at the NIH to complete two additional years of doctoral research. He has now transferred to the Mo unt Sinai School of Medicine in New York City wh ere he is finishing his last year of study toward the M.D. degree. After graduation he plans to comple te a residency in diagnostic radiology.