<%BANNER%>

Quantitative Fluorescence Molecular Tomography

Permanent Link: http://ufdc.ufl.edu/UFE0042325/00001

Material Information

Title: Quantitative Fluorescence Molecular Tomography Algorithms,Simulations,and Phantom/in Vivo Experiments
Physical Description: 1 online resource (122 p.)
Language: english
Creator: Tan, Yiyong
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: cancer, cell, drosophila, fluorescence, mice, molecular, quantitative, stem, tomography, vivo
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: A novel method for quantitative fluorescence molecular tomography (FMT) is developed in this thesis research, including image reconstruction algorithms and imaging hardware. Numerical simulations and phantom experiments are performed to test and validate the implemented reconstruction software and experimental system. Both macro- and meso-scale animal experiments are conducted to evaluate the developed quantitative FMT method. The image reconstruction algorithms are implemented in the framework of finite element method, while the experimental system is constructed using a non-contact, multi-angle transmission scheme. Shape-from-silhouette based volume carving approach is used to render the 3D models of the actual samples. With a free-space light propagation model, the readout from the CCD is converted into the photon density normal to sample surface in order to match the model-based tomographic reconstruction. Depending on the sample size, light propagation in biological tissue is described with two different models. For macro-scale objects, a diffusion equation based FMT reconstruction algorithm is implemented. For meso-scale objects, a radiative transfer equation (RTE) based FMT reconstruction algorithm is adopted. In particular, diffuse optical tomography (DOT) guided FMT reconstruction method is developed to improve the quantification of image reconstruction. The method utilizes measurements at both the excitation and emission wavelengths to reconstruct fluorophore absorption coefficient, the absorption coefficient, and the reduced scattering coefficient in turbid media. Simulations and phantom experiments using targets containing indocyanine green (ICG) indicate that with the optical property distribution reconstructed by DOT, the qualitative and quantitative accuracy of the recovered fluorophore absorption coefficient is improved significantly over that without such a priori knowledge. The applications are demonstrated using both macro- and meso-scale animals. The macro-scale animal experiments are conducted using a mouse model containing a near-infrared (NIR) fluorophore-magnetic nanoparticle hybrid probe which allows the studies of tumor cell quantification andaffinity. The meso-scale animal experiments are performed using a Drosophila pupae model for dynamic monitoring of the red fluorescent protein (DsRed) reporter. The DsRed is an indicator of stem cell stress-induced death events in Drosophila pupae. The in vivo FMT results obtained are cross-validated with fluorescence and confocal microscopy using in vitro samples. The recontructed results presented suggest that the DOT guided FMT method described in this thesis provides a promising noninvasive tool for in vivo quantitative tomographic molecular imaging.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Yiyong Tan.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Jiang, Huabei.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042325:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042325/00001

Material Information

Title: Quantitative Fluorescence Molecular Tomography Algorithms,Simulations,and Phantom/in Vivo Experiments
Physical Description: 1 online resource (122 p.)
Language: english
Creator: Tan, Yiyong
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: cancer, cell, drosophila, fluorescence, mice, molecular, quantitative, stem, tomography, vivo
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: A novel method for quantitative fluorescence molecular tomography (FMT) is developed in this thesis research, including image reconstruction algorithms and imaging hardware. Numerical simulations and phantom experiments are performed to test and validate the implemented reconstruction software and experimental system. Both macro- and meso-scale animal experiments are conducted to evaluate the developed quantitative FMT method. The image reconstruction algorithms are implemented in the framework of finite element method, while the experimental system is constructed using a non-contact, multi-angle transmission scheme. Shape-from-silhouette based volume carving approach is used to render the 3D models of the actual samples. With a free-space light propagation model, the readout from the CCD is converted into the photon density normal to sample surface in order to match the model-based tomographic reconstruction. Depending on the sample size, light propagation in biological tissue is described with two different models. For macro-scale objects, a diffusion equation based FMT reconstruction algorithm is implemented. For meso-scale objects, a radiative transfer equation (RTE) based FMT reconstruction algorithm is adopted. In particular, diffuse optical tomography (DOT) guided FMT reconstruction method is developed to improve the quantification of image reconstruction. The method utilizes measurements at both the excitation and emission wavelengths to reconstruct fluorophore absorption coefficient, the absorption coefficient, and the reduced scattering coefficient in turbid media. Simulations and phantom experiments using targets containing indocyanine green (ICG) indicate that with the optical property distribution reconstructed by DOT, the qualitative and quantitative accuracy of the recovered fluorophore absorption coefficient is improved significantly over that without such a priori knowledge. The applications are demonstrated using both macro- and meso-scale animals. The macro-scale animal experiments are conducted using a mouse model containing a near-infrared (NIR) fluorophore-magnetic nanoparticle hybrid probe which allows the studies of tumor cell quantification andaffinity. The meso-scale animal experiments are performed using a Drosophila pupae model for dynamic monitoring of the red fluorescent protein (DsRed) reporter. The DsRed is an indicator of stem cell stress-induced death events in Drosophila pupae. The in vivo FMT results obtained are cross-validated with fluorescence and confocal microscopy using in vitro samples. The recontructed results presented suggest that the DOT guided FMT method described in this thesis provides a promising noninvasive tool for in vivo quantitative tomographic molecular imaging.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Yiyong Tan.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Jiang, Huabei.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042325:00001


This item has the following downloads:


Full Text

PAGE 1

1 QUANTITATIVE FLUORESCENCE MOLECULAR TOMOGRAPHY: ALGORITHMS, SIMULATION S AND PHANTOM /IN VIVO EXPERIMENTS By YIYONG TAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Yiyong Tan

PAGE 3

3 To my p arents, w ife and d aughter

PAGE 4

4 ACKNOWLEDGMENTS I would like to take this chance to convey my gratitude to all those people who have helped me complete this thesis. Firstly, I am profoundly indebted to my supervisor and mentor Prof. Huabei Jiang. He carefully chose the projects and tailored them to my interest and capacity. Prof. Jiang was always present to listen to me and help me stay in right and most efficiency track. With his encouragements and instructions, I exceeded expectation I set for myself and buil t my foundation and confidence for my future research career. I will forever be grateful to have been a Ph.D. student of Prof. J iang. I would also like to thank Prof. Johannes (Hans) van Oostrom Prof. David Gilland and Prof. Jorg Peters for being on my committee and guiding my thesis progress. I learned a lot from their professional instructions and academic courses. I am very thankful to the collaborators on this thesis project: Prof. Lily Yang in Surgery Department of Emory University, Prof Lei Zhou from Cancer / Genetics Center of Univ. of Florida and Dr. Steven L. Ponder from Imaging Diagnostic Systems Inc It was their efforts that facilitate d the application and highlight ed the significance of this thesis. Finally, I want to thank Dr Qizhi Zhang, Dr.Zhen Yuan and Dr.Lei Yao whose tireless efforts and help made possible much of what appears in this thesis I learn ed a lot from my labmates, and I thank all the lab members for their help and constructive discussion

PAGE 5

5 TABLE OF CONTENTS ACKNOWLEDGMENTS .................................................................................................. 4 page LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 ABSTRACT ................................................................................................................... 10 CHAPTER 1 INTRODUCTION .................................................................................................... 12 Principle of Fluorescence ........................................................................................ 12 Fluorescent Dye ...................................................................................................... 13 Fluorescence Measurement ................................................................................... 13 Sensing Devices ............................................................................................... 14 Data Acquisition Techniques ............................................................................ 15 Fluorescence Molecul ar Tomography (FMT) .......................................................... 17 Theory and Algorithm ....................................................................................... 17 Forward problem ........................................................................................ 18 Inverse problem ......................................................................................... 20 Experimental System ....................................................................................... 22 Marco scale FMT experimental systems ................................................... 22 Meso scale FMT experimental systems ..................................................... 23 Hybrid system with other imaging methodology ......................................... 25 Applications ...................................................................................................... 26 Whole body in vivo fluorescence imaging .................................................. 26 Brain imaging ............................................................................................. 26 Clinical applications ................................................................................... 27 The Aims, Novelty, Significance and Contents of the Dissertation ......................... 28 2 FMT ALGORITHM IMPLEMENTATION AND SIMULATIONS ................................ 32 Diffusion Equation Based Method ........................................................................... 33 Algorithm .......................................................................................................... 33 Simulation ......................................................................................................... 35 Radiative Transfer Equation (RTE) Based Method ................................................. 35 Algorithm .......................................................................................................... 35 Simulation ......................................................................................................... 39 DOT Guided FMT ................................................................................................... 39 Algorithm .......................................................................................................... 40 Simulation ......................................................................................................... 42

PAGE 6

6 3 EXPERIMENTAL SYSTEM AND METHOD ........................................................... 47 Experimental System .............................................................................................. 47 Shape Extraction for Arbitrarily Shaped Objects ..................................................... 49 Camera Calibration .......................................................................................... 49 Multi Camera Calibrations ................................................................................ 51 Visual Hull Method ........................................................................................... 52 Free Space Data Extraction Model ......................................................................... 54 Detector Model ................................................................................................. 54 Source M odel ................................................................................................... 55 Implementation in FMT System .............................................................................. 56 4 PHANTOM EXPERIMENTS ................................................................................... 66 Regular Shaped Objects ......................................................................................... 67 2D Experiments ................................................................................................ 70 3D Experiments ................................................................................................ 71 Arbitrarily Shaped Objects ...................................................................................... 71 5 APPLICATION IN A MOUSE MODEL .................................................................... 80 Introduction ............................................................................................................. 80 Method and Experiments ........................................................................................ 82 3D DOT Reconstruction of Mice ............................................................................. 83 Qu antification of Cy 5.5 ATF Labeled Tumor Cell ................................................. 83 Evaluation of A ffinity ............................................................................................... 85 Conclusions ............................................................................................................ 87 6 APPLICATION IN A DROSOPHILA PUPA MODEL ............................................... 91 Static Fluorophore Concentration Imaging .............................................................. 93 Cy5.5 Microtube Imaging .................................................................................. 93 DsRed Whole Body Imaging ............................................................................ 94 Dynamic DsRed Concentration Imaging ................................................................. 95 Dynamic measurement of DNA accessibility in live animals via FMT. .............. 96 Fluorescence Recovery after Phototbleaching (FRAP) and FMT ..................... 97 7 CONCLUSIONS ................................................................................................... 106 LIST OF REFERENCES ............................................................................................. 109 BIOGRAPHICAL SKETCH .......................................................................................... 122

PAGE 7

7 LIST OF TABLES Table page 4 1 Optical properties of the embedded organs and the background used in the experiments ........................................................................................................ 74

PAGE 8

8 LIST OF FIGURES Figure page 2 1 Diffusion approximation based FMT simulation for centimeter scale. A) Reconstructed bar target B) Reconstructed two cylinder targets: pink dots indicate exact position. ....................................................................................... 45 2 2 RTE based FMT reconstruction at different target depths. Green circle indicates the exact size and position of the target. ............................................. 45 2 3 2D simulation comparison between FMT without A) and with B) DOT guidance ............................................................................................................. 46 3 1 DOT guided FMT experiment system. ................................................................ 58 3 2 Graphic user interface of the system .................................................................. 59 3 3 Coordinates system definition: top view from observe point above the sample stage and Y axis direction is from paper internal to outside. .............................. 60 3 4 Use calibrated CCD model parameters and universe world coordinate of grids to predict pixel values of corners in the image. .......................................... 60 3 5 Visual hull sc heme demonstration. ..................................................................... 61 3 7 Sources (red dots) and detectors (blue dots) of different projections. ................ 62 3 8 Lambert's cosine law .......................................................................................... 63 3 9 Scattering geometry for a diffusive object of volume V surrounded by air for free space model ................................................................................................ 63 3 10 Photon density model of source on the air/sample interface .............................. 64 3 11 Outline of experimental procedures .................................................................... 65 4 1 Photograph of the CCD based CW FMT system. ............................................... 74 4 2 Reconstructeda, s and m xa images for a representative experimental case. ................................................................................................................... 75 4 3 Reconstructed m xa values in the target with and without DOT recovered a and s distributions when different ICG concentration was used. ....................... 76 4 4 Reconstructed 3D images for a representative case (ICG concentration in the target=1 M ). ..................................................................................................... 77

PAGE 9

9 4 5 Arbitrarily shaped phantom experiment. A) Phantom and imaging system. Inclusion is the arbitrarily phantom. B) Raw boundary signal ............................. 77 4 6 Exact positions of the targets in the finite element mesh A), and reconstructed a B) s C), and m xa D) images for a representative case (ICG concentration=1 M ). ................................................................................. 78 4 7 Reconstructed m xa values in the fluorescent targets with and without DOT guidance when different ICG concentration was used. ....................................... 79 5 1 Mouse in Experiment ........................................................................................ 88 5 2 Reconstructed 3D a and s images of a typical mouse (mm1): (a) a image (b) s image ........................................................................................................ 88 5 3 Recovered 3D FMT images from a mouse. ....................................................... 89 5 4 Comparison of signal intensity of the mammary tumor in the mice received uPAR targeted NIR 830 ATF IONPs and non targeted NIR 830 MSA IONP ( with ATF vs without ATF). ................................................................................ 89 5 5 Detection of local recurrent mammary tumor and lung metastasis using reconstructed 3D images of FMT method and NIR 830ATF peptide optical imaging probes. ................................................................................................. 90 6 1 A pupa in experiment ........................................................................................ 100 6 2 Comparative Cy5.5 tube experiment of the diffusion and RTE based FMT reconstruction for a micr otube embedded pupa. .............................................. 101 6 3 3D view of reconstructed Cy 5.5 microtube in the pupa. .................................. 101 6 4 Reconstructed in vivo FMT (top row a f) in vitro confocal microscope (bottom row ad) and epifluorescence microscope (bottom row e & f) images: Column a, b, c and d: transverse slices ; Column e and f : sagittal slices. ....... 102 6 5 3D FMT tomography of 15 stages during the pupation development. .............. 102 6 6 Microscope images of pupa (shell removed) in early stage. ............................. 103 6 7 Quantifying the change of DNA accessibility in li ve animals with FMT and FRAP. .............................................................................................................. 104 7 1 Workflow framework of the d issertation ............................................................ 108

PAGE 10

10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy QUANTITATIVE FLUORESCENCE MOLECULAR TOMOGRAPHY: ALGORITHMS, SIMULATION S AND PHANTOM /IN VIVO EXPERIMENTS By Yiyong Tan December 2010 Chair: Huabei Jiang Major: Biomedical Engineering A novel method for quantitative fluorescence molecular tomography (FMT) is developed in this thesis research including image reconstruction algorithms and imaging hardware. Numerical simulations and phantom experiments are performed to test and validate the implemented reconstruction software and experimental system. Both macroand meso scale animal experiments are conducted to evaluate the developed quantitative FMT method The image reconstruction algorithms are implemented in the framework of finite element method, while the experimental system is constructed using a noncontact, multiangle transmission scheme. Shapefromsilhouette based volume car ving approach is used to render the 3D models of the actual samples. With a free space light propagation model, the readout from the CCD is converted into the photon density normal to sample surface in order to match the model based tomographic reconstruct ion. Depending on the sample size, light propagation in biological tissue is described with two different models F or macro scale objects, a diffusion e quation based FMT

PAGE 11

11 reconstruction algorithm is implemented. For meso scale objects, a r adiative t ransfer e quation (RTE ) based FMT reconstruction algorithm is adopted. In particular, d iffuse optical tomography (DOT) guided FMT reconstruction method is developed to improve the quanti fication of image reconstruction. The method utilizes measurements at both the excitation and emission wavelengths to reconstruct fluorophore absorption coefficient the absorption coefficient, and the reduced scattering coefficient in turbid media. Simulations and phantom experiments using targets containing indocyanine green (ICG) indicate that with the optical property distribution reconstructed by DOT, the qualitative and quantitative accuracy of the recovered fluorophore absorption coefficient is improved significantly over that without such a priori knowledge. The applications are demonstrated using both m acr o and m eso scale animals The macroscale animal experiments are conducted using a m ouse model containing a near infrared ( NIR) fluorophoremagnet ic nanoparticle hybrid probe which allows the studies of tumor cell quantification and affinity. The mesoscale animal experiments are performed using a D rosophila p upae model for dynamic monitor ing of the red fluorescent protein ( DsRed ) reporter The DsRed is an indicator of stem cell stressinduced death events in D rosophila p upae. The in vivo FMT results obtained are crossvalidated with fluorescence and confocal microscopy using in vitro samples. The recontructed results presented suggest that the DOT guided FMT method described in this thesis provides a promising noninvasive tool for in vivo quantitative tomographic molecular imaging .

PAGE 12

12 CHAPTER 1 INT R ODUCTION Fluorescent molecules are extremely valuable tools in biological research preclinical study and clinical diagnosis. Fluorescence can be used to analyze the regulation and expr ession of genes, to locate proteins in cells and tissues, to follow metabolic pathways, to map the tumor to reveal neuron activity and to study the location migration communication of cells. Fluorescence molecular tomography (FMT), as a novel method to obtain in vivo information of fluorescent probe in wholebody small animals under natural state instead of cells in culture dishes and slides ha s transferred from pure numerical simulations to a fast evolving approach for real in vivo experiments and pre clinical applications in the past ten years. Thanks to extensive fluorescent probe research, new signal acquisition techniques and its unique in vivo depth penetration compared to traditional widely used in vitro florescence microscope FMT rece ives more and more attentions a nd becomes a promising tool for small animal imaging .1 5 Pilot clinical research using indocyanine green (ICG) for the breast cancer diagnosis and brain blood volume monitor have also been reported .6, 7 Principle of Fluorescence Different from incandescence result ed from high temperature, fluorescence is induced by excitation light instead and produces very little heat. For this reason fluorescence has also b een referred to a s "cold light" with exterior excitation. B ioluminescence which is self emitting and requires no exterior excitation, is another well known cold light In the standard model of molecules the electrons occupy orbits and energy levels. A molecule that abs orbs light will result in that electrons are moved into a higher energy

PAGE 13

13 orbit states All these excited electronic states are unstable, and later the electrons will lose absorbed energy and fall back to lower energy states. Most absorbed energy is transfer red to heat in form of vibrations within the molecule, but a small portion of electrons emit part of the absorbed energy as photons Th ese emitted photons are so called fluorescence. Fluorescent Dye Recently, biocompatible fluorophore in the visible and n ear infrared spectral bands have been developed with stunning speed: organic dyes such as carbocyanines, fluorescein derivatives and tetrapyrroles, inorganic molecules including lanthanide chelates, various artificial made inorganic quantum dots hairpin shaped oligonucleotide molecular beacons protein fluorescent dye ( YFP, GFP, DsRed derivatives etc) and so on. Especially, quantum dots the artificially made inorganic semiconductor crystal, recently receives increased attention as tunable fluorophores b ecause their emission spectra can be t weake d by design the size of the nanocrystal and can be excited using a monochromatic source while provide desired spectrum of emission. S tudy also shows quantum dots can increase anisotropy factor contrast as well .8 Moreover, most f luorophore can be bound into any desired carrier such as ligand, aptamer nanoparticles and microbub b les to have desired function and enhanced specificity. Fluorescence Measurement On the other hand, advances in fluorescence sensing techniques and acquisition approach enable high quality noninvasively in vivo fluorescence imaging. Photomultiplier Tube (PMT) Photodiode a nd chargecoupled device (C CD) are three widely used light sensing techniques. Fluoresc ence acquisition techniques have also been evolved to new generations: confocal microscope multi photon microscope, t ime -

PAGE 14

14 correlated single p hoton counting (TCSPC) l aser scanning m icroscopy full spectrum filterless fluorescent microscope and planar fluor escence imaging systems. Sensing Devices PMT based systems have been widely used for past two decades and gradually been replaced by high performance p hotodiode based system in the last ten years PMT is a high sensitive and high gain light detector providing current output directly proportional to light intensity. PMT is quite expensive and need careful maintenance: t he vacuum in the PMT will degrade over time; the high voltage supply need regular checking as contaminants attracted to terminals can caus e c urrent leakage and failure PMTs are susceptible to shocks and vibration and can be easily damaged by overexposure light. Photodiode is a semiconductor device containing a pn structure for the detection of light. Light absorbed in the pn structure generates a photocurrent. A Photodiode is a simple, robust, low cost low voltage device and is less susceptible to shocks, movement and overexposure damage s. A valanche photodiode can also get high sensitive and quantum yield compar able to PMT, but it needs prea mplif iers to obtain high gain and the detector area is usually 1mm*1mm which is much smaller than 1cm*1cm of PMT. CCD bec o me s increasingly popular recently in optical imaging field to provide more boundary data and fast data acquisition time. Intensified C CD (ICCD) and Electron Multiplying CCD (EMCCD) are two major advancements for detect ultralow and ultrafast light signal. ICCD uses an image intensifier coupled to the CCD chip to increase the sensitivity down to single photon level and can capture nanosec ond phenomena. Since readout and dark current noises are negligible, no cooling

PAGE 15

15 component is needed for ICCDs. The image intensifier will bring distortion and amplify signal with high gain, so ICCD is only suitable to work at very low light. EMCCD technology is relative new and is introduced in 2000. EMCCDs unite the quantum efficienc y of CCDs and the gain of ICCDs so it can be used as a regular CCD or ICCD by controlling gain. The performance is depended on charge transfer and dark current noises so EMCC Ds must be strongly cooled ( < 6 0 C ) and milliseconds scale read out time is necessary for best performance Data Acquisition Technique s Traditionally, biological fluorescence imaging has been performed in microscopy of in vitro specimens stained with fluorescent dyes. The rich biology information provided by fluorescent dyes has promoted the advance of microscope and in return is much better understood with unprecedented resolution and clarity. Microscopic techniques have also been applied in vivo to study skin or other exposed structures using confocal microscope multiphoton microscope Time Correlated Single Photon Counting (TCSPC) Laser Scanning Microscopy and full spectrum filterless fluorescent microscope How e ver, these approaches are typically limi ted by the depth limit ( 5 0 6 00um), and have been restricted to observ e subsurface fluorescence C onfocal microscope enables the reconstruction of threedimensional structures from the obtained ph oto n s of different focus depth by using a spatial pinhole to eliminate out of focus light from specimens The increased resolution is at the cos t of decreased signal intensity, so long exposures are often required. The imaging depth of confocal microscope is around 200 micrometers. Multiple photon absorption requires less energy of each photon with longer wave length which can greatly decrease the scatteri ng effect and increase penetration depth.

PAGE 16

16 The widely use d multi photon microscope is two photon fluorescent microscope (2PFM) which enhances the penetration depth and at the same time provides clarity image up to 500um depth and allows imaging 1mm deeper r egions for low scatting sample .9 P hoto bleaching and photodamage are limited to the focal region in 2PFM while in 1PFM the whole light path is affected. Although 2PFM has a lower resolution and requires special excitation l asers it provides long er observation time, deeper tissue penetration, more efficient light detection and less phototoxicity Hyperspectral fluorescence microscopy (HFM) is an em erging field based on hyperspectral or multispectral imaging concepts. HFM reveal s fluorescence emission associated spectrum profile of biological samples for each pixel by using tunable optical filter like l iquid crystal tunable filters (LCTF) or acoustooptic tunable filter (AOTF) The HFM has been used to differentiate the contributions of autofluorescence from exogenous fluorescent signals present in the sample and greatly increase signal/noise ratio Wavelength scanning provides full spectrum informati on of fluor o phore dy e while the extra scan time greatly decreases temporal resolution. Snapshot techniques have been developed for hyperspectral imaging to capture the spatial and spectral information simultaneously 10 The fluorescence lifetime is obtained by measurement of the time delays between the excitation pulse and the fluorescence photons. TimeCorrelated Single Photon Counting (TCSPC) laser scanning microscope contains information on both the fluorescence lifetime and the fluorescent intensity for each pixel. The fluorescence lifetime is the average time the molecules remain in excited state before emitting a photon and can indicate fluorophore dynamic state under different environments.

PAGE 17

17 As a widely used tool for in vivo imaging of small animals, conventional planar fluorescen ce imaging systems create a two dimensional fluorescence distribution projectional map.1113 When signal is relative strong, t he planar method can reveal 2D location of fluorophore in living small animal. Howe ver, it is difficult to recover the depth, size, and fluorophore concentration accurately from the projection images. Fluorescence M olecular T omography (FMT) Fluorescence molecular tomography (FMT) can provide valuable in vivo information in living subjects wit h unparalleled depth penetration and accurate 3D localization .14 Unlike above described fluorescence measurement methods which usually imaging surface flourophore, fluorescence molecular tomography can i mage fluorophore up to 5 centimeter de ep FMT images are reconstructed by an optimization process to minimize the discrepancy between theoretic computation and measured boundary fluorescence intensity distributions. F M T can provide cross sectional or full threedimensional (3D) images Recently, FMT gradually becomes a promising tool for small animal in vivo imaging in preclinical research 1 5 FMT enables researchers to rapidly and easily obtain tomographic images o f in vivo biomarkers with various advantages such as unprecedented depth penetration, n o invasive via NIR light, wholebody animal imaging and true quantification. A lthough quantitative accuracy and the resolution need further improvement FMT is evolving with encouraging advances in resolution and quantitative accuracy. Theory and Algorithm Light propagation inside tissue is described based on scale s of applications : Maxwell equations are the fundamental theory and are usually used at the microscale (micro meter) and scattering free situations ; T he radiative transport equation (RTE) is

PAGE 18

18 the approximation of Maxwell equation under assumptions valid at the meso scale (millimeter) with scatteri ng and absorption heterogeneity ; D iffusion equation is the approximat ion of RTE under assumptions valid at the macroscale (centimeter) RTE can describe mesoscale and macroscale, but f rom engineering point of view, it is impractical and unnecessary when d iffusion approximate equation is accurate enough. Biological applic ations of f luorescence tomography are ma jorly in macro scale like m ice or rats and extend recently in meso scale sample s as Drosophila pupae. In radiative transport theory, the propagation of light through a biology tissue medium is formulated based on a conservation law that describes gains and losses of photons in different directions due to scattering and absorption The diffusion approximation (DA) to the RTE is widely used in small animal applications due to its low computation cost and memory effici ency The DA is valid when the reduced scattering coefficient is much large r than the absorption coefficient; the point of observation is far away from excitation source point For both RTE and DA the tomography algorithms are usually compos ed of two part s: f orward p roblem and i nverse p roblem .15 Forward p roblem Forward Problem is to calculate light intensity distribution using known or assigned optical property distributions and light source. F orward solution is unique and accurate to predicate light intensity of any location on boundary or internal domain. Analytical modeling solution. Green's func tions provide a method to solve the diffusion equation or the RTE analytically. The Green's function method is accurate when the source is a spatial and temporal function whose induced light distribution can be solved by convolution. Analytical solutions only exist and is practical for simple homogeneous regular shaped objects 16

PAGE 19

19 Statistical modeling techniques Monte Carlo simulation based modeling is t he most commonly used statistical technique in optical tomography field and is often used as the gold standard to validate other techniques Monte Carlo modeling uses a pseudorandom number generator to trace individual photon trajectories Given enough photons, statistically valid propagation direction, photon diminish rule based on known absorption and reduced scatting coefficients, Monte C arlo modeling can reveal how many photons (light intensity) in any interested small volume region of the objects Numerical techniques for complex geometries. Finite element method (FEM) is a general technique which can be applied to any geometry. FEM is v ery suitable to model arbitrarily shaped object s and was first introduced into optical tomography by Arridge et al .17, 18 and later introduced to frequency domain fluorescence molecular tomography 5 While FEM has become the standard method of choice for modeling complex domains in optical imaging, the finite differenc e method (FDM) ,19, 20 finite volume method (FVM) 21and boundary element method (BEM) 22 have also been used in various applications. It is too computationally expensive to solve the RTE fully by numerical method in a practical application T he RTE can be solved using the PN approximation, 23 by expansion into a rotated spherical harmonic basis24 and by discrete ordinates approximation.25 Discrete ordinates approximation solves the full RTE on a regular grid using FDM FVM or FEM by specify ing discrete directions of light propagation A priori i nformation can greatly improve accuracy of forward problem. One big trend in FMT is to plug a prior i anatomical information and optical property into the forward problem to improve quantification accuracy A priori information can be obtained

PAGE 20

20 from MRI, CT, photoacoustic tomography (PAT), DOT and anatomy structure. T he a priori information and the FMT image s should be acquired simultaneously for accurate image registration, but MRI, CT and PAT methods all will add extra cost and complicate FMT system even bring extra system error and comprise image quality Al th oug h MRI CT and PAT can provide better resolution of anatomical structure, DOT method requires zero extra effort and cost for existing FMT system. In addition, it provides complete quantitative absorption and reduced scattering coefficient distribution as a priori information which can be directly plugged into FMT algorithm Inverse p roblem I nverse problem is a procedure to use actual observations to reconstruct the value of the characterizing parameters in the system under a defined model. Inverse problem is much more difficult to solve than forward problem because: (1) different spatial values dist ribution of the model parameters can have same boundary signal s; (2) re covering the values of the model parameters may require analysis of a huge parameter combinations which result in more unknown parameter than known boundary observation; (3) inverse problems are typically ill posed. A little bit of variation in measurement boundary data due to noise will result in significant changes in the reconstructed images. To stabilize this process, regularization of original data is always added into the reconstruction procedure. How much regularization to add is usually experimentally determined by finding a best tradeoff between robust solution and minimum artifacts due to alteration of original data. Therefore, i nverse problems only find one of best solution ins tead of the unique correct solution and the solving process usual becomes an optimization problem to find best probabilistic solution. Three reconstruction algorithms are widely used for FMT in verse problems

PAGE 21

21 Algebraic reconstruction technique (ART) The Algebraic Reconstruction Technique (ART) is an iterative algorithm for the reconstruction of a twodimensional image from a series of onedimensional angular projections, and the method is already successful used in medical Computed Tomography (C T) scanning. Normalized Born m ethod is used to account for optical heterogeneit ies inside reconstruction domain. In this method the fluorescent signal of each project is normalized with excitation signal of same projection before ART, which simpl ify experiment because the positiondependent coupling factors are canceled out. One disadvantage of normalized Born is the l ost of absolute quantitative information and physics unit due to the normalization process of raw data before reconstruction. In addition, ART based reconstruction simplifies the scattering effect of excitation and emission light 26, 27 Newtons optimization method. In mathematics Newton's method is a very efficient approach to find roots of equations in one or more dimensions It can also be used to find local maxima and local minima of functions as these extrema are the roots of the derivative functions. Based on a set of coupled diffusion equations that describe the propagation of both excitation and fluorescence emission light in highly scattering media, the reconstruction algorithm is centered on Newtons iter ative method where the update of variable is calcu lated based on Jacobian matrix consisting of the derivatives of boundary light intensity at each boundary observation node with respect to fluorescent yield or fluorescent lifetime. 5 28 T runcated Newtons optimization scheme also demonstrated as a much fast er method t o handle large inverse problem .29, 30, 33 Bayesian framework probability based reconstruction. Bayesian estimation is a n approach to the inverse problems that maximum likelihood estimation needs to be

PAGE 22

22 found. It focuses on finding optimal solutions which also honest ly include s error bars in the estimate. The general idea is to start with a constant likelihood function describing the experimenter's a priori knowledge, then the updat e of parameter and parameter covariance is estimate d by a daptive extended Kalman filter (AEKF) Bayesian framework is statistically stric t and computationally efficient to provide a likelihood function of photon density with error estimation L arge 3D optical imaging problems can be implemented within clinically practical computational resources by using automatic progressive parameter reducing inverse zonation and estimation ( APPRIZE ) and datadriven zonation (DDZ) 31, 32, 33 Experimental System T here are various methodologies to implement 3D FMT for different application requirement s : the pulsed or continuous source spatially modulated or multis pectral illumination time or frequency resolved data and polarization or phase sensitive signal .13 Typical FMT system s for general purpose collect all b oundary signal s including highly scattering light to match RTE or diffusion model Recently for special applications and much better resolution, some new FMT system s especially for meso scale objects only selectively collect photon s with certain scattering rule and theoretical model is also modified to match the data collection scheme. Since only fraction of photons are collected, the signal will be very we a k and extra exposure time is required. Marco scale FMT e xperimental system s Fluorescence M olecular T o mography (FMT) experiment system usually has scanning excitation sources on boundary and paired transmission detectors on opposite side or reflectance detectors on same side. It can be implemented using contact fiber optics or non contact CCD camera The s ystem can be frequency domain and

PAGE 23

23 continuous wave for different applications N on contact scheme is a nature choice for in vivo applications of arbitrary sh aped objects, which is crucial for in vivo FMT to be a univers al tool as MRI, CT and PET and becom e a standard medical imaging instrument Typical FMT has been implemented for various applications: Godavarty et al. described a 3D frequency domain reconstruction algorithm 33 suited for a gain modulated intensified CCD (ICCD) setup and succeed to reconstruct ICG concentration for a breast mimicking phantom using a Bayesian framework. Corlu et al succeed ed to use a CCD and optical fiber based continuous wave system to reconstruct ICG concentration of patients breast using normalized finite element method. The system is a contact system need match f luid to enable regular shape reconstruction. Ntziachristos et al present ed a series of in vivo application s by us ing non contact CCD multi projection transmission system with normalized B orn method. 3436 Roy et al demonstrated frequency domain based reflectance fluorescent tomography in phantom study .37 Early photons travel preferentially along the shortest path connecting the photon source to the detector s and experience few scattering events compared with the diffuse photons in the medium. Therefore imaging with timegated detection of early photons is used to reduce the amount of scattered photons contained in the measurements and could lead to better defined forward problems and inverse reconstruct ion .3840 Meso scale FMT experimental systems Recently some new techniques are developed for special ap plication s. In these techniques, only part of scattering photons are collected on boundary, so usually these techniques provide better resolution and with fast reconstruction speed: Laminar optical tomography (LOT ) collect scattering on chosen depth; Opti cal projection tomography

PAGE 24

24 (OPT)27 is good tool for less scattering applications Early photons method only collect s scattering photons within a time window. Single Scattering tomography collect s scattering photons with certain angle. Laminar optical tomogra phy ( LOT) LOT b ridg ed the gap between micrometer and millimeter depth resolution, by combining optical tomographic techniques with a microscopy based setup to allow imaging with 200u m resolution over depths of 0 2.5 mm which surpasses the depth capabiliti es of optical coherence tomography but with a lower spatial resolution LOT is also suited to spectroscopy when multiple narrowband sources are used .41, 42 Optical projection tomo graphy (OPT). OPT is a linear equation based reconstruction technique using a filtered back projection. The method is suitable for small samples with relatively low scattering. By rejecting multi scattering photons to fit directly CT reconstruction algorit hm, OPT produce s high resolution 3D images of both fluorescent and nonfluorescent biological specimens with a thickness up to 15 millimeters OPT microscopy allows high onsite temporal resolution in mapping the tissue distribution of RNA and protein expression in intact embryos or organs for developmental biology and gene function research .27 Single scattering optical tomography ( SSOT) SSOT unlike early photon method, do not rely on time gating for separating singlescattered photons from strong scattered light SSOT utilizes angularly selective intensity measurements to reconstruct the total attenuation coefficient of an inhomogeneous meso scale regime medium. SSOT produces highquality images even in relatively thick samples where the singlescattering ap proximation is expected to break down. So far SSOT is a optical

PAGE 25

25 tomography technique and single scattering fluorescence tomography will be reported in near future 43 Hybrid system with other imaging methodology Commonly used imaging methodology for small animal includes pos itron emission tomography (PET), single photon emission computed tomography (SPECT), magnetic resonance imaging (MRI), ultrasound, X ray computed tomography (CT) and hybrid system The imaging systems for preclinical applications obtain higher resolution a nd detection sensitivity compared with their clinical counterparts because of smaller imaging domain and better penetration. S tandard imaging methods and FMT method can complement each other using either hardware approach (hybrid experimental system) or so ftware approach (multi methodology image registration). The hybrid system or multi methodology imaging are usually for two purposes: 1 FMT has its unique advantage for in vivo animal imaging and can provide additional functional information for other methods such as high molecular specificity nonionizing radiation, optical probe stability ( no intensity decay over time like isotope) and the potential for simultaneous investigations of multiple targets using spectral probes without overlapped emission. 2 Oth er methods provide a priori information which improves the qualitative and quantitative accuracy of FMT A priori information has been found to be particularly effective in improving the quantitative accuracy by guiding and constraining the FMT reconstruct ion algorithm. Hybrid FMT systems reported so far includes: FME/MRI ,4446 FMT/CT ,47, 48 FMT/PET/CT ,49 FMT/ Photoacoustic tomography ( PAT ) .5052

PAGE 26

26 Applications O ver the past decade in vivo Imaging systems for small animals have become increasingly popular The imaging tool developed for small animal preclinical research includes: MRI X ray micro CT Micro PET Micro SPECT Ultrasound PAT Optical C oherence T omography ( O CT ) and Optical T omography (Biol uminescence & fluorescence) Fluorescence molecular tomography has high contrast, high specificity, and biggest reagent arsenal for almost every aspect of biology molecular events and has been widely used for fundamental biology research and preclinical research .1113, 5358 W hole body in vivo f luorescence imaging Although in vivo tomography imaging allows visualization of biology in its intact and native physiological state, it is a technically cha llenging process for several reasons. First, thick and opaque animal tissue absorbs and scatters photons and generates strong autofluorescence which obscure signal and deteriorate quantification. Second, fluorescent dye for complicated in vivo application require biological stabi lity (relative stable value of quantum yield) preferentially accumulat ion at the intended target site s, and high imaging contrast specific to the target s. Third fast metabolism wash out demands restrict time window for experiment s. Despite of these difficulties, great progress is obtained in tumor mapping of m urine tumor, lung carcinomas and breast cancer, chemotherapeutic effect monitoring, angiogenesis related vascular volume analysis and genes expression profiling identification 2, 4, 40, 59 67 Brain imaging The bloodbrain barrier (BBB) is the separation between circulating blood and cerebrospinal fluid (CSF) maintained by the choroid plexus in the central nervous

PAGE 27

27 system (CNS). Most com mercial dye s are h ydrophilic with huge molecular weight and cannot passed BBB : Cy55, ICG, Ca 2+ indicator (fluo 4, c alcium o range and crimson) v oltage sensitive dye and e nzyme indicator (with huge molecular weight) like ProSense At present, th e delivery of these dyes to brain is achieved only for research purpose with following approaches : 1. Invasive direct injection 2. Permeabil ization of tight junctions a osmotic disruption (mannito b biochemical openi ng (RMP 7 Alkermes, histamine). 3. Focused ultrasound (FU S) induced BBB opening The dye usually should have a low molecular weight (~ 500) and high lipophilicity (logP 1 4, P= partition coefficient) to cross the bloodbrain barr ier (BBB) in sufficient amounts. To do non invasive brain research, some dyes is desig ned and synthesized to pass BBB and succeed to facilitate in vivo applications: T umor l ocalization and treatment with n anoparticle NPCP CTX Cy5.5 and MPAPCy5.5 ; Alzheimers disease mechanism study with n anoparticle ( I CQ BCA NPs ) and s mall commercial available dye s such as AOI987 ( MW 410.1) and NIAD4 ( MW 334 ) .47, 68 69, 70 Clinical a pplications The clinical application s of fluorescen ce tomography are even more challenge due to three reasons: Firstly, o nly one infrared dye ( I ndocyanine green ICG) is a pproved by FDA for clinical usage. Secondly, d epth is ten times deeper than small animal s and signal is extreme wea k compared to mouse experiment s. Thirdly, it is hard to get approved for human experiments and patient s also are not willing to have chemical fluorophore in body for research purpose. There are still very promising clinical research achievement s in breast cancer and brain imaging field despite of th ose difficulties: Corlu et al reported that it is possible to detect and reconstruct breast tumor fluorescence in

PAGE 28

28 vivo with fluorescence molecular tomography. The large tumor contrast obtained with a nontargeted exogenous fluorophore (ICG) pictures a promising future as more molecularly targeted dyes get approved for clinical us age .7 Liebert et al first demonstrat e that exogenous ICG introduced intravenously to healthy human volunteers, can be excited and detected noninvasive ly inside the brain.71 The Aims, Novelty, Significance and Contents of the Dissertation The goal of this Ph.D. thesis research is to develop a noncontact 3D fluorescence molecular tomography (FMT) system (both hardware and software) for quantitative in vivo imaging of spatial distribution of fluorescent probes/r eporters in both macroand meso scale animals. The goal is achieved through the completion of the following three major tasks: Development of robust finite element based reconstruction algorithms for quantitatively accurate recovery of fluorescent concentrations for arbitrarily shaped animals in both macroand meso scale. Implementation of a practical experimental system for in vivo FMT of meso scale animals .Realization of quantitative fluorescent imaging for in vivo animal applications. The thesis summ arizes all the results obtained during my graduate study and systematically describes the methods developed and their theoretical background. Development of hardware and finite element based reconstruction algorithms for quantitative FMT We made considera ble effort to develop a robust easy to use FMT system for in vivo animal imaging. The novel features of this system includes: 1, an automatic system. The data flow from the acquisition of the raw signals to the 3D rendering of reconstructed images is strea mlined: Raw data are evaluated with a computer algorithm for signal optimization; Initial state for reconstruction is

PAGE 29

29 autotomatically chosen without the users intervention; Reconstruction domain is determined by raw fluoresecence intensity distribution an d 3D representation settings are optimized to highlight the regions of interest (ROI). 2, Capability of imaging both maco and meso scale samples using the same hardware and software. The computer program automatically determines the method of processing t he experiment data based on the dimension of the samples. 3, Multi angle non contact measurements for arbitrarily shaped objects and automatic selection of source and detector positions based on the contour of the samples. The coupling coefficients of the free space calibration model are determined and validated through well controlled phantom experiments. 4, Fast computation and efficient memory management for FMT image reconstruction. 5, Compatible data formats for easy coregistration with other imaging m odalities such as lCT or MRI. Implementation of the RTE based reconstruction algorithm for dynamic FMT of meso scale animals (Drosophila pupae) Light propagation in tissue is described in accordance of the sample size: Theoretical studies have shown that t he RTE is the most accurate model for mesoscale imaging, but no experimental validation is reported in this regard. We implemented the RTE based model into our non contact FMT system with finite element based algorithm and first applied to FMT imaging of meso scale animal like Drosophila pupae. We monitor DsRed distribution inside Drosophila pupae and the results are consistent with confocal slice images. The dynamic change of DSRed are also presented and validated by in vitro confocal and microscope sessi on results. The major advantage of in vivo FMT over in vitro confocal microscope in dynamic monitor applications is that in vivo FMT can avoid sacrificing animals and

PAGE 30

30 obtain more reliable information from the same animal over time without changing the experimental setting, thus increasing the statistical validity of the data by minimizing experimental variations. In addition, dynamic monitoring of the meso scale animals can be realized only by FMT. Realization of absolute quantitative fluorescent imaging C urrently most FMT methods are based on a linear algorithm and the assumption of uniform optical property distribution or the optical distribution obtained indirectly from other imaging methods like MRI, CT or Photoacoustic Tomography (PAT). These methods usually provide semi quantitative analysis with arbitrarily units or quantitative analysis based on a calibration curve. We use finite element based d iffuse optical tomography (DOT) guided FMT method to provide truely quantitative fluorescent images: DOT guided method provides a priori optical properties of tissue for quantitative analysis of FMT. The DOT guided FMT method is tested and validated by simulations, phantom experiments and mouse experiments. The results consistently show better quantification and image quality over other FMT methods. This thesis consists of a total of seven chapters: In Chapter 1, a brief review of fluorescent imaging and FMT is given. The principles of fluorescence light, fluorescent dyes, traditional fluorescent imaging methods and their recent developments are reviewed. For the newly emerging FMT method, the associated forward and inverse problems, hardware for both macroand meso scale imaging, and the preclinical applications are discussed. In Chapter 2, image reconstructio n algorithms based on both the diffusion equation and radiative transport equation are implemented using the finite element

PAGE 31

31 method. In addition, the DOT guided FMT reconstruction algorithm is presented. Numerical simulations using all the implemented reconstruction algorithms are conducted. In Chapter 3, the experimental system for FMT is described in detail. In addition, a method for contour extraction of arbitrarily shaped objects and a model of free space light propagation in noncontact geometry are dis cussed. In Chapter 4, tissuemimicking phantom experiments are used to validate the theoretical models presented in Chapter 2 and with the experimental system described in Chapter 3. Furthermore, the proposed DOT guided quantitative FMT method is tested and evaluated with phantom experiments. In Chapter 5, Application of the reconstruction algorithms and imaging system described in Chapters 2 and 3 to macroscale mouse imaging is presented. In these preclinical experiments, tumor bearing mice containing N IR dye ATF nanoparticle probes are imaged. The FMT results obtained indicate that our method has the potential to become a useful tool for monitoring of tumor progression, detection of early stage cancer, chemotherapy evaluation, surgery guidance and drug delivery. In Chapter 6, In Vivo application of the RTE based FMT method to monitoring fluorescent protein (DsRed) in meso scale Drosophila pupae is given. Fluorescence recovery after photobleaching (FRAP) is studied for validating the findings from the dyn amic FMT monitoring. In Chapter 7, the overall conclusions from this thesis research and future directions are given.

PAGE 32

32 C HAPTER 2 FMT ALGORITHM IMPLEMENTATION AND SIMULATIONS Diffuse Optical Tomography (DOT ) has demonstrated promising application s in biom edical imaging field. Examples include detection of cerebral hemorrhages 72 functional imaging of brain activity ,73, 74 diagnosis of rheumatoid arthritis ,75, 76 and cancer mapping .7, 77, 78 These applications rely on the fact that various disease processes and most physiological changes affect the optical properties of biological tissue. The optical properties are the absorption coefficienta, the reduced scattering coefficient s The differences in these optical properties between healthy and pathological tissues provide the contrast for optical tomography technology 79 The contrast and selectivity of DOT are usually not satisfactory; therefore, exogenous fluorescent dye can greatly enhance contrast and selectivi ty O n the other hand, DOT reconstruction complement s FMT by providing all optical information needed in FMT. We here implement t he DOT guided FMT algorithm as a complete optical tomography method and investigate FMT applications in both macro scale and meso scale. For mesoscale modeling, RTE based algorithms is implemented; for macro scale application, diffusion based algorithm is implemented. Simulation is to model a real life or hypothetical situation on a computer so that it can be studied to see how t he system works in perfect conditions without any experiment al errors. By changing optical variables, simulation can be made to predict boundary experimental signal and test inverse algorithm s. Simulation also can optimize algorithm s and experimental setti ngs. A typical simulation procedure includes : 1 target size and variable value is predefined; 2 simulated experiment data is obtained through forward solutions and used as the input of reverse algorithm s; 3 reconstructed targets

PAGE 33

33 are compared to exact pr edefined targets ; 4 change simulation condition (for example source detector numbers and distribution) and repeat 13 to find best setting s that can provide best conformity between predefined targets and reconstructed targets. 3D FMT finite element algor ithm implementation is tested by numerical simulation s in this chapter and phantoms in Chapter 3 Implementation of adjoint sensitive method algorithm, one dimension variable bandwidth storage strategy and fast solver for symmetric matrix dramatically reduc e computation time and memory cost. In addition, as an improvement of methodology, we combined fluorescence molecular tomography (FMT) with diffuse optical tomography (DOT), which allows us to study the impact of heterogeneous optical property distributi on on FMT and provide quantitative FMT Diffusion Equation Based Method Algorithm Our finite element based algorithms for both DOT and FMT have been described in detail 5, 80, 81 Here we outline the algorithms base d on the following coupled diffusion equations that describe the propagation of excitation and emission light in tissue: 0(r) S (r) (r) (r) (r) Dx x a x xx (1) 0(r) (r) (r) (r) (r) (r) Dx a ma m mm x m (2) where x,m is the photon density for excitation (subscript x) or emission light (subscript m), Dx,m is the diffusion coefficient, m x,a is the absorption coefficient for excitation and emission light due to contributions fro m both nonfluorescing chromophores and fluorophores, and m xa is the fluorescence quantum yield. Sx(r) is

PAGE 34

34 the excitation source term, which for a point source can be written as S = S0 r0), where S0 r0) is the Dirac delta function for a source centered at r0. The diffusion coefficient can be written as Dx,m = )) r ( ) r ( ( 3 / 1m x m xs a where ) r (m xs is the reduced scattering coefficient. In this study, we use the nonzero photon density or t ype III boundary conditions: Dx m x m n = x m, where n is the unit normal vector to the boundary surface, and is the coefficient related to the internal reflection at the boundary. Making use of finite element discretization, we obtain the matrix representations of Eqs. (1) and (2) and realize other derived matrix relationships through differentiation, which lead to a set of equations capable of inverse problem solution: (5) ) ( ( (4) } { A } b { } ]{ [ 3 } {b } ]{ [A c m x, o m x, T m x, m x, T m x, m x, m x, m x, m x, m x, m x, m x, m x, w here i j a i j m x ij m x,m xD ) (A ds ) ( S ) (bi j M 1 j j x i i x ds ) ( ) ( ) b (i j M 1 j j m k 1 k N 1 j i j j x a k i mk m x and k j i is a set of locally spatially varying Lagrangian basis functions; Dx, xa or m xa ; x.m is the Jacobian matrix consisting of the derivatives of x,m luorescent property profiles; I is identity matrix; o m x and c m x are the observed and the computed excitation or emission photon density at the boundary sites, respectively.

PAGE 35

35 In FM T reconstruction Dx, xa are known or treated as uniform when we calculate x in forward solution with Eqs (1) Dm, ma are know n or treated as uniform as well We plug ged Dx,m, m x,a, x and boundar y m to reconstruct the only unknown m xa in Eqs.(2) by using equation (3),(4),(5). Simulation For brief, we present two typical 3D FMT simulations as shown in Fig. 2 1. Fig. 2 1 A) is the reconstruction of a simulated bar target in a sq uare background; Fig. 2 1 B) is the reconstruction of two simulated cylinder targets in a square background. The reconstruction results show correct value, shape and position of targets. R adiative Transfer Equation (RTE) B ased Method Algorithm The RTE in t he steady state for FMT can be described as following equations: Excitation: r q d r r r r rx S x s x a s1 n x x x (1) Emission: r r 4 1 d r r r r rx S a m s m a s1 n m x m m m (2) where x,m r is the photon density for excitati on (subscript x) or emission light (subscript m), m xs r is scattering coefficient, m xa r is the absorption coefficient for excitation or emission light due to contributions from both nonfluorescing chromophores and fluorophores respectively, and m xa r is the fluorescent yield. 1 nS denotes a unit vector in the direction of interest. Here 1 nS is t he angular direction, n=2 or 3 denotes the physical domain which is considered isotropic in the

PAGE 36

36 sense that the probability of scattering between two directions depends only on the relative angle between those directions. R is the spatial domain, and R denotes its boundary. The kernel is the scattering phase function describing the probability density that a photon with an initial direction will scatter to direction In this study we assume that the scattering phase function depends only on the angle between the incoming and outgoing directions, and thus (3 ) The two dimensional Henyey Greenstein scattering function, the most widely adopted and highly accurate phase function of scattering kernel for light propagation, is used here: 82 cos 2 1 1 2 12 2g g g (4) where is the angle between the input direction and output direction The anisotropy factor, g ( 1 1 g ), defines the shape of the probability density. C onsidering the relatively homogenous optical propert ies and s >> a for early stage D rosophila (<4 th day) pupae in our current study g is set to zero for fast computation. The boundary conditions (BC) for the RTE assume that no photons travel in an inward direction at the boundary R that is, R r all for 0 n 0 r (5) wh ere n is the outward unit normal on S The BC, also known as free surface BC, imply that once a photon escapes the domain it does not reenter it. The BC can be modified to include a boundary source r0 at the source position Ri and can be written as follows : 83

PAGE 37

37 0 n \ S r 0 0 n r r r qi i i i x0 (6) The solution to the RTE with the chosen boundary conditions is existent and unique. 84 Making use of finite element discretization, we obtain the matrix representations of Eqs. (1) and (2) and realize other derived matrix relationships through differentiation, which leads to a set of equations capable of inverse problem solution: } b { } ]{ A [m x m x m x (7) } { A } b { } ]{ A [m x m x m x m x m x (8) ) ( ) I (c m x o m x T m x m x T m x (9) where ij m x,) (A R S i S s R Si j s a R S i j R S i jr d d ,r d r r d d r r dS d r r n r d d r, r1 n 1 n 1 n 1 n 1 n ( 10) 0 j R S i j i x1 ndS d r r n ) (b (11) R s x a i i mr d d r r 4 1 r ) b (1 n m x (12) j i is a set of locally spatially xs xa or m xa ; x.m is the Jacobian matrix consisting of the derivatives of x,m with

PAGE 38

38 or fluorescent property profiles; I is an identity matrix; matrix; o m x and c m x are the observed and the computed excitation or emission photon density at the boundary sites, respectively. In addition, r needs to be expanded as the sum of coefficients multiplied by the Lagrangian basis function: n aN nj N mj mj nj mj nj N j j jr r r1 1 1, (13) where r nj and mj are the nodal spatial and angular basis functions, mj nj is the radiance at the spatial nodal point nj and direction mj nN is the number of spatial nodes of the mesh, and aN is th e number of angular directions. The ray effect may disturb the standard FE techniques when solving the RTE, since it can produce oscillating results or it can visually be seen as photon rays radiating from the source into the direction of the discretization angles 85, 86 To overcome the ray effect, the streamline diffusion modification (SDM) is used in the FE solut ion of the RTE. In the SDM, the test function is written in the form r rj j instead of the standard form of a test function ( rj ). The parameter is the smoothing paramet er which is a spatially varying constant that depends on the local absorption and scattering 85 In this study, s and a were assumed as uniform and their values, a =0.005 mm1 and s =0. 2 mm1, were obtained through an optimization scheme 87 rx was calculated by solving equation (7) (under subscript x). m xa was reconstructed by iteratively solving Eqs. (7) (9) (under subscript m) from presumably uniform initial

PAGE 39

39 estimates until the difference between the measured and computed photon density at the emission wavelength is minimized. The RTE based fluorescence inverse computation requires over a magnitude more memory and time than the diffusion equation based FMT and it is usually realized through parallel computation. We have made an effort t o realize the RTE based fluorescence inverse computation in a single personal computer by incorporating the following features into our algorithm: Element by element unsymmetric solver using hybrid BiCGStab(1) version to save memory 88 and one time fast computation of the Jacobian matrix using adjoint sensitivity method.89 Simulation Two dimensional simulations were conducted to test the RTE reconstruction algorithm described above. The 2D mesh used has 1622 nodes (the direction, mj=16) For the test geometry shown in Fig. 2 2 a total of 14 source and 55 detector positions were used. The initial value used for m xa was 1e8/mm with the updating constrain of m xa >0 The quantum efficiency With a 3GHz personal computer the RTE based fluorescence reconstruction need ed about 0.3 GB memory and 150 minutes to complete. The results are shown in Fig. 2 2 We see that the RTE based method can correctly recons truct the size, position and m xa value (1em) at different depth. DOT Guided FMT While exogenous fluorescent probes improve the contrast and selectivity of the targets of interest, unknown absorption coefficient (a) and reduced scattering coefficient ( s ) distribution in tissue complicate fluorescent tomographic reconstruction

PAGE 40

40 .90, 91 Theoretically, propagation of both excitation and fluorescent emission light in tissue is described by the coupled diffusion equations. The emission photon density is directly related to the excitati on photon density, which in turn depends on the optical property distributions in tissue. In fact, the exogenous fluorescent probe itself in a relatively large target usually has strong absorption and becomes a significant heterogeneity affecting excitation photon density distribution. In the approach we directly reconstruct optical heterogeneities using DOT and apply reconstructed s and a distributions to the FMT reconstruction. Both DOT an d FMT reconstructions are conducted using an optical fiber free system based on a non contact multi angle transmission scheme, coupled with finite element reconstruction algorithms. Our simulation and experimental results suggest that the optical heterogeneous nature of the tar get itself especially when its size is relatively large must be considered for quantitatively correct FMT in the framework of finite element based reconstruction methods. Algorithm DOT guid ance information is a natural way to improve the quantification and qualification of FMT. DOT method provides information on reduced scattering coefficient distribution, absorption distribution and excitation light distribution. All these distributions are the prerequisite for accurate fluorescence modeling and FMT recons truction. The challenge of this algorithm is to get DOT reconstruction as accurate as possible, since incorrect reconstruction will bring in extra error propagation from DOT to FMT. Although DOT reconstruction error is inevitable due to experiment error an d inverse procedure of DOT, due to our simulation and experimental results, DOT guidance information can

PAGE 41

41 help improve the quantification and qualification of FMT reconstruction. In some case boundary artifacts, which defined as extra high value within 3 mi llimeter close to the boundary of DOT, will cause error in FMT. To overcome this issue, we assign background value to boundary artifacts when plugging into FMT reconstructions. In our DOT guided FMT procedure, Dx and xa are reconstructe d by iteratively solving Eqs. (3) (5) (under subscript x) from presumably uniform initial estimates until the difference between the measured and computed photon density at the excitation wavelength is minimized. The recovered Dx and xa are then used to interpolate Dm and ma at the emission wavelength using optical property spectra available from the literature. 1 9 Similar iterative procedure is applied to reconstruct m xa with the interpolated Dm and ma in place in order to minimize the difference between the measured and computed photon density at the emission wavelength. We note that the accuracy for obtaining quantitative recovery of absorption and reduced scattering coeffic ient is critical for the DOT guided FMT method presented here. We have made great efforts in improving our DOT reconstruction algorithm and experimental system to obtain accurate optical property reconstruction. It has been demonstrated in recent years by several groups including our own 9295 that quantitative reconstruction of both absorption and scattering coefficients is possible using CW DOT when a priori information coupled with effective normalization schemes are used. In our CW DOT method, in addition to the use of a priori information obtained by initial search based on raw experimental signal and normalization scheme, the hybrid regularization schemes of Marquardt and Tikhonov play an important role in comba ting the illposed problem involved.87 Our previous phantom study shows that with DOT guidance,

PAGE 42

42 statistically we can obtain better FMT reconstruction in mouse mimicking phantom experiments .96 He terogeneity can also be addressed by normalized Born method and using a priori information from CT, MRI and PAT. Like low resolution disadvantage in DOT guided method, these approaches also have their own issues need to be solved for better performance. Us ing b oundary excitation data to normaliz e emission data can reveal the heterogeneity distribution and cancel out boundary geometry effects but it averages the heterogeneity effect and optical property distributions hidden inside boundary signal are not fu lly utilized In addition, t he heterogeneity of inside tumor or tissue adjacent to fluorophore will affect the fluoresce nt reconstruction most. F luorophore targeted tumor has very high heterogeneity which cannot be recovered by boundary data. In some case s when boundary heterogeneities (moles, surgery scars and subdermal tumor s) are high and the boundary data will not reveal internal heterogeneity and will bring extra error in internal FMT reconstruction. Although MRI, CT and PAT can provide better resolution of anatomical structure, DOT method requires zero extra effort and cost for existing FMT systems. In addition, it provides quantitative absorption and reduced scattering coefficient distribution as a priori information which can be direc tly plugged into FMT algorithm. Simulation Numerous simulation studies reported to date have shown that tissue optical property distribution does pose significant impact on fluorescent image reconstruction. 26, 31, 97101 Experimental results using various FMT approaches were generally consistent with the simulation findings in presence of heterogeneity, although most reported FMT experiments were limited to homogeneous optical property approximation

PAGE 43

43 or a priori optical property distribution, 2, 4, 37, 80, 102 Soubret et al described a normalized Born approximation approach and studied FMT image quality under optical heterogeneity using phantom and animal data 103 Roy et al investigated the impact of optical heterogeneity using a gradient based constrained truncated Newton method. 37 Milstein et al presented a phantom study in presence of optical heterogeneity with a Bayesian framework based FMT method. 101 Herv et al reported improved FMT reconstruction using a normalized Born approximation approach with reconstructed a distribution while assuming s is homogeneous .104 We propose the FEM based DOT guided FMT for better quantitative accuracy and simulations mimicking experiment situat ions are administered. Prior to the phantom experiments, we have conducted considerable numerical simulations using various ICG containing target positions, and target to background contrast levels in terms of fluorophore concentration.These simulations have demonstrated that much improved image quality especially quantitative accuracy of the recovered m xa image can be obtained with a priori knowledge of the a and s distributions, whereas in unif orm optical property treatment the quality of the recovered fluorescence images are significantly degraded with overestimated target size, poor spatial resolution and underestimated value of recovered m xa Since the focus of th e study i s the phantom evaluation and preclinical applications, we just show a representative simulation case for brevity (Fig. 23 ), where the m xaimage was reconstructed without (a) and with (b) a priori a and s information. In this simulation, a 6mm diameter target was positioned at ( 5, 6). The target had a=0.025/mm, s =2/mm

PAGE 44

44 and m xa =0.02/mm. The background had a =0.005/mm, s =1/mm and m xa = 4x1012/mm. we see that when a priori a and s distributions are used, the m xa image is accurately reconstructed in terms of the si ze (estimated from the full width at half maximum (FWHW) of the fluorescent profiles), location and m xa value of the target, whereas when a and s distributions are assumed uniform, the recovered target size is overestimated and the recovered m xa value of the target contains over 60% error compared to the exact value of 0.02/mm.

PAGE 45

45 Fig ure 2 1 Diffusion a pproximation b ased FMT simulation for centimeter scale A) Reconstructed bar target B) Reconstructed two cylinder targets: pink dots indicate exact position. F igure 2 2 RTE based FMT reconstruction at different target depths. Green circle indicates the exact size and position of the target. A B

PAGE 46

46 Fig ure 2 3 2D simulation comparison between FMT without A ) and with B) DOT guidance A B

PAGE 47

47 CHAPTER 3 EXPERIMENTAL SYSTEM AND METHOD A desirable signal collecting platform in small ani mal preclinical research should have following features: n oncontact with objects like modern MRI CT and PET, robust signal collecting independent of spatial position, adaptive to different size sample and practical image registration method for easy comparison to the results of other medical imaging instruments. Fig. 3 1 is the CCD based non cont act continuous wave ( CW ) DOT guided FMT system developed to fulfill stated requirements : The animals like rat m ice and Drosophila can be measured in their nature state, the data acquisition is same for both big and tiny objects and reconstructed images have spat ial coordinates for easy regist ration to other imaging method. The no contact method used in our small animal system can be handily transferred to implement noncontact breast cancer diagnosis. Currently commercial mammal graph for breast cancer diag nosis need to compress breast, which make it unpleasant experience for the patient and lose true spatial information for image registration to other methodology and post surgery. DOT guided quantitative FMT noncontact system can have several advantages com pared with current methods : Non contact scheme to avoid unwanted compress effect s ; O riginal uncompressed shape will help register optical tomography result s with other methods; Functional imaging for soft tissue; C omfortable test environment for patients. Experimental System In our imaging system (see Fig. 3 1), the excitation light can be delivered to the phantom at multiple points in both X and Y directions via linear stages (300nm precision). For each source position, one set of excitation light data from the opposite

PAGE 48

48 side of the phantom is recorded by a 1024x1024 pixels CCD camera (Princeton Instrument, Trenton, NJ ). E mission light is collected b y pla cing a band pass filter in front of the CCD. Our system has been constructed such that multi angle transm ission light can be collected for more accurate image reconstruction ( four angles are used in most case with 0.5 arcsecond angle resolution). This is realized by rotating the phantom three 90 to collect transmission light at four diff erent phantom project ions A graphic user interface (Fig.3 2) coded with Visual C ++ 6.0 is used to control the entire data acquisition. The software contains a host of device control features and is designed to maximize the flexibility of image acquisition and analysis. Motori zed linear stage control and rotator control help improve precision and repeatability of data in experiments. The scanner controller and CCD camera are synchronized, so the sampling can be controlled to ensure that when the camera is actively acquiring a f rame, sample, laser and CCD are all remain still. Some routine experiments are programmed so that only human intervention is to initiate the experiment by hitting mouse on start button. E ach procedure is fully automated and streamlined to ensure fastest speed to obtain tomography images after signal sampling. In some extreme experiments with low light level, t o reduce systematic errors, baseline measurements need to be performed once on a homogeneous phantom in the absence of fluorophores .105 Thus the ratio of the baseline measurements to calc ulated forward simulation data with the same homogeneous geometry, multiplying the measured data from the inhomogeneous phantom s, served as the true input for both DOT and FMT reconstruction.

PAGE 49

49 Shape Extraction f or Arbitrarily Shaped Object s It is our aim to develop a noncontac t system and able to accurately locate position and concentration of fluorophore accumulated on the organ or tumor of live mouse in its nature state. To improve the accuracy on shape extraction of live mouse and simplify experimental pr ocedure, shapefrom silhouett e (also referred as visual hull ) method is implemented to render 3D model of sample. The 3D model is utilized in the whole experiment data process pipeline: finite element mesh generation, mesh optimization and post process, data mapping and extraction from raw experimental data, 3D fluorescent imaging inverse reconstruction algorithm and image registration method for MRI and CT imaging A rubber alligator is used to demonstrate the shapefrom silhouette method for 3D modeling o f arbitrarily shaped object s Camera C alibration Camera calibration often referred to as camera resectioning, is a way of examining an image, or a video, and deducing what the camera situation was at the time the image was captured. Camera calibration is u sed primarily in robotic applications, and when object model virtually based on real input. Rendering programs are all based on a virtual camera. In order for the modeled objects to be the equivalent of the real objects, we need to make sure that our virtual camera is in match with our real camera when we shot the photograph. Camera calibration achieves this and deduct s where the real camera was relative to the scene. Our system uses Tsai's camera model which is based on the pinhole model of perspective projection. Given the position of a point in 3D world coordinates the camera model predicts the position of the point's 2D pixel coordinates. 11parameter pinhole camera model is used to describe our CCD The internal parameters describe how the camera forms an image while the external parameters

PAGE 50

50 describe the camera's position and orientation in the world coordinate frame. Calibration data for the model consists of 3D (x,y,z) world coordinates of a set of points and corresponding 2D coordinates (Xf,Yf) (typica lly in pixels) of the feature point s in chessboard photo. In Tsais model following 11 parameter s fully determine a virtual camera.106 5 intrinsic camera parameters: f equivalent focal length as in the pin hole camera model 1st order radial lens distortion coefficient Cx, Cy coordinates of center of field of view uncertainty factor for experimental system error and situation s 6 extrinsic came ra parameters: Rx, Ry, Rz, Tx, Ty, Tz rotational and translational components in the world's coordinate frame Cx, Cy do not need to be experimentally determined, we can obtain the value from camera parameter for Princeton 1024* 1024 CCD with 13um pixel size, when bin =4 Cx=Cy=128, for the field of view close to lens center =0 =1. F, Rx, Ry, Rz, Tx, Ty and Tz are determined using calibration data obtained from chess board method. 106 We define the center of chess board picture is zero and calculate each corners coordinates based on the true side length of each square. The square corner position is estimated by eye and then the subpixel accu racy is obtained determined by corner finder algorithm b ased on autocorrelation method ( Harris corner finder method). 107

PAGE 51

51 Multi Camera Calibration s To get 3D model with mil limeter accuracy, 72 projection ( one projection every 5 degree) silhouettes are used to rend 3D sample contour. Therefore we need to do camera calibration for each angle of all 72 projections. Calibrate same camera in different project separately usually r esult in different focus length and scale factor value due to non uniqueness of optimization problem (>10 parameters) when using imperfect experimental data. In experiment s, since the CCD rotation center and sample rotation center is not perfectly centric and CCD focus plane is not perfectly parallel to calibrate plane it is impractical and tedious to calibrate CCD camera for all 72 angles. In order to obtain virtual CCD parameter to describe our CCD camera for all 72 angle projections by modifying only ro tation angle of y axis ( R y) value of one virtual camera, the multi camera calibration is implemented through following procedure: 1. Find the camera rotation center If the center of chess board is the rotation center of the CCD camera, center of chess board corners should have same pixel value 128 for different angles of view (256*256 region of interest). Z direction of chess broad can be move d forward or backward for better center alignment For tiny objects like Drosophila pupae, the sample stage is a ne edle tip. As long as the needle tip is in the center of image and remains the same pixel value while the needle tip spin, the needle tip is right in center position. 2 Mount chessboard pattern on the stage, the center of chessboard is in the center of rota tion 3.6mm square is used for big macrolens 0.7mm square is used for small microlens. The coordinates is defined as in Fig 3 3. Calibrate 30, 0,30 degree CCD camera and obtain 3 set of calibration parameters. Use each of this 3 calibration parameter to generate CCD model for 30 0 30 degree by

PAGE 52

52 modifying R y and predict CCD pixel positions for all three positions .The best calibration parameter model among the three model set gives minimum error in predicted pixel s of 30, 0,30 degree. Fig. 3 4 shows the raw input data of calibration data and predicted pixel positions of corners from ca librated CCD model at 0 degree. 3 Modify Ry value ( New Ry value of n ew p osition =Ry of chosen CCD model position ( /6,0, /6) + Radius difference between the new position and chosen CCD model position) of the best camera model in accordance with 72 different angle positions of the CCD camera. If CCD camera and sample stage are always fixed, the calibration is a onetime only process. In addition, camera calibration model is th e foundation that determines performance of all following data process Theref ore multiple repeated calibrations and extra effort are worthwhile to obtain the best camera model for following procedures. Visual Hull Method Visual hull is a geometric entit y created by shapefrom silhouette 3D reconstruction technique. This technique assumes the foreground object in an image can be separated from the background. Under this assumption, the original photos can be convert ed into a foreground/background binary s ilhouette image by defining a threshold. The foreground silhouette is the 2D projection of the corresponding 3D foreground object. Along with the camera viewing parameters, the silhouette defines a back projected generalized cone that contains the actual o bject. T he visual hull is the maximal object that has the same silhouettes of all projects as the original object ,108, 109 Visual hull algorithm has been explored extensively recently and has three major categories: volumebased (e.g. voxel carving methods ),110 polygonbased (e.g.. polyhedral visual hull ) 111 or imagebased 112 and m odel fitting helps t o obtain a

PAGE 53

53 smoother result .113 The basic principle of algorithm is implemented through following steps: First, co n vert digital photo to binary silhouette images of each projection by defining a threshold for boundary. Second, t he true camera projection position (note: position for the virtual camera model not the camera position) is determined through camera calibrations. Compute camera projecti on matrices according to determined coordinates. Third, the space of interest is divided into discrete voxels. Test every voxel by projecting it into image planes defined by projection matrices of all camera projection. I f the projected point is contained in the silhouette for all 72 camera position s, the voxel is inside the visual hull. The union of all the vox els tested be inside visual hull is the 3D model of the objects The resolution of visual hull is determined by the voxels size and could be improved by adding more camera projections. As Fig. 3 5 shows: the overlap of three camera projections is the visual hull of the sample and t he visual hull can be refined by adding more camera posit i ons We use a rubber alligator to access our visual hull method and a mock FMT experiment is used to test performance. As Fig. 3 6 and 37 shows the alligators shape is well reconstructed and the detector and source positions are mapped and the error is less than 1mm. The major drawback is that concave surfaces cannot be reconstructed very well. Fortunately, very few regions of the small animals used for imaging actually are concave ( i.e. the regions under the forearm and hind legs) and no current research f ocused on these convexity parts When the complete (convex and concave) surface

PAGE 54

54 areas are needed, more advanced surface extraction approaches should be used such as 3D surface cameras 114, 115 or spatially m odulated illumination patterns. Because DOT and FMT has a resolution about 1~5 millimeters and the variations of t he order of the mean free path do not have an impact on the reconstructed images. Therefore, Visual Hull approximation can provide adequate accuracy. In addition, this means that while the surface may change during silhouette acquisition due to breathing, an average surface rendered by 72 camera position is accurate enough for FMT or DOT. Free Space Data E xtraction Model Detector Model Arbitrarily shaped objects need extra process since CCD collect signal from image plane instead of sample boundary. Ntziachristos et al proposed a model to collect light intensity signal from arbitrarily shaped objects and experimentally testified by reconstructing a fluorescent target with normalized Born approximation method. 35, 116 ) 6 ( S d ) r r ( ) r ( J 1 ) r ( Jd s n d det ) 7 ( S r dA cos cos r r ) sin NA ( f ) r r ( ) r r (d 2 d d d d The focus plane of CCD is treated as 2D virtual pixel detector array with certain number aperture (NA) W hen small aperture assumption is met ,) (det dr J is the actual signal obtained by each virtual detector on CCD focus plane. As we can see, ) ( r Jn is related to ) (det dr Jby above equation ( 6 ) and ( 7 ), under some simplification ) (dr r can be calculated for corresponding experimental situation in Fig 3 9 :

PAGE 55

55 ) sin (dNA f : The NA is considered through this Gaussion function, which modeled as a Gaussian of full width at half maximum of 2NA, NA is determined by parameters of lens CCD and actual experimental dimension situation s. ) (dr r : define the visibility or directivity calibration factor and discard surface points not visible from the detector. It is determined by projection area on the sample surface of each virtual detector on focus plane, detection area dA and experimental calibration factor. The value can be set as 1 for most experiments. 1/ 2r rd : solid angle term r rd is the distance between specific surface detector node and correspondi ng virtual detector on focus plane. cos : Lambert's cosine dependence term (Fig 3 8) is the angle between the normal vector of surface node and normal vector of virtual detector on focus surface d is the angle between light path and the virtual detection surface normal vector, in most in focus case, d can be treated as 0. Source model When an arbitrarily shaped object is measured, each source have different incident angle which results in various reflection lost so true photon density of each sour ce that propagated into the object should be calculated individually. As in Fig. 3 10: i t t i i i i tn n n E E t cos cos/ ) cos( 2 /0 0 || (8) These equations are called the Fresnel Equations for parallel polarized light. t t i i i i i t n n n E E t cos cos / ) cos( 2 /0 0 (9) These equations are called the Fresnel Equations for perpendicularly polarized light.

PAGE 56

56 In scenario of our experiments finite element mesh of the arbitrarily shaped object has absolute coordinates centered at rotation center of rotator stage. For each node, surface normal vector towards free space and distance from focus plane of CCD lens could be obtained Numerical aperture and visibility factor were determined for each virt ual CCD detectors based on experimental situation and manufacturing parameters of the lens and CCD camera. Thereafter, solid angle value of each virtual detector on the CCD focus plane and Lambert's cosine law were used to convert detectable CCD signal int o photon density of each surface detector node on t he finite element mesh With known surface normal vector of each source and Fresnel equations, light reflection loss of each source was calculated to obtain actual photon density which was used as assigned light intensity value of each source node on the finite element mesh Implementation in FMT S ystem The complete experimental procedure is described in Fig 3 11. The DOT guided FMT for arbitrarily objects is established and tested. The data acquisition and processing programs are streamlined for maximum repeatability. The well known shape from silhouette 3D recovery approach was used to render th e 3D model of the actual sample used. 117 In this method, the CCD camera was modeled as a pin hole camera positioned in a coordinate system where its intrinsic and extrinsic parameters (i.e., space projective transformation matrices) were calibrated using the widely used chess board method, 118 allowing the calcu lation of the extrinsic parameters for a new position while the intrinsic parameters stayed the same for all the positions. A total of 72 projection images (one projection every 5 by rotating CCD 360) were used in our calculation giving a submillimeter r esolution. The projection images were then converted into foreground/background binary silhouette images (the 2D

PAGE 57

57 projections of the 3D foreground object). Each silhouette defined a back projection of a cone that contained the geometrical information of the actual phantom. Thus the overlapped volume of all 72 cones defined by silhouettes gave the 3D visual hull model of the arbitrarily shaped phantom A photoluminescent plate was used as the background screen which provided evenly distributed illumination an d good contrast. A 3D finite element mesh of the 3D model was then generated with the point cloud obtained by visual hull method: use point wrap function in Amira 3.1 to obtain the surface. The surface need to be post processed to obtain a good uniform sur face which is the prerequisite of good mesh generation to minim ize mesh effect on final tomography reconstruction. The process includes: 1, F lip the edge s in triangle element which have aspect radio (maximum length/minimum length) bigger than 4 ; 2 Remove dihedral angle below 60 degree; 3 Remove coplanar face; 4 Remove all intersection to ensure a closed surface After these post processes, the surface is smoothed and refined. We can simpl if y the refined mesh to any desired face number based on desired me sh node numbers (for example, a 3000 faces surface usually can generate a 4500 nodes 3D mesh). The compute tetragen function in Amira 3.1 can be used to generate 3D finite element mesh and optimized the mesh quality. The generated mesh is optimized to de crease half band width for a fast computation. There is also a need to map the photon density read from the CCD camera (actually the virtual detector along the focal plane of the CCD camera) onto the arbitrarily shaped surface of the phantom. We have adopt ed a method developed by Ntziachristos et al 35, 116 that was able to realize the mapping accurately in their FMT studies. In this method, a Lambert's cosine law and solid angle based light propagation

PAGE 58

58 model is est ablished to correlate the photon density at the virtual detector and that at the animal surface. With this model, given the relative geometric relationship between the virtual detector and the phantom surface, the numerical aperture of the virtual detector s and the visibility factor that can be experimentally calibrated, one can accurately convert the read out from the CCD into the photon density at the phantom surface for tomographic reconstruction. Figure 3 1 DOT guided FMT experiment system

PAGE 59

59 Fig ure 3 2 G raphic u ser i nterface of the system

PAGE 60

60 Figure 3 3 Coordinates system definition: t op view from observe point above the sample stage and Y axis direction is from paper internal to outside. A Figure 3 4 Use calibrated CCD m odel parameters and universe world coordinate of grids to predict pixel values of corner s in the image A) original input data for CCD calibration, B) pixel positions predicted by CCD model obtained by CCD calibration B

PAGE 61

61 Figure 3 5 Vi sual hull scheme demonstration. The intersection of the silhouette projection cone (shown in pink, blue and green) is a cross section of the visual hull. Additional silhouette photos from new viewpoints will further constrain the intersection region, carve and refine 3D visual hull.

PAGE 62

62 Fig ure 3 6 Original r ubber alligator and its 3D finite element mesh. F inite element mesh is generated based on 3D dot contour obtained by visual hull method. Fig ure 3 7 Source s (red dots) and detector s (blue dots) of dif ferent projections.

PAGE 63

63 Figure 3 8 Lambert's cosine law Figure 3 9 Scattering geometry for a diffusive object of volume V surrounded by air for free space model 116

PAGE 64

64 Fig ure 3 10. P hoton density model of source on the air /sample interface

PAGE 65

65 Figure 3 1 1 Outline of e xperimental p rocedures

PAGE 66

66 CHAPTER 4 PHANTOM EXPERIMENTS Algorithm s and hardware system are evaluated by using regular shaped and arbitrarily shaped 3D phantoms. Targets of arbit rarily shaped phantoms also mimic mouse anatomy structures as in vivo situation Indocyanine green (ICG) is the only FDA Approved dye with strong optical absorption in the near infrared (NIR) region, where light can penetrate deepest into biological tissue ICG is well characterized with quantum yield data and fluor ophore absorption coefficient m xa In our phantom experiments ICG is used to evaluate our DOT guided FMT algorithm. According to Beers law, fluorophore absorption coefficient m xa is proportional to fluorophore concentration since quantum yield is stable at low concentration in most preclinical applications. Here we present a systematic study using phantom experiments under the condition of heterogeneous s and a distributions for fluorescence image reconstruction. In our approach, we directly reconstruct optical heterogeneities using DOT and apply reconstructed s and a dis tributions to the FMT reconstruction. Both DOT and FMT reconstructions are conducted using a noncontact multi angle transmission scheme, coupled with finite element reconstruction algorithms. Our experimental results suggest that the optical heterogeneous nature of the target itself, especially when its size is relatively large, must be considered for quantitatively correct FMT in the framework of finite element based reconstruction methods. We note that several methods (CT, MRI and PAT) to address heterogeneity issue in real experimental study have been combined with FMT .119122 G uided methods like

PAGE 67

67 MRI and CT also can only indirectly reveal optical property by heuristic or anatomical assignment. PAT can provide distr ibution of absorption coefficient but it cannot provide the distribution of scattering coefficient. This information is critical for quantitative FMT as it was demonstrate d in our former experimental results28 and Milstein et al s simulation results .101 While it is true that PAT offers better spatial resolution than DOT, its ability of tissue penetration is not as good as DOT. In addition, all optical based DOT guided FMT can be conveniently implemented wit hout adding extra hardware, while additional ultrasonic detection/data acquisition is needed for PAT/FMT combination. Regular Shaped O bject s We have performed a series of regular phantom experiments ( Fig. 4 1) to evaluate the merits of considering the heterogeneous a and s distributions in the fluorescence reconstruction. Reconstructed optical and fluorescence images for a representative 2D case is given in Fig. 4 2 while the recovered m xa values from a number of 2D cases having different ICG concentrations are presented in Fig. 43 Validation of 3D image reconstruction algorithm is given in Fig. 4 4 for a representative 3D case. In the representative case, the 6mm diameter target contained 1 M ICG and was positioned at ( 5, 6). The DOT recovered a and s distributions are given in Figs. 4 2 (a) and 4 2 (b). Fluorescence image is reconstructed under four scenarios: DOT recovered a and s distributions ( Fig 4 2 (c)), uniform s but DOT recovered a distributions ( Fig 4 2 (d)), uniform a but DOT recovered s d istributions ( Fig 4 2 (e)), and uniform aand s distributions ( Fig 4 2 (f)). Examining the images shown in Fig 4 -

PAGE 68

68 2 it can be seen that when the DOT recovered a and s distributions are used, the fluorescence image is most accurately reconstructed in terms of the size/shape and m xa value of the target. We can clearly see that the uniform aand s assumption re sults in the worst m xa reconstruction ( Fig 4 2 (f)), where the recovered target size is significantly overestimated, and the recovered m xa of the target has about 75% error relative to the literature value of 0.03. We can also see that the s distribution plays a critical role in determining an accurate FMT reconstruction. With heterogeneous s but uniform a distributions ( Fig 4 2 (e)), the recovered image quality is better and the reconstructed m xa value is closer to the literature value( m xa =0.03), compared with that using heterogeneous a but uniform s distributions as shown in Fig 4 2 (d). This finding is consistent with our simulations and the simulations conducted by Eppstein et al .31 For fluorescence reconstructi on, the aand s distributions at the excitation wavelength (785nm) were used to interpolate their values at the emission wavelength (830nm) using the absorption spectrum data available in the literature. 123126 Based on the literature, 127 the quantum efficiency of ICG is stable u nder the condition of low ICG concentrations as used in our experiments. Thus, the reconstructed fluorescent yield (m xa) under different ICG concentrations should be solely determined by m xa which is directly proportional to the ICG concentration according to the Beer Lambert Law m xa value was also used to provide a basis of comparison between calculated m xa and that obtained via spectroscopic methods 91, 128 In this work we used a

PAGE 69

69 quantum yield value of 0.016 which was measured under nearly the same experimental condition as in our experiments .91 Quantitative performance of our reconstruction approach is further evaluated us ing the 6mm diameter target containing different ICG concentrations (0.1, 0.2, 0.4, 0.6 and 1.0 M ). In the image reconstructions, the m xa values (4x1012mm1) of the background phantom were used as the initial guess for all these experimental cases. Reconstructed m xa values with and without DOT recovered a and s distributions are plotted in Figs. 4 3 where values from the literature are also presented for comparison. For better comparison, the m xa value at 0.1 M from the literature was used to calibrate the m xa value at 0.1 M from the FMT reconstructions. It is clear that the results with the DOT recovered optical property distributions are in good agreement with that from the l iterature, whereas the recovered m xa value, with the uniform optical property distribution assumption, is significantly away from the exact value and basically not quantitatively correct, compared to that of spectroscopy method. While the 2D reconstruction described above is simple, computationally fast and yet able to provide reasonably accurate results, we have also implemented the DOT guided FMT approach in 3D and tested it using several phantom experiments. 3D cases results are consistent with results of 2D cases. In the 3D phantom experiments, a 6x6mm cylindrical solid target containing ICG was embedded in a 30x30x30mm cubic back ground. A total of 260 source and 260 detector positions were used for image reconstruction with a finite element mesh of 4354 nodes.

PAGE 70

70 2D Experiments Experiments were conducted using a continuous wave (CW) 785nm diode laser at 65mW for the FMT experiments and at 2mW for the DOT experiments It takes less than 1 minute (Integral time 50ms per source) for a set of excitation light data and less than 10 minutes (Integral time 5s per source) for a set of emission light data. Since experiment running requires CCD X axis linear stage, Y axis linear stage, sample rotator, CCD rotator works together and each one need work with respect to status of others. System is synchronized to make sure every move of each component is desired and happens on specific time window. In the experiments, the laser beam was focused directly onto the sample to serve as a point source. The focal plane of the CCD represented a collection of virtual detectors. The excitation source positions at the phantom surface were determined precisely by the X, Y linear stage, while the detector positions for both the excitation and emission light collection were accurately with an accuracy of coordinate mapping of 0.25mm. For each set of transmission imaging data, we used 25 detectors and 25 sources that covered a 24x24mm central area of a 30x30mm phantom. A finite element mesh with 1186 nodes was used for all the image reconstructions. In the phantom experiments, a 6mm diameter solid cylindrical target containing ICG was embedded in a 30x30x90mm sol id cuboid, mousesize background phantom. The background phantom was composed of 1% Intralipid, India ink and 1% Agar powder, providing a s of 1.0/mm and a aof 0.005/mm. The target had a s of 2.0/mm and a was contributed by both the India ink (0.005/mm) and m xa of ICG. To test the

PAGE 71

71 quantitative performance of the system, we prepared a series of target s containing low concentration ICG (0.1uM 0.2uM, 0.4 uM, 0.6uM and 1uM). 3D Experiments In the experiments, the laser beam was focused directly onto the sample to serve as a point source. The focal plane of the CCD represented a collection of virtual detectors of area dA= 1mm2. 7 layers of sources with averagely 13 sources at each layer were used. Signals from a total of 363 source (corresponding to 4 CCD positions/angles) and 363 detector positions were collected transmission data for both DOT and FMT We used a finite element mesh of 4598 nodes and 22539 tetrahedron elements. R egular shaped DOT guided FMT q ualitative and quantitative performance improvements for a representative case (1uM ICG in the target) as we can see in Fig. 4 4 where the exact fluorescent target (red), the recovered fluorescent target with (gold) and without (blue) DOT guidance are fused together for easy comparison. Examining the images shown in Fig. 4 4 it can be seen that when the DOT recovered a and s distributions are used, the FMT imag e is most accurately reconstructed in terms of the size/shape, position (partially overlapped with the exact position) and m xa value (0.026 mm 1) of the target. We can clearly see that the uniform a and s assumption results in unsatisfied m xa reconstruction, where the recovered target shape is distorted, the recovered m xa value of 0.01 mm1 has about 67% error with respect to the exact value of 0.03 mm 1, a nd the target position is shifted. Arbitrarily S haped O bject s Herein, we first demonstrate the combination of CW FEM based DOT/FMT with 3D freespace noncontact detection and fully consider the distribution of both

PAGE 72

72 absorption and scattering coefficients with arbitrarily shaped phantom geometries. In addition, the phantom mimics a realistic in vivo (mouse) anatomy heterogeneous optical situation and animal boundary shape The experiments were conducted using an optical fib er free system based on a non cont act multi angle transmission scheme, and the 3D fluorescence images were recovered using our finite element based FMT reconstruction algorithm with DOT guidance. Our approach is demonstrated using a series of phantom experiments with low ICG concentration (0.1, 0.2, 0.4 and 1.0 M ) targets. In the phantom experiments, a 6x6mm cylindrical solid target containing ICG was embedded in a mouse size arbitrarily shaped phantom. The background phantom was composed of 1% i ntralipid, I ndia ink and 1% a gar powder, providing a s of 1.0/mm and a a of 0.03/mm. Five optical heterogeneities mimicking lung, heart, liver and stomach in a mouse were embedded in the background and their optical properties are listed in Table 4 1 according to the literature.129 The stomach also served as the fluorescent target containing ICG at a variety of low concentrations (0.1, 0.2, 0.4 and 1 M ). Based on the literature,127 the quantum efficiency of ICG is stable u nder the condit ion of low concentrations as used in our experiments. Thus, the reconstructed fluorescent yield ( m xa ) under different ICG concentrations s hould be solely determined by m xa which is directly proportional to the ICG concentration as the Beer Lambert Law states. Recons tructed m xa value was compared with m xa obtained via spectroscopic methods 91, 128 In this work we used the reported quantum yield, =0.016 from the lit erature which was measured under nearly the same experimental conditions as in our experiments .91

PAGE 73

73 Reconstructed optical and fluorescence images for a representative case are given in Fig. 4 6 : a s and m xa images are given in Fig. 4 6 (b), Fig. 4 6. (c) and Fig. Fig. 4 6 (d) respectively. W hile the recovered m xa values from a number of cases having different ICG concentrations are pr esented in Fig. 4 7 For the representative case, the phantom the exact position, shape and size of the optical heterogeneities are shown in Fig. 4 6 (a) and table 41 while the recovered a s and m xa images are given in Fig. 4 6 (b), Fig. 4 6 (c) and Fig. 4 6 (d) respectively. In Fig. 4 6 (d), the exact fluorescent target (orange), and the fluorescent target reconstructed with (yellow) and without (blue) DOT guidance are fused together for easy comparison where the isosurface plot of a image is also depicted in black mesh. Examining the images shown in Fig. 4 6 (d), it can be seen that when the DOT recovered a and s distributions are used, the FMT image is most accurately reconstructed in terms of the size/shape, position (partially overlapped with exact position) and m xa value (0.025 mm1) of the target. We can clearly see that the uniform a and s assumption results in unsatisfied m xa reconstruction, where the recovered target shape is distorted, the recovered m xa value of 0.007 mm1 has about 80% error with respect to the exact v alue of 0.03 mm1, and the target position is significantly shifted. Reconstructed m xa values with and without DOT recovered a and s distributions for target containing different ICG concentrat ions are plotted in Fig. 4 7 Values determined by spectroscopic method from the literature are also presented for

PAGE 74

74 comparison. It is clear that the results with the DOT guidance are in good agreement with that from the literature whereas the m xa values recovered with the uniform optical property distribution assumption is shifted significantly away from the literature values. T able 4 1 Optical properties of the embedded organs and the background used in the experiments Lung Heart Liver Stomach ( also the ICG Target ) Background a (mm 1 ) 0. 3 0. 1 1 0.45 0.2 0.03 s (mm 1 ) 2. 5 1.1 2.5 1.8 1.0 Fig ure 4 1 Photograph of the CCD based CW FMT system.

PAGE 75

7 5 Figure 4 2. Reconstructeda, s and m xa images for a representative experimental case. A ) aimage B ) s image C ) m xa with DOT recovered a and s distributions D ) m xa with uniform s but DOT recovered a distributions, E ) m xa with uniform a but DOT re covered s distributions, and F ) m xa with uniform aand s distributions. A B C E F D

PAGE 76

76 Figure 4 3. Reconstructed m xa values in the target with and without DOT recovered a and s distributions when different ICG concentration was used. The m xa value from literature (0.1,0.2,0.4 and 0.6 M ) were obtained by spectroscopic methods,21 while the m xa literature value (1 M ) was obtained by a micromolar aqueous solution with a spectrofluorometer [8].

PAGE 77

77 Fig ure 4 4 Reconstructed 3D images for a representative case (ICG concentration in the target=1 M ). Figure 4 5 Arbitrarily shaped phantom experiment. A ) Phantom and imaging system. Inclusion is the arbitrarily phantom. B ) Raw boundary signal A B

PAGE 78

78 Figure 4 6 Exact positions of the targets in the finite element mesh A ), and reconstructed a B ) s C ), and m xa D ) images for a representative case (ICG concentration=1 M ). In (d), the exact fluorescent target (orange), and the fluorescent target recovered with (yellow) and without (blue) DOT guidance are all shown. The insert is a close view of the recovered fluorescent target with and without DOT guidance relative to the exact target. Here the isosurface plot of the absorption image is also depic ted (black mesh).

PAGE 79

79 Figure 47 Reconstructed m xa values in the fluorescent targets with and without DOT guidance when different ICG concentration was used. For easy comparison, the recovered data shown were calibrated against the m xa value at 0.1 M from the literature. Error bars are the deviation of each concentration point value and individual concentration point values are the mean value of five different experiments. The significance of differ ences was assessed between data with DOT guidance and data without DOT guidance using t test, with DOT guidance method (all pvalues less than 0.05).

PAGE 80

80 CHAPTER 5 APPLICATION IN A MOU SE MODEL Diffuse optical tomography (DOT) guided fluorescence molecular tomography (FMT) is utilized to provide quantitative analysis in our breast cancer preclinical study. In this method, we conducted a full body optical property reconstruction of mice a nd applied the reconstructed optical property into FMT reconstruction as functional a priori information to minimize the heterogeneity effects. First, we conducted a well controlled experiment in which the volume, quantity and location of the tumor cells a re known to validate quantitative and qualitative performance of the method in mice In this experiment, mouse mammary tumor 4T1 cells were prelabeled with a tumor targeting peptide conjugated with a near infrared fluorescence (NIR) dyes, Cy5.5. Following injection of different numbers of the NIR dyetargeting peptidelabeled living cancer cells into three subcutaneous locations of the mice FMT was performed on the mice. Results of FMT reconstruction show that cell quantification and tumor localization are improved with DOT guidance. We further applied this method to evaluate target specificity and detection sensitivity of a newly synthesized NIR dye (NIR 830) labeled urokinase plasminogen activator receptor (uPAR) targeted magnet iron oxide nanopart icle s (IONPs) after systemic delivery Our results show the signal intensity in the orthotopic mammary tumor in mouse that received NIR 830 dye labeled and uPAR targeted IONPs is 10 times higher than that of the mouse injected with NIR 830 dye labeled and non targeted mouse serum albumin (MSA) IONPs. Introduction Theoretically propagation of both excitation and fluorescent emission light in tissue is described by the coupled diffusion equations where the emission photon

PAGE 81

81 density of the fluorophores is dir ectly related to the excitation photon density, which in turn depends on the optical property distribution in tissue. Research conducted by several groups has shown the impact of optical property distribution 97, 98, 101 and methods have been reported to improve image quality of FMT in the presence of heterogeneities FMT in phantom study 98, 103 FMT is already proved as a promising tool to trac k cancerous tissue in mice .4, 59 I n these studies, inhomogeneities in tissue are usually treated by normalizing the excitation signal towards emission signal via normalized Born approximation method.103 The normalized Born method uses original boundary data to minimize the error resulted from uniform optic al property assumption. Heterogeneities of mouse can also be better evaluated for FMT reconstruction with a priori MRI or CT information 45, 130 To our best knowledge, whole body optical heterog eneities of mice are not systematically evaluated by directly using diffusion optical tomography. Diffusion equation based full optical property reconstruction can reveal information hidden in raw experimental data and further improve the FMT reconstruction. The resolution of the reconstructed heterogeneity is not as good as MRI or CT method, but DOT can directly reconstruct optical property and it is easily implemented into FMT system with zero extra cost, requiring only few extra minutes to regular FMT experimental procedure. Our method has proved be an efficient and practical method to improve FMT imaging reconstruction for both regular and arbitrarily shaped phantom experiments in presence of heterogeneous reduced scattering coefficient ( s ) and absorpt ion coefficient ( a ) distributions .28, 96 Here we first applied DOT guided FMT method and evaluate its performance in mice. To help evaluate the method qualitatively and quantitatively, we label ed the living 4T1 breast tumor cells with

PAGE 82

82 amino terminal fragment (ATF) peptides conjugated to Cy 5 .5 dye and then injected different amount of tumor cell (100000,200000 and 500000) in three marked positions on the mouse. The results proved that with DOT g ui dance, tumor cell quantification and localization are both improved. We further applied proposed method to evaluate affinity and sensitive limit of a newly developed nanoparticle NIR830 ATF IONP with tumor cell targeting amino terminal fragment (ATF) The ATF modified dye and pure dye are injected in tail of live mouse induced with breast tumor. Based on DOT guided FMT results, the affinity of ATF modified dye toward tumor cell is increased ten folds Further FMT test indicated that NIR830 ATF IONP ca n localize trace mount of cancer tumor cells (r ecurrent tumor and metastasis cases). Method and Experiments Data acquisition and process procedure were described in detail in our previous phantom studies 28, 96 Bri efly, for each angle, the source and detector nodes in the finite element mesh were automatically chosen to cover the region of interest and absolute coordinates of detectors were mapped automatically towards CCD pixel grid. For each angle, XY positioner d elivered laser beam according to coordinates of source positions in 3D finite mesh and CCD collected transmission signal of the opposite si d e. For example, in tumor cell quantification experiments, signals from a total of 238 source positions (corresponding about 60 per side in all 4 projections by rotating mouse 0,90,180 and 270 degree) and 284 detector positions on opposite side were collected. Mesh is generated according to the shape of the mice and we used a finite element mesh of 3855 nodes and 17999 t etrahedron elements.

PAGE 83

83 3D DOT R econstruction of Mice Reconstructed optical property images of a typical mouse are given in Fig. 2 and isosurface is plotted using 70% of the maximum reconstructed value. The 3D absorption and scattering images show that the lu ng (upper part ) and liver (lower part) are reconstructed with high absorption and scattering value. We note that the accuracy of reconstructed absorption and scattering coefficient is critical for the DOT guided FMT method presented here. We have made gr eat efforts in improving our DOT reconstruction algorithm and experimental system to obtain accurate optical property reconstruction. It has been demonstrated in recent years by several groups including our own 9295 that quantitative reconstruction of both absorption and scattering coefficients is possible using CW DOT when a priori information coupled with effective normalization schemes are used. In our CW DOT method, in addition to the use of a priori information and normalization scheme, the hybrid regularization schemes of Marquardt and Tikhonov play an important role in combating the ill posed problem involved.87 Our previous phantom study shows that with DOT guidance, statistically we can obtain better FMT reconstruction in mouse mimicking phantom experiments .96 Quantification of Cy 5.5 ATF L abeled T umor C ell According to Beers law, fluorophore absorption coefficient m xa is proportional to fluorophore concentration since quantum efficiency is stable at low concentration in most preclinical applications I n FMT reconstruction algorithm, m xa value is determined by light intensity distribution, optical heterogenei ty distribution and boundary data, which are all related to heterogeneous absorption coefficient ( a ) distributions and reduced

PAGE 84

84 scattering coefficient ( s ) distributions. Since mice have highly heterogeneous optica l property in different organs and tissues, same concentration of dye in different parts of mouse body will be reconstructed with different m xa value when optical property distributions are treated uniform. In addition, DOT reconstructi on itself can provide additional information to reveal tumor position since high a value correlates high blood volume which can result from angiogenesis of the late stage tumor and high s value reveal tumor with big particle size distribution since tumor cells/nuclei are considerably enlarged relative to normal ones .131, 132 Therefore, DOT reconstructed reduced scattering coefficient ( s ) and absorption coefficient ( a ) distributions are introduced in our study to help obtain reliable quantitative analysis in FMT. The recombinant amino terminal fragment ( ATF ) of uPA was produced from a bacterial expressing system using our established protocol ATF peptides were then labeled with Cy5.5 dye and used to label breast cancer 4T1 cells. After labeling, the intercellular dye and nonspecific binding are washed thoroughly. C ells were counted and specific numbers of cells (top:100,000 cells, middle:200,000, bottom:500,000) were in j ected into the haired Balb / C mouse for optical imaging. E xperime nts were conducted using a continuous wave (CW) diode laser at 660 nm (25mW) as excitation and 710 nm band pass filter for emission light acquisition The injected tumor shape may vary depending on the anatom ic structure of the injection site but its volume is remained approximately same and proportional to tumor cell number. I sosurface of the reconstructed fluorophore absorption coefficient m xa is used to present the 3D rec onstruction where the volume of the isosurface indicat es the

PAGE 85

85 tumor cell amount. The reconstructed tumor images are given in Fig. 3 where it can be seen that when the DOT recovered a and s distributions are used, the fluorescence image is most accurately reconstructed in terms of the volume ( which is proportional to tumor cell number) and position (partially overlapped with the exact position). We can also see that the uniform a and s assumption results in unsatisfied m xa reconstruction where the recovered target position s are shifted from the exact positions and the recovered size s are not proportional to their tumor cell number. We also noticed that DOT guided FMT may somehow increase the sensitivity in tumor imaging, the 100,000 cells in up position is not revealed by FMT without DOT guidance. Evaluation of A ffinity Oxy hemoglobin and deoxy hemoglobin have same absorption at about 790nm and the sum of absorption of oxy hemoglobin and deoxy hemoglobin reach minimum at about 800 nm in visible light region. We developed NIR IR 830ATF IONPs (excitation at 785nm and emission at 830nm) to improve preclinical mouse imaging quality. T he NIR 830 dyes can yield much higher signal/noise ratios with less absorption than Cy 5.5, and is also much stable than indocyanine green (ICG). We here use DOT guided FMT to characterize NIR830 dyeATF IONP m xa is reconstructed by DOT guided FMT and is used to quantify affinity. Quantification of the affinity in the mouse can improve specificity and facilitate probe development. A ffinity are normally measured in vitro by using confocal scanning laser microscopy, quartz crystal microbalance (QCM), atomic forc e microscopy (AFM) and surface plasmon resonance (SPR) 133135 Planar or FMT based whole body

PAGE 86

86 monitoring methods are also reported. 136137 These methods usually can just provide qualitative or semi quantitative analysis. The affinity to targeted organs or tumors in the mouse is the key parameter to evaluate developed probe in preclinical applications. Although quantitative measurement of the adsorption kinetics and affinity in a liv e mouse is extremely challenging the variation in affinity is the frequently used substitute index for most applications W e tentatively utilized FMT to quantify affinity of a newly developed NIR830ATF IONP by comparing reconstructed m xa value of NIR830 IONP dye with and without ATF. We administrated un targeted NIR 830 MSA IONP and NIR830 ATF IONP via the tail vein of the tumor bearing mice DOT guided FMT reconst ruction, at the exact same measurement and reconstruction settings was used. The ATF targeting greatly increase s the signal intensity in the tumors over nontargeted, NIR 830 dyeMSA IONP with a difference over 10fold as shown in Fig 5 4. Th e high specific affinity of NIR830 ATF IONPs can greatly increase the sensitivity thus makes challenging cases like early stage tumor, recurrent tumor and metastasis detectable. To confirm whether NIR 830ATF IONPs dye can help detect spontaneous recurrent tumors and lung metastasis of breast cancer, we excised the original breast cancer tumor. After the mouse recovered, both the planar fluorescent (IVIS system) and FMT measurement detected metastasis and recurrent tumor with consistent results.

PAGE 87

87 Conclusions DOT guided FMT can help to obtain more accurate quantitative and qualitative results in a well control led tumor cell count experiment using Cy5.5 ATF nanoparticle dye. We further used the method to characterize NIR830ATF IONPs nanoparticle dye developed by an established protocol138. We tentatively evaluate the affinity with FMT, o ur study suggested that the affinity to tumor cell increased over 10 fold compared to NIR830IONP without ATF. Trace cancer cells in recurrent tumor and metastasis can be detected and FMT results are consistent with planar fluorescence imaging. Our results show that NIR830AT F IONP is very suitable for preclinical cance r research. Its high specificity to tumor cell and stability after administration could help follow up pathology study, mark surgical margins more accurately and detect possible circulating cancer cells in blood; DOT guided quantitative FMT can be a promising tool in preclinical study such as fluorescent dye quality assess, tumor progression monitor early stage cancer detection, chemotherapy evaluation and drug delivery

PAGE 88

88 B B Fig ure 51. Mo use in Experiment. A ) Photography of the experimental system, where the insert is the 3D finite element mesh of the mouse for the region of interest B ) The detector node distribution of a typical sagittal projection. Blue stars are the voxel of 3D mouse m odel and red circles are the projected detector nodes on the finite element mesh used. Figure 52. Reconstructed 3D a and s images of a typical mouse (mm1): A ) a image B ) s image A A

PAGE 89

89 Figure 53 Recovered 3D FMT images from a mouse in sagittal A ) and coronal projection B ). The exact tumor location (red mesh), and the isosurface plot of the fluorescent target recovered with ( in gold en) and w ithout ( in green ) DOT guidance are shown together for comparison. Figure 54 Comparison of signal intensity of the mammary tumor in the mice received uPAR targeted NIR 830 ATF IONPs and non targeted NIR 830 MSA IONP (with ATF vs wi thout ATF). (A ) Tail vein injection of 40 pmol of NIR 830ATF IONPs for 6 days (B ) Tail vein injection of 40 pmol of NIR 830 MSA IONPs for 6 days A B B A 100000 200000 500000

PAGE 90

90 Figure 55 Detection of local recurrent mammary tumor and lung metast asis using reconstructed 3D images of FMT method and NIR 830ATF peptide optical imaging probes The primary tumor of the mouse was removed by surgery. Local recurrent tumor in the mammary fat pad and lung metastasis developed two weeks following surgery. The mouse received the tail vein delivery of 40 ug of NIR 830 ATF peptides for 72 hrs. Bioluminescence imaging showing the locations of the tumor lesions. (a) Planar fluorescent image by IVIS system. (b) Reconstructed 3D images by FMT method. A B

PAGE 91

91 CHAPTER 6 APPLICATION IN A DROSOPHILA PUPA MODEL Basic biologic research (e.g., genetics epigen e tics, and stem cell s) often use animals such as D rosophila, zebrafish, and xenopus laevis as the study models of mesoscopic scale biological events. Characterization of molecular events in these models currently relies on timeconsuming in vitro tissue sectioning based microscopic techniques including fluorescence microscopy ( ~ 40 m depth), confocal microscropy ( ~ 200um depth) and twoor multi photon fluorescence microscopy ( ~ 500um depth). In particular, the in vitro microscopic imaging can be performed only at certain fixed time points, making a dynamic characterization of cell events of the same animal impossible. In contrast, the emerging f luorescen ce molecular tomography (FMT) is capable of providing information about specific biological events of intact live animal in its natural state, and has already shown promises in imaging small animals such as mice with applications including cancer detection, drug discovery and basic mechanism studies 12, 13, 57, 59, 139, 140 Recently, Vinegoni et al showed the possibility of in vivo tomographically imaging millimeter or mesoscopic scale animals where green fluorescent protein (GFP) ex pressing cells inside D rosophila pupae were imaged using a FMT approach based on F ermi simplification to the Fokker Planck solution of photon transport theory This simplified model is effective only in strongly forwardscattering regimes. In their study, this approximation was satisfied through the use of a polarizer to reject highly diffusive photons due to the relatively strong polarization property GFP contains, which allows to collect only 10% of the boundary photons .141

PAGE 92

92 Compared to GFP, DsRed shows almost no polarization property since biexponential anisotropy decay reveals a fast ( 2116ps) depolarization in DsRed .142, 143 T hus polarization based method used for GFP is not applicable to the DsRedbearing pupae we study here. T he radiative transfer equation (RTE) is perfect to deal with such cases for mesoscopic fluorescence tomography and the RTE based FMT can be applied to imag ing fluorescent dyes without polarization. In the RTE framework, s ince all boundary photons are collected, the laser power can be reduced to minimize possible damage the live millimeter scale animals Previously Klose et al 144 implemented the RTE based FMT reconstruction algorithm using finite difference method and demonstrated successful image reconstruction in a slab like geometry phantom 19and in vivo mouse imaging in a match f luid filled slab like container with finite difference method .145 Amit Joshi et al .146 presented the RTE based fluorescence tomography on a com putational mouse model to localize fluorophore co ncentrati on distribution. Our RTE based fluorescence reconstruction algorithm is implemented with finite element method for arbitrarily shaped object s. This algorithm is validated using a Cy5. 5 dye containing microtube embedded in a pupa The RTE based algorithm is also tested using DsRed bearing live pupae, and the reconstructed images are compared with images obtained from confocal microscop y. In our platform, the FMT microscope was horizontally mounted, and the sample laid vertically on a rotation stage. A laser beam was focused to the horizontal scan positions on the pupas surface with a planoconvex lens The sample was rotated to collect signals from different angle and the fluorescent signals were captured with a

PAGE 93

93 microscope lens mounted on a CCD camera The v ertical scanning of the laser beam was realized via a linear stage. The tomographically collected fluorescence measurements were then used to localize and quantify the signal from IRER{ubiDsRed} through a radiative transfer equation based tomographic inverse algorithm. Static Fluorophore Concentration Imaging Cy5.5 Microtube Imaging The FMT experimental system and d ata acquisition procedure were described in detail in our previous studies 28, 96 Briefly, the exper iments were conducted using a continuous wave (CW) 660nm diode laser as excitation and 710nm band pass filter for emission light acquisition. The 3D shape of the pupa was obtained with volume carving method based on silhouettes of 72 projections .117 Signals from a total of 484 sources at 4 CCD positions and 484 detector positions at the opposite side of the source location were collected.T he 3D experimental data was mapped to seven transverse slices and the 2D RTE based FMT algorithm was utilized for reconstruction. A 3 mm long silica capillary microtube ( outer diameter = 150 m and inner diameter = 100 m ) containing Cy5.5 dye served as a target The tub e was inserted through the pupa as shown in Fig 6 2 (a). The D rosophila pupa wa s positioned vertically on a rotation stage. Laser beam was focused to source point at the pupas shell with a planoconvex lens. The sample was rotated to 4 projection positions and transmi tted light were captured with a 1024x 1024 pixels CCD camera. Since an early stage pupa is composed of mostly fat tissue like larva, the optical propert ies of the p upae can then be assumed homogenous. The reconstructed FMT images (transverse slice s) using both the diffusion and RTE based methods are given in Fig 6 2b~6 2 h We see that s trong boundary artifact s exist in all the images using the

PAGE 94

94 diffusion based method, while the RTE based method shows overall much improved images. We also note that t he images with the RTE based method (bottom row, Fig. 6 2b~62h) give a target localization error of up to 100 m. Fig. 6 3 shows the 3D rendering of the reconstructed images obtained from these 7 transverse slices where the golden rod indicates the exact position and size of the Cy5.5containing microtube. DsRed Whole Body Imaging Drosophila is a widely used model for genetic and molecular biology research The genetic prowes s of this organism allows florescent markers such as the DsRed gene to be readily inserted into interested loci as a report er for various applications .147, 148 The pupa stage of D rosophila undergoes extensive tissue remodeling controlled by genetic casade before it develops into an adult fruit fly. In our experiments, DsRed fluorescent reporter was inserted into the middle of IRER (Irradi ation responsive enhancer region) and can only be expressed in IRERopen cells (Zhang et al 2008) IRER is open in undifferentiated proliferating embryonic stem cells and is responsible for stress induced cell death of these cells. Therefore, by monitoring the DsRed signal from different organs during development, we will be able to monitor the epigenetic status of IRER as well as follow some stem cell activities and events .149 The in vivo e xperiments were conducted following exactly the same procedure as the Cy5.5 microtube experiments except that a 535nm CW diode laser was used as exci tation and that a 585nm band pass filter was applied for emission light acquisition. Low laser power (6 W) was used during the 10 minute scanning for a full set of data collection (1 second CCD exposure time for each source position ). We noticed that such a low laser power did not interrupt the undergoing biology development of the sample,

PAGE 95

95 and that all the pupae were developed in to normal fruit flies after the optical imaging Confocal microscropy of certain slices from the pupae of same development stage w ere used to validate our reconstructed FMT results. Fig. 6 4 presents the FMT images for a typical day 2 pupa, in comparison with the confocal microscopy. In this figure, the first schematic shows the positions/sections corresponding to the FMT/confocal sli ces; the top row gives the FMT images, while the bottom row displays confocal microscope (bottom row ad) and epifluorescence microscope (bottom row e & f) images As can be seen, larger congregated DsRedcontaining organs are clearly visualized ( slice s b and d). Major features in more complicated structure are also identifiable in the FMT images ( slice s a and c ), although the FMT images are relatively blurred compared to the confocal image s. Sagittal slices given in slices e and f were obtained through the interpolation using 9 transverse slices. We immediately note that t he sagittal FMT images agree well with the corresponding images obtained from cryostat sections of fixed DsRedexpressing pupae. Although it still needs to be verified by a tissue specific marker, it seems that at about 1 day post pupation (PP), the strong signal at the anterior corresponds to the degenerating salivary gland while the one at the posterior corresponds to the midgut. Signals from both regions were captured by the FMT images. Dynamic DsRed Concentration Imaging Epigenetic regulation, by limiting the accessibility of DNA and the expressivity of genes, plays a fundamental role in determining the differentiation potential of stems cells as well as the properties of differentiated cells. Deregulation of epigenetic status has been implicated in many diseases such as cancer, cardiovascular disorders, and mental diseases. In contrast to static genetic changes, epigenetic regulations are

PAGE 96

96 dynamic, responsive to environmental and dietary factors, and under m any circumstances, reversible. The dynamic nature of epigenetic regulation demands innovative techniques that allow continuous monitoring of epigenetic status in live animals. However, most biochemical methodologies for measuring epigenetic modification and DNA accessibility rely on homogenizing large amount of cells, which is inapplicable for monitoring dynamic epigenetic changes in live animals. In this study, we explored the applicability of using a fluorescent reporter, inserted into an epigenetically regulated region, in monitoring and semi quantitative assessment of epigenetic status and DNA accessibility in vivo Our analysis indicated that the expression of this ubiquitinDsRed reporter accurately reflects the epigenetic status, i. e. the accessibility of DNA, in the tested locus. This reporter allowed us to monitor epigenetic changes of this locus during development as well as in response to histone modification compounds administrated with food. In addition, we showed that it is possible to semi quantitatively measure DNA accessibility in live animals by measuring the fluorescence recovery after photobleaching. Dynamic measurement of DNA accessibility in live animals via FMT. Much cell death and proliferation occur during metamorphosis. However, the pupal shell is opaque and deflects the fluorescent signal, making it impossible for fluorescent microscopy to accurately localize and quantify the signal. To solve this problem, we have developed an initial platform for monitoring the dyn amic signal from IRER {ubi DsRed} using fluorescence molecular tomography (FMT) 96, 150 The main advantage of monitoring epigenetic change in live animals is that we could follow the dynamic epigenetic status dur ing organism development, cellular differentiation and migration, and continuous response to environmental factors. To (g ) (i )

PAGE 97

97 interrogate if longitudinal observations would be possible, we followed the expression of IRER{ubiDsRed} over the whole pupation period (~115 hours in our setting as in Fig. 6 5 ) with recordings done at 215 hour interval. The relatively low energy required for FMT monitoring appeared to be well tolerated by the animal and did not disrupt the development process. Adult flies emerged from t he pupal case at the end of recording without any noticeable defect. Repeated recordings of three pupae indicated that the dynamic pattern of IRER{Ubi DsRed} during pupation is highly repeatable (Fig. 67 ). The results are further confirmed by using micros cope to image pupa whose shell is removed as in Fig. 6 6 F luorescence R ecovery after P hototbleaching (FRAP) and FMT M uch of the DsRed signal could be due to expression of the protein prior to the testing time. So the level of fluorescent signal monitored during development actually reflects the epigenetic status of the cell prior to the time of observation. To determine whether we could directly assay the status of epigenetic modification in live animals, we devised a scheme to monitor FRAP. Our initial da ta indicated that it is fully feasible to monitor the recovering of fluorescent signal, i.e. de novo synthesis of the reporter protein, in live animals using FRAP (bottom inlets in Fig. 6 7 ). The FRAP analysis was focused on the developing midgut in the posterior of the pupa, which has a very strong signal during early pupation. The overall signal in this area starts to increase at early pupation, reach its peak at about 30hr P.P. and starts to decline at about 80 hr P.P.. The FRAP analysis indicated that c orresponding with the overall increase of the signal, the rate of FRAP is relatively faster at 14 hr P.P. (photobleaching done at 14 hr P.P ., fluorescent signal is monitored up to 5 hour after bleaching ). In contrast, FRAP was almost absent at 75 hr P.P., which is still 5 hrs prior to the precipitous decline of

PAGE 98

98 fluorescent signal in this region. This result strongly suggests that FRAP measurements, in combination of monitoring overall fluorophore DsRed concentration by FMT provide better measurement of DNA accessibility for region of interest at a given time point. The relationship of FMT and FRAP can be further revealed with derivative FMT together with normalized FRAP in Fig. 6 8 Derivative FME indicates overall DsRed concentration change rate and normal ized FRAP indicates regenerate speed proportionally to gene openings. The tendencies of two methods for salivary gland and the midgut almost overlap each other after subtracting 1.2 at 14 hour stage FRAB The 1.2 difference is the sudden decrease of DsRed concentration which we think is due to a sudden increase in dying metabolism attributed to cell dying process naturally programmed to get rid of useless larva organ tissues and provide nutrition and space for new adult fruit fly organs at about 10% stage o f pupation. In summary to realize in vivo monitoring of DSRedexpressing cell distribution In Drosophila pupae using fluorescence molecular tomography (FMT ), t he radiative transfer equation (RTE) based FMT reconstruction algorithm is implemented using finite element method for mesoscopic millimeter scale imaging. The RTE algorithm is validated using both simulated and phantom experimental data. For the in vivo experiments, DsRed fluorescent reporter was inserted into the middle of IRER (Irradiation Respons ive Enhancer Region) of Drosophila pupae and expressed only in IRERopen cells. Quantitatively accurate fluorescence images can be reconstructed from both simulated and phantom data. The in vivo images obtained agree well with the

PAGE 99

99 in vitro images obtained from confocal microscopy both qualitatively and semi quantitatively. DsRed bearing stem cells in Drosophila pupae can be markedly imaged using our FMT approach. D ynamic in vivo monitoring of biological events in mesoscopic scale animals can greatly facilitate basic biologic research such as genetics, epigenetic and stem cells. I nitial dynamic monitoring and FRAB are very promising in reveal ing epigenetic regulation activity during pupation, although much work lies ahead for us in order to further validate and develop the methodology. I mprovements are certainly needed especially with respect to the spatial resolution data acquisition speed and experimental protocol To enable cell level event monitoring, the resolution needs to be improved from the curren t 100 m to 10 m. Since the low resolution is most caused by scattering photo n s, techniques to capture early or less scattering photon s 40, 43 will provide much better boundary signal and higher resolution. H igh speed dat a acquisition will allow realtime dynamic monitoring of biologic events and can be implemented with fast Galvo scanner for light delivery and EMCCD for data collection, making a frame rate of 100 possible Experimental protocol need to be improved for bet ter statistic validity: the design of the reporter needs to be modified to minimize the impact of mRNA stability on FRAP analysis. The novel FMT&FRAP analysis needs to be carefully validated and improved with better resolution. Nonetheless, the in vivo mon itoring of stem cells demonstrated in this thesis has paved the way for us to continually optimize our FMT method for improved performance.

PAGE 100

100 Fig ure 6 1. A p upa in experiment

PAGE 101

101 Figure 62 Compara tive Cy5.5 tube experiment of the diffusion and RTE based FMT reconstruction for a microtube e mbedded pupa. a) Photograph of the pupa under test (top) and the source point/ node distribution (green cross es ) in coronal projection (bottom). (b)( h ) : reconstructed FMT images for seven consecutive transverse slice s w ith the diffusion based method ( top row) and the RTE based method ( bottom row) The green circle indicates the exact position of the dyecontaining microtube. Figure 63 3D view of reconstructed Cy 5.5 microtube in the pupa. A ) reconstructed Cy5. 5 dyecontaining micro tube in blue isosurface plot. B ) reconstructed Cy5.5 dyecontaining micro tube (blue)and the exact position of the tube (golden) C ) another view of the reconstructed Cy5.5 dyecontaining micro tube (blue)and the exact position of the t ube (golden) A B C

PAGE 102

102 Figure 64 Reconstructed in vivo FMT (top row a f) in vitro confocal microscope (bottom row ad) and epifluorescence microscope (bottom row e & f) images: Column a, b, c and d: transverse slices ; Column e and f : sagittal slices. Fi gure 65 3D FMT tomography of 15 stages during the pupation development.

PAGE 103

103 Figure 66 Microscope images of pupa (shell removed) in early stage Sudden decrease in DsRed expressing at 10% stage is detected.

PAGE 104

104 Figure 67 Quantifying the change of DNA accessibility in live animals with FMT and FRAP The top inlets are representative whole pupa constructions. The three lines represent the developmental change of signal level at the midgut region measured on three pupae. The inlets below the lines are FRAP measurements performed at 14hr, 4 5 hr, 7 5 hr, and 94 hr post pupation (P.P.). For FRAP, each line represents an independent measurement on a different pupa at the same developmental stage (hours P.P.). FRAP measurements indicated that the region is already closed at about 75 hr P.P., even though the level of fluorescent signal did not start to decrease for another few hours.

PAGE 105

105 Figure 68 Correlation between derivative FMT together with normalized FRAP. For better comparison, the hour (x axis) are aligned and the value (y axis) are normalized. Actual sampling hour discrepancy between FMT and FRAB for each stage of five stages in plot is within 5 hours.

PAGE 106

106 CHAPTER 7 CONCLUSIONS In this thesis, DOT guided quantitative FMT are proposed: simulation test exper imental implementation and preclinical applications in both macroscale and meso scale samples are investigated. A quantitative 3D non contact FMT systems was built and continuously innovated during the investigation and it is capable of revealing dye dist ribution for macroscale and meso scale animals in their natural states. The workflow mainframe is described in Fig 7 1. To obtain full quantitative FMT, a DOT guided FMT approach is proposed. In this approach, reconstruction is conducted in both excitati on and emission wavelength. Reconstruction in excitation wavelength can provide full optical heterogeneity information for quantitatively accurate FMT reconstruction. The simulation results in Chapter 2 and phantom experimental results in Chapter 4 show the improvement clearly. In C hapter 5, we also use our DOT guided FMT to evaluate the newly developed fluorescent dyes ( NIR830MSA IONP and Cy 5.5 ATF) which show great potential in accurately localizing tumor boundary to enable well defined excision region for tumor surgery. The reconstructed images in Chapter 5 also show good improvement in FMT with DOT guidance. Our study suggested that the affinity to tumor cell increased over 10 fold compared to NIR830MSA IONP without ATF. Trace cancer cells in recurrent tumor and metastasis can be detected and FMT results are consistent with planar fluorescence imaging. Our results show that NIR830MSA IONP nanoparticle is very suitable for preclinical cancer research. Its high specificity to tumor cell and stability after administration could help follow up pathology study, mark surgical margins more accurately and detect possible circulating cancer cells in blood; DOT guided quantitative

PAGE 107

107 FMT proved be a promising tool in preclinical study such as tumor progression m onitor, early stage cancer detection, chemotherapy evaluation and drug delivery. This thesis proposed RTE based FMT reconstruction for meso scale sample like Drosophila pupa. Light transport ation within the object is described through two approaches (RTE and DA) according to the size of objects In our Drosophila experiments, the distance between detector and source is less than 1mm so DA is no longer a good model We implemented RTE based reconstruction and validate the method with simulation, Cy5.5 microt ube experiments and DsRed dynamical monitoring experiments. The results we got in Chapter 6 shows that RTE based FMT is a good model for mesoscale animal study to provide dynamical semi quanti tative monitoring within 100 m localization error. The FMT imag e is validated by confocal microscope semi quantitatively and qualitatively. To help study the stem cell activities, we also developed an in vivo Fluorescence Recovery After P hotobleaching ( FRAP) procedure and coordinate it with our dynamic FMT monitoring Based on our monitoring of full developmental course of Drosophila pupa, we got consistent results from FRAP and RTE. There is a dramatic drop in in vivo FMT at 10% stage, and overall dynamic plot of FMT and FRAP are very consistent with each other. We also further confirm this phenomen on with upright fluorescent microscope. The sudden increase dying metabolism factor might be attributed to cell dying process naturally programmed to get rid of useless larva organ tissues and provide nutrition and space for developing new organs of adult fruit fly

PAGE 108

108 Figure 71. Workflow f ramework of the d issertation

PAGE 109

109 LIST OF REFERENCES 1. J. Reynolds, T. Troy, R. Mayer, A. Thompson, D. Waters, K. Cornell, P. Snyder, and E. Sevick Muraca, "Imaging of sp ontaneous canine mammary tumors using fluorescent contrast agents," Photochem. Photobiol. 70, 87 94 (1999). 2. J. Grimm, D. G. Kirsch, S. D. Windsor, C. F. B. Kim, P. M. Santiago, V. Ntziachristos, T. Jacks, and R. Weissleder, "Use of gene expression profi ling to direct in vivo molecular imaging of lung cancer," P. Natl. Acad. Sci. USA 102(40), 14404 14409 (2005). 3. A. Becker, C. Hessenius, K. Licha, B. Ebert, U. Sukowski, W. Semmler, B. Wiedenmann, and C. Grtzinger, "Receptor targeted optical imaging of tumors with near infrared fluorescent ligands," Nat. Biotechnol. 19 (4), 327 331 (2001). 4. X. Montet, J. L. Figueiredo, H. Alencar, V. Ntziachristos, U. Mahmood, and R. Weissleder, "Tomographic fluorescence imaging of tumor vascular volume in mice," Radio logy 242(3), 751 758 (2007). 5. H. Jiang, "Frequency domain fluorescent diffusion tomography: a finiteelement based algorithm and simulations," Appl. Optics. 37, 53375343 (1998). 6. J. Steinbrink, A. Liebert, H. Wabnitz, R. Macdonald, H. Obrig, A. Wunder, R. Bourayou, T. Betz, J. Klohs, and U. Lindauer, "Towards Noninvasive Molecular Fluorescence Imaging of the Human Brain," Neurodegenerative Dis 5 (5), 296 303 (2008). 7. A. Corlu, R. Choe, T. Durduran, M. Rosen, M. Schweiger, S. Arridge, M. Schnall, and A Yodh, "Threedimensional in vivo fluorescence diffuse optical tomography of breast cancer in humans," Opt. Express. 15(11), 66966716 (2007). 8. S. Roddaro, A. Fuhrer, C. Fasth, L. Samuelson, J. Xiang, and C. Lieber, "Strong g Factor Anisotropy in Hole Q uantum Dots Defined in Ge/Si Nanowires," Arxiv. Pre. 0706.2883 (2007). 9. P. Theer, M. Hasan, and W. Denk, "Twophoton imaging to a depth of 1000 m in living brains by use of a Ti: Al_2O_3 regenerative amplifier," Opt. Lett. 28(12), 10221024 (2003). 10. L. Gao, R. Kester, and T. Tkaczyk, "Compact Image Slicing Spectrometer (ISS) for hyperspectral fluorescence microscopy," Opt. Express 17 1229312308 (2009). 11. V. Ntziachristos, J. Ripoll, L. H. V. Wang, and R. Weissleder, "Looking and listening to light: the evolution of wholebody photonic imaging," Nat. Biotechnol. 23(3), 313 320 (2005).

PAGE 110

110 12. J. Rao, A. Dragulescu Andrasi, and H. Yao, "Fluorescence imaging in vivo : recent advances," Curr. Opin. Biotech. 18(1), 17 25 (2007). 13. F. Leblond, S. Davis, P. Valds, and B. Pogue, "Preclinical whole body fluorescence imaging: Review of instruments, methods and applications," J Photoch. Photobio. B. 98(1), 77 94 (2010). 14. E. E. Graves, R. Weissleder, and V. Ntziachristos, "Fluorescence molecular imaging of smal l animal tumor models," Curr. Mol. Med. 4 (4), 419 430 (2004). 15. S. Arridge and J. Schotland, "Optical tomography: forward and inverse problems," Arxiv. Pre. 0907.2586 (2009). 16. A. S R, "The theoretical basis for the determination of optical pathlengths in tissue: temporal and frequency analysis," Phys. Med. Biol 37 (7), 15311560 (1992). 17. S. Arridge, M. Schweiger, M. Hiraoka, and D. Delpy, "A finite element approach for modeling photon transport in tissue," Med. Phys. 20 299 30 9 (1993). 18. S. Arridg e and M. Schweiger, "Photonmeasurement density functions. Part 2: Finite element method calculations," Appl. Optics. 34 (34), 80268037 (1995). 19. A. Klose, V. Ntziachristos, and A. Hielscher, "The inverse source problem based on the radiative transfer eq uation in optical molecular imaging," J Comput Phys. 202(1), 323345 (2005). 20. J. Culver, R. Choe, M. Holboke, L. Zubkov, T. Durduran, A. Slemp, V. Ntziachristos, B. Chance, and A. Yodh, "Threedimensional diffuse optical tomography in the parallel plan e transmission geometry: Evaluation of a hybrid frequency domain/continuous wave clinical system for breast imaging," Med. Phys. 30, 235247 (2003). 21. K. Ren, G. Abdoulaev, G. Bal, and A. Hielscher, "Algorithm for solving the equation of radiative transf er in the frequency domain," Opt Lett 29(6), 578580 (2004). 22. J. Ripoll and V. Ntziachristos, "Iterative boundary method for diffuse optical tomography," J Opt. Soc. AM A. 20(6), 11031110 (2003). 23. E. Aydin, C. De Oliveira, and A. Goddard, "A comparison between transport and diffusion calculations using a finite element spherical harmonics radiation transport method," Med. Phys. 29 2013 2023 (2002). 24. V. Markel, "Modified spherical harmonics method for solving the radiative transport equation," W aves in Random and Complex Media 14(1), 13 19 (2004).

PAGE 111

111 25. A. Klose, "The forward and inverse problem in tissue optics based on the radiative transfer equation: A brief review," J Quant. Spect. Rad. Trans. 111 (11) 18521853 (2010) 26. V. Ntziachristos and R. Weissleder, "Experimental threedimensional fluorescence reconstruction of diffuse media by use of a normalized Born approximation," Opt. Lett. 26(12), 893895 (2001). 27. C. Vinegoni, D. Razansky, J. L. Figueiredo, M. Nahrendorf, V. Ntziachristos, and R. Weissleder, "Normalized Born ratio for fluorescence optical projection tomography," Opt. Lett. 34(3), 319 321 (2009). 28. Y. Tan and H. Jiang, "Diffuse optical tomography guided quantitative fluorescence molecular tomography," Appl Opt 47 (12), 201120 16 (2008). 29. R. Sevick Muraca, "Truncated Newton's optimization scheme for absorption and fluorescence optical tomography: Part I theory and formulation," Opt. Express 4 353371 (1999). 30. R. Sevick Muraca, "Truncated Newton's optimization scheme for a bsorption and fluorescence optical tomography: Part II Reconstruction from synthetic measurements," Opt. Express 4 372 382 (1999). 31. M. Eppstein, D. Dougherty, D. Hawrysz, and E. Sevick Muraca, "Threedimensional Bayesian optical image reconstruction wi th domain decomposition," IEEE. T. Med. Imaging. 20(3), 147 163 (2001). 32. M. Eppstein, D. Hawrysz, A. Godavarty, and E. Sevick Muraca, "Threedimensional, Bayesian image reconstruction from sparse and noisy data sets: near infrared fluorescence tomography," P. Natl. Acad. Sci. USA. 99 (15), 9619 (2002). 33. A. Godavarty, E. Sevick Muraca, and M. Eppstein, "Threedimensional fluorescence lifetime tomography," Med. Phys. 32, 992 1000 (2005). 34. V. Ntziachristos, J. P. Culver, and B. W. Rice, "Small animal optical imaging," J Biomed Opt. 13(1), 011001 (2008) 35. V. Ntziachristos, E. Graves, R. Schultz, and J. Ripoll, "Fluorescence molecular tomography: New detection schemes for acquiring high information content measurements," I. S. Biomed. Imaging. 2 14751 478 ( 2004) 36. V. Ntziachristos, E. A. Schellenberger, J. Ripoll, D. Yessayan, E. Graves, A. Bogdanov, L. Josephson, and R. Weissleder, "Visualization of antitumor treatment by means of fluorescence molecular tomography with an annexin V Cy5.5 conjugate," P. Natl. Acad. Sci. USA. 101 (33), 1229412299 (2004).

PAGE 112

112 37. R. Roy, A. Thompson, A. Godavarty, and E. Sevick Muraca, "Tomographic fluorescence imaging in tissue phantoms: a novel reconstruction algorithm and imaging geometry," IEEE. T. Med. Imaging. 24(2), 137 154 (2005). 38. G. M. Turner, G. Zacharakis, A. Soubret, J. Ripoll, and V. Ntziachristos, "Completeangle projection diffuse optical tomography by use of early photons," Opt. Lett. 30 (4), 409 411 (2005). 39. G. M. Turner, A. Soubret, and V. Ntziachrist os, "Inversion with early photons," Med. Phys. 34(4), 1405 1411 (2007). 40. M. J. Niedre, R. H. de Kleine, E. Aikawa, D. G. Kirsch, R. Weissleder, and V. Ntziachristos, "Early photon tomography allows fluorescence detection of lung carcinomas and disease progression in mice in vivo ," P. Natl. Acad. Sci. USA. 105(49), 19126 19131 (2008). 41. E. Hillman, O. Bernus, E. Pease, M. Bouchard, and A. Pertsov, "Depthresolved optical imaging of transmural electrical propagation in perfused heart," Opt E xpress 15(26 ), 17827 17841 (2007). 42. E. Hillman, D. Boas, A. Dale, and A. Dunn, "Laminar optical tomography: demonstration of millimeter scale depthresolved imaging in turbid media," Opt. Lett. 29(14), 16501652 (2004). 43. L. Florescu, J. Schotland, and V. Markel, "Single scattering optical tomography," Phys. Rev. E. 79(3), 36607 (2009). 44. D. Hogemann, V. Ntziachristos, L. Josephson, and R. Weissleder, "High throughput magnetic resonance imaging for evaluating targeted nanoparticle probes," Bioconjugate. Chem. 13 (1), 116 121 (2002). 45. M. Niedre and V. Ntziachristos, "Elucidating structure and function in vivo with hybrid fluorescence and magnetic resonance imaging," P. IEEE. 96 (3), 382 396 (2008). 46. D. E. Sosnovik, M. Nahrendorf, N. Deliolanis, M. Novikov, E. Aikawa, L. Josephson, A. Rosenzweig, R. Weissleder, and V. Ntziachristos, "Fluorescence tomography and magnetic resonance imaging of myocardial macrophage infiltration in infarcted myocardium in vivo ," Circulation 115 (11), 1384 1391 (2007). 47. D. Hyde, R. de Kleine, S. A. MacLaurin, E. Miller, D. H. Brooks, T. Krucker, and V. Ntziachristos, "AOI987:Hybrid FMT CT imaging of amyloid beta plaques in a murine Alzheimer's disease model," Neuroimage 44(4), 13041311 (2009). 48. D. Hyde, R. Schulz, D. Brooks, E. Miller, and V. Ntziachristos, "Performance dependence of hybrid x ray computed tomography/fluorescence molecular

PAGE 113

113 tomography on the optical forward problem," J Opt. Soc. AM A. 26(4), 919923 (2009). 49. M. Nahrendorf, E. Keliher, B. Marinelli, P. Waterman, P. Feruglio, L. Fexon, M. Pivovarov, F. Swirski, M. Pittet, and C. Vinegoni, "Hybrid PET optical imaging using targeted probes," P. Natl. Acad. Sci. USA. 107 (17), 79107915 (2010) 50. D. Razansky and V. Ntziachristos, "Hybrid photoacoustic fluorescence mol ecular tomography using finiteelement based inversion," Med. Phys. 34(11), 42934301 (2007). 51. D. Razansky, C. Vinegoni, and V. Ntziachristos, "Multispectral photoacoustic imaging of fluorochromes in small animals," Opt. Lett. 32(19), 2891 2893 (2007). 52. D. Razansky, C. Vinegoni, and V. Ntziachristos, "Imaging of mesoscopic scale organisms using selectiveplane optoacoustic tomography," Phys. Med. Biol. 54(9), 27692777 (2009). 53. P. Bandettini, "What's new in neuroimaging methods?," Ann. NY. Acad. Sc i. 1156, 260 293 (2009). 54. N. Blow, In vivo molecular imaging: the inside job," Nat. Methods. 6 (6), 4 65469 (2009). 55. J. Frangioni, In vivo near infrared fluorescence imaging," Curr. Opin. Chem. Biol. 7 (5), 626634 (2003). 56. B. Giepmans, S. Adams, M. Ellisman, and R. Tsien, "The fluorescent toolbox for assessing protein location and function," Sci Signaling 312(5771), 217224 (2006). 57. K. Licha and C. Olbrich, "Optical imaging in drug discovery and diagnostic applications," Advanced drug delivery reviews 57(8), 10871108 (2005). 58. T. Misgeld and M. Kerschensteiner, In vivo imaging of the diseased nervous system," Nat. Rev. Neurosci. 7 (6), 449463 (2006). 59. N. C. Deliolanis, J. Dunham, T. Wurdinger, J. L. Figueiredo, B. A. Tannous, and V. Ntzi achristos, In vivo imaging of murine tumors using completeangle projection fluorescence molecular tomography," J Biomed Opt. 14(3), 030509 (2009). 60. J. Baeten, J. Haller, H. Shih, and V. Ntziachristos, In Vivo Investigation of Breast Cancer Progression by Use of an Internal Control," Neoplasia 11 (3), 220227 (2009). 61. X. Montet, V. Ntziachristos, J. Grimm, and R. Weissleder, "Tomographic fluorescence mapping of tumor targets," Cancer. Res. 65 (14), 6330 6336 (2005).

PAGE 114

114 62. H. K. Choi, D. Yessayan, H. J. Choi, E. Schellenberger, A. Bogdanov, L. Josephson, R. Weissleder, and V. Ntziachristos, "Quantitative analysis of chemotherapeutic effects in tumors using in vivo staining and correlative histology," Cell. Oncol. 27(3), 183190 (2005). 63. C. Bremer, V. N tziachristos, B. Weitkamp, G. Theilmeier, W. Heindel, and R. Weissleder, "Optical imaging of spontaneous breast tumors using protease sensing 'Smart' optical probes," Invest Radiol. 40(6), 321327 (2005). 64. E. A. Schellenberger, A. Bogdanov, A. Petrovsky V. Ntziachristos, R. Weissleder, and L. Josephson, "Optical imaging of apoptosis as a biomarker of tumor response to chemotherapy," Neoplasia 5 (3), 187 192 (2003). 65. C. Bremer, M. Mustafa, A. Bogdanov, V. Ntziachristos, A. Petrovsky, and R. Weissleder, "Steady state blood volume measurements in experimental tumors with different angiogenic burdens A study in mice," Radiology 226(1), 214 220 (2003). 66. C. B. Bremer, M. Mustafa, A. A. Bogdanov, V. Ntziachristos, A. Petrovsky, and R. Weissleder, "Steady state blood volume measurements in experimental tumors with differing angiogenic burden," Radiology 225, 632633 (2002). 67. K. Licha, B. Riefke, V. Ntziachristos, A. Becker, B. Chance, and W. Semmler, "Hydrophilic cyanine dyes as contrast agents for near infrared tumor imaging: Synthesis, photophysical properties and spectroscopic in vivo characterization," Photochem. Photobiol. 72(3), 392398 (2000). 68. O. Veiseh, C. Sun, C. Fang, N. Bhattarai, J. Gunn, F. Kievit, K. Du, B. Pullar, D. Lee, and R. Ellenb ogen, "Specific targeting of brain tumors with an optical/magnetic resonance imaging nanoprobe across the bloodbrain barrier," Cancer. Res. 69(15), 62006207 (2009). 69. D. AbiHaidar and T. Olivier, "Confocal reflectance and twophoton microscopy studies of a songbird skull for preparation of transcranial imaging," J Biomed Opt. 14, 034038 (2009). 70. J. Klohs, M. Grafe, K. Graf, J. Steinbrink, T. Dietrich, D. Stibenz, P. Bahmani, G. Kronenberg, C. Harms, and M. Endres, In vivo imaging of the inflammator y receptor CD40 after cerebral ischemia using a fluorescent antibody," Stroke 39(10), 28452852 (2008). 71. A. Liebert, H. Wabnitz, H. Obrig, R. Erdmann, M. Moeller, R. Macdonald, H. Rinneberg, A. Villringer, and J. Steinbrink, "Noninvasive detection of f luorescence from exogenous chromophores in the adult human brain," Neuroimage 31(2), 600 608 (2006). 72. T. Austin, "Optical tomography of the neonatal brain," Eur. Radiol. 17, 29262933 (2007).

PAGE 115

115 73. D. Kepshire, S. Davis, H. Dehghani, K. Paulsen, and B. Pogue, "Reflect depth experiment 2D better than 3D Fluorescence tomography characterization for subsurface imaging with protoporphyrin IX," Opt E xpress 16(12), 85818593 (2008). 74. Y. Xu, H. Graber, and R. Barbour, "Image correction algorithm for function al threedimensional diffuse optical tomography brain imaging," Appl Opt. 46(10), 16931704 (2007). 75. U. Netz, A. Scheel, A. Klose, A. Hielscher, and J. Beuthan, "Optical tomography in the diagnosis of rheumatoid arthritis method and implementation," Med Laser Appl 22(1), 15 22 (2007). 76. Z. Yuan, Q. Zhang, E. Sobel, and H. Jiang, "Threedimensional diffuse optical tomography of osteoarthritis: initial results in the finger joints," J Biomed Opt. 12 034001 (2007). 77. U. Mahmood and R. Weissleder, "N ear infrared optical imaging of proteases in cancer," Mol. Cancer. Ther. 2 (5), 489 496 (2003). 78. S. Shen, Y. Chiou, J. Wang, M. Yen, R. Lee, C. Lai, and C. Chang, "Diffusionweighted singleshot echoplanar imaging with parallel technique in assessment o f endometrial cancer," AM. J. Roentgenol. 190 (2), 481 488 (2008). 79. A. Klose, U. Netz, J. Beuthan, and A. Hielscher, "Optical tomography using the time independent equation of radiative transfer Part 1: Forward model," J Quant. Spect. Rad. Trans. 72(5), 691713 (2002). 80. C. Wu, H. Barnhill, X. Liang, Q. Wang, and H. Jiang, "A new probe using hybrid virus dye nanoparticles for near infrared fluorescence tomography," Opt. Commun. 255 (4 6), 366374 (2005). 81. H. Jiang, K. Paulsen, U. Osterberg, B. Pogue, and M. Patterson, "Optical image reconstruction using frequency domain data: simulations and experiments," J Opt. Soc. AM A. 13(2), 253266 (1996). 82. J. Heino, S. Arridge, J. Sikora, and E. Somersalo, "Anisotropic effects in highly scattering media," Phy s. Rev. E. 68(3), 31908 (2003). 83. J. Tervo and P. Kolmonen, "Data fitting model for the kernel of integral operator from radiation therapy," Math. Comp. Model 28(3), 59 77 (1998). 84. N. Qatanani, A. Barham, and Q. Heeh, "Existence and uniqueness of the solution of the coupled conductionradiation energy transfer on diffusegray surfaces," Srvy. In Math. Appl. 2 43 58 (2007). 85. T. Tarvainen, M. Vauhkonen, V. Kolehmainen, and J. Kaipio, "Hybrid radiativetransfer diffusion model for optical tomography, Appl. Optics. 44(6), 876886 (2005).

PAGE 116

116 86. G. Kanschat, "A robust finite element discretization for radiative transfer problems with scattering," E. W. J. Numer.Math. 6 265 272 (1998). 87. N. Iftimia and H. Jiang, "Quantitative optical image reconstruction of turbid media by use of direct current measurements," Appl. Optics. 39(28), 52565261 (2000). 88. G. Sleijpen, H. Vorst, and D. Fokkema, "BiCGstab (l) and other hybrid Bi CG methods," Numer. Alogrithms. 7 (1), 75 109 (1994). 89. F. Fedele, J. Laible, an d M. Eppstein, "Coupled complex adjoint sensitivities for frequency domain fluorescence tomography: theory and vectorized implementation," J Comput. Phys. 187 (2), 597 619 (2003). 90. M. Patterson and B. Pogue, "Mathematical model for timeresolved and fre quency domain fluorescence spectroscopy in biological tissues," Appl. Optics. 33(10), 19631974 (1994). 91. E. Sevick Muraca, G. Lopez, J. Reynolds, T. Troy, and C. Hutchinson, "Fluorescence and absorption contrast mechanisms for biomedical optical imaging using frequency domain techniques," Photochem. Photobiol. 66(1), 55 64 (1997). 92. A. Scheel, M. Backhaus, A. Klose, B. Moa Anderson, U. Netz, K. Hermann, J. Beuthan, G. Mller, G. Burmester, and A. Hielscher, "First clinical evaluation of sagittal laser optical tomography for detection of synovitis in arthritic finger joints," Ann. Rheum. Dis. 64(2), 239 245 (2005). 93. A. Li, Q. Zhang, J. Culver, E. Miller, and D. Boas, "Reconstructing chromosphere concentration images directly by continuous wave diffuse optical tomography," Opt. Lett. 29 (3), 256 258 (2004). 94. Y. Xu, X. Gu, T. Khan, and H. Jiang, "Absorption and scattering images of heterogeneous scattering media can be simultaneously reconstructed by use of dc data," Appl. Optics. 41(25), 54275437 (20 02). 95. Y. Pei, H. Graber, and R. Barbour, "Normalized constraint algorithm for minimizing inter parameter crosstalk in DC optical tomography," Appl. Opt 35 39633978 (1996). 96. Y. Tan and H. Jiang, "DOT guided fluorescence molecular tomography of arbit rarily shaped objects," Med Phys 35(12), 57035707 (2008). 97. V. Ntziachristos, A. Hielscher, A. Yodh, and B. Chance, "Diffuse optical tomography of highly heterogeneous media," IEEE Trans. Med. Imaging 20(6), 470478 (2001).

PAGE 117

117 98. R. Roy, A. Godavarty, and E. Sevick Muraca, "Fluorescence enhanced optical tomography using referenced measurements of heterogeneous media," I EEE Trans. Med. Imaging 22(7), 824 836 (2003). 99. A. Sahu, R. Roy, A. Joshi, and E. Sevick Muraca, "Evaluation of anatomical structure and non uniform distribution of imaging agent in near infrared fluorescence enhanced optical tomography," Opt E xpress 13(25), 1018210199 (2005). 100. J. Rolland and H. Barrett, "Effect of random background inhomogeneity on observer detection performance," J Opt. Soc. AM A. 9 (5), 649 658 (1992). 101. A. Milstein, S. Oh, K. Webb, C. Bouman, Q. Zhang, D. Boas, and R. Millane, "Fluorescence optical diffusion tomography," Appl. Optics. 42(16), 30813094 (2003). 102. E. Shives, Y. Xu, and H. Jiang, "Fluorescence lifetime tomography of turbid media based on an oxygensensitive dye," Opt. Express. 39, 52565261 (2000). 103. A. Soubret, J. Ripoll, and V. Ntziachristos, "Accuracy of fluorescent tomography in the presence of heterogeneities: Study of the normalized bor n ratio," I EEE Trans. Med. Imaging 24(10), 13771386 (2005). 104. L. Herv, A. Koenig, A. Da Silva, M. Berger, J. Boutet, J. Dinten, P. Pelti, and P. Rizo, "Noncontact fluorescence diffuse optical tomography of heterogeneous media," Appl. Optics. 46 (22), 48964906 (2007). 105. C. Li and H. Jiang, "A calibration method in diffuse optical tomography," Journal of Optics A: Pure and Appl. Optics. 6 844 852 (2004). 106. R. Tsai, "A versatile camera calibration technique for highaccuracy 3D machine vision metr ology using off the shelf TV cameras and lenses," Radiometry, 221244 (1992). 107. C. Harris and M. Stephens, "A combined corner and edge detector," 1988. 108. N. Ahuja and J. Veenstra, "Generating octrees from object silhouettes in orthographic views," IE EE. T Pattern. Anal. 137 149 (1989). 109. A. Laurentini, "The visual hull concept for silhouettebased image understanding," IEEE. T Pattern. Anal. 150 162 (1994). 110. J. Hasenfratz, M. Lapierre, J. Gascuel, and E. Boyer, "Real time capture, reconstructio n and insertion into virtual world of human actors," Vis Video Graph. Conf 4956 ( 2003) 111. W. Matusik, C. Buehler, and L. McMillan, "Polyhedral visual hulls for real time rendering," Eur Workshop. Rend. 115 125 ( 2001)

PAGE 118

118 112. W. Matusik, C. Buehler R. Raskar, S. Gortler, and L. McMillan, "Imagebased visual hulls," Proc of the 27th A nn C onf on Comp. G raph I nter T ech 369374 ( 2000) 113. J. Carranza, C. Theobalt, M. Magnor, and H. Seidel, "Freeviewpoint video of human actors," ACM. T. Graphic 22 569 577 ( 2003) 114. R. B. Schulz, J. Ripoll, and V. Ntziachristos, "Noncontact optical tomography of turbid media," Opt. Lett. 28(18), 1701 1703 (2003). 115. R. B. Schulz, J. Ripoll, and V. Ntziachristos, "Experimental fluorescence tomography of tiss ues with noncontact measurements," IEEE. T. Med. Imaging. 23(4), 492 500 (2004). 116. J. Ripoll, R. B. Schulz, and V. Ntziachristos, "Freespace propagation of diffuse light: Theory and experiments," Physical Review Letters 91 (10) ,103901 (2003). 117. A. Mu layim, U. Yilmaz, and V. Atalay, "Silhouettebased 3 D model reconstruction from multiple images," IEEE. Trans. Sys. Man Cyber. 33(4), 582 591 (2003). 118. Z. Zhang, "A flexible new technique for camera calibration," IEEE. T Pattern. Anal. 22(11), 13301334 (2000). 119. H. Zhao, F. Gao, Y. Tanikawa, K. Homma, and Y. Yamada, "Timeresolved diffuse optical tomographic imaging for the provision of both anatomical and functional information about biological tissue," Appl. Optics. 44 (10), 19051916 (2005). 120. D. Razansky and V. Ntziachristos, "Hybrid photoacoustic fluorescence molecular tomography using finiteelement based inversion," Med. Phys. 34, 42934301 (2007). 121. A. Serdaroglu, B. Yazici, and V. Ntziachristos, "Fluorescence molecular tomography based on a priori information." Biomed Opt Conf SH46 (2006) 122. X. Song, J. Yi, and J. Bai, "A parallel reconstruction scheme in fluorescence tomography based on contrast of independent inversed absorption properties," I. J. Bio. Imaging. 2006 1 7 (2006). 123. H. Van Staveren, C. Moes, J. van Marie, S. Prahl, and M. van Gemert, "Light scattering in Intralipid10% in the wavelength range of 400 1100 nm," Appl. Optics. 30 (31), 45074514 (1991). 124. E. Hull, M. Nichols, and T. Foster, "Quantitative broadband near infrared spectroscopy of tissue simulating phantoms containing erythrocytes," Phys. Med. Biol. 43 33813404 (1998).

PAGE 119

119 125. S. Flock, S. Jacques, B. Wilson, W. Star, and M. van Gemert, "Optical properties of Intralipid: a phantom medium for light propag ation studies," Lasers in Surg Med 12 510 51 9 (1992). 126. M. Landsman, G. Kwant, G. Mook, and W. Zijlstra, "Light absorbing properties, stability, and spectral stabilization of indocyanine green," J Appl Physiology 40(4), 575 583 (1976). 127. R. Philip, A. Penzkofer, W. Bumler, R. Szeimies, and C. Abels, "Absorption and fluorescence spectroscopic investigation of indocyanine green," J Photoch. Photobio. B. 96(1 3), 137148 (1996). 128. B. Yuan, N. Chen, and Q. Zhu, "Emission and absorption properties of indocyanine green in Intralipid solution," J Biomed Opt. 9 497 503 (2004). 129. G. Wang, W. Cong, K. Durairaj, X. Qian, H. Shen, P. Sinn, E. Hoffman, G. McLennan, and M. Henry, In vivo mouse studies with bioluminescence tomography," Biomed. Eng 4 2 35260 (2002). 130. Y. Lin, W. Barber, J. Iwanczyk, W. Roeck, O. Nalcioglu, and G. Gulsen, "Quantitative fluorescence tomography using a combined tri modality FT/DOT/XCT system," Opt. E xpress 18(8), 78357850 (2010) 131. J. Mourant, A. Hielscher, A. Eick T. Johnson, and J. Freyer, "Evidence of intrinsic differences in the light scattering properties of tumorigenic and nontumorigenic cells," Cancer. Cytopathol. 84 (6), 366374 (1998). 132. A. Cerussi, D. Jakubowski, N. Shah, F. Bevilacqua, R. Lanning, A. B erger, D. Hsiang, J. Butler, R. Holcombe, and B. Tromberg, "Spectroscopy enhances the information content of optical mammography," J Biomed Opt. 7 60 71 (2002). 133. H. Van Genderen, H. Kenis, P. Lux, L. Ungeth, C. Maassen, N. Deckers, J. Narula, L. Hofstra, and C. Reutelingsperger, In vitro measurement of cell death with the annexin A5 affinity assay," Nat. Protoc. 1 (1), 363 367 (2006). 134. C. Tamerler, E. Oren, M. Duman, E. Venkatasubramanian, and M. Sarikaya, "Adsorption kinetics of an engineered gold binding peptide by surface plasmon resonance spectroscopy and a quartz crystal microbalance," Langmuir 22(18), 77127718 (2006). 135. M. C. Weiger, J. J. Park, M. D. Roy, C. M. Stafford, A. Karim, and M. L. Becker, "Quantification of the binding affinity of a specific hydroxyapatite binding peptide," Biomaterials 31 (11), 29552963 (2010). 136. X. Yang, C. Shi, R. Tong, W. Qian, H. Zhau, R. Wang, G. Zhu, J. Cheng, V. Yang, and T. Cheng, "Near IR Heptamethi ne Cyanine Dye Mediated Cancer Imaging," Clin Cancer. Res. 16 (10), 28332844 (2010)

PAGE 120

120 137. J. Baeten, J. Haller, H. Shih, and V. Ntziachristos, In vivo investigation of breast cancer progression by use of an internal control," Neoplasia (New York, NY) 11(3 ), 220 227 (2009). 138. L. Yang X.Peng Y.Wang X.Wang Z. Cao C.Ni, P.Karna X.Zhang, W.Wood X.Gao S.Nie and H.Mao, Receptor Targeted Nanoparticles for In Vivo Imaging of Breast Cancer, Clin. Can cer Res. 15(14), 47224732 ( 2009 ). 139. N. C. Delio lanis, R. Kasmieh, T. Wurdinger, B. A. Tannous, K. Shah, and V. Ntziachristos, "Performance of the redshifted fluorescent proteins in deeptissue molecular imaging applications," J Biomed Opt. 13 (4), 044008 (2008). 140. C. McCann, P. Waterman, J. Figueiredo, E. Aikawa, R. Weissleder, and J. Chen, "ProSense: Combined magnetic resonance and fluorescence imaging of the living mouse brain reveals glioma response to chemotherapy," Neuroimage 45(2), 360 369 (2009). 141. C. Vinegoni, C. Pitsouli, D. Razansky, N. Perrimon, and V. Ntziachristos, In vivo imaging of Drosophila melanogaster pupae with mesoscopic fluorescence tomography," Nat. Methods. 5 (1), 45 47 (2008). 142. A. Heikal, S. Hess, G. Baird, R. Tsien, and W. Webb, "Molecular spectroscopy and dynamics of intrinsically fluorescent proteins: coral red (dsRed) and yellow (Citrine)," P. Natl. Acad. Sci. USA. 97 (22), 1199612001 (2000). 143. M. Garcia Parajo, M. Koopman, E. Van Dijk, V. Subramaniam, and N. Van Hulst, "The nature of fluorescence emission in the red fluorescent protein DsRed, revealed by singlemolecule detection," P. Natl. Acad. Sci. USA. 98(25), 1439214397 (2001). 144. A. Klose and A. Hielscher, "Fluorescence tomography with simulated data based on the equation of radiative transfer," Opt. Lett. 28 (12), 10191021 (2003). 145. A. Hielscher, "Optical tomographic imaging of small animals," Curr. Opin. Biotech. 16 (1), 79 88 (2005). 146. A. Joshi, J. Rasmussen, E. Sevick Muraca, T. Wareing, and J. McGhee, "Radiative transport based frequency domain f luorescence tomography," Phys. Med. Biol. 53 20692088 (2008). 147. D. Grover, J. Yang, D. Ford, S. Tavar, and J. Tower, "Simultaneous tracking of movement and gene expression in multiple Drosophila melanogaster flies using GFP and DsRED fluorescent reporter transgenes," BMC Res. Notes. 2 (1), 58 (2009). 148. Y. Wang, Y. John, and S. Chien, "Fluorescence proteins, livecell imaging, and mechanobiology: seeing is believing," Annu.Rev.Biomed.Eng. 10, 1 38 (2008).

PAGE 121

121 149. Y. Zhang, N. Lin, P. Carroll, G. Chan, B Guan, H. Xiao, B. Yao, S. Wu, and L. Zhou, "Epigenetic blocking of an enhancer region controls irradiationinduced proapoptotic gene expression in Drosophila embryos," Dev. Cell. 14(4), 481493 (2008). 150. C. Vinegoni, C. Pitsouli, D. Razansky, N. Perrimon, and V. Ntziachristos, In vivo imaging of Drosophila melanogaster pupae with mesoscopic fluorescence tomography," Nat Methods 5 (1), 45 47 (2008).

PAGE 122

122 BIOGRAPHICAL SKETCH Yiyong Tan got a B achelor of Engineering (2000) and M aster of Engin eering (2003) in precision instrument engineering in Wuhan University China He later studied at Univ ersity of Uta h USA and got M aster of Science in bioanalytical chemistry (2005) Yiyong Tan began his Ph.D program in b iomedical e ngineering department o f Univ ersity of Florida in 2005 and worked on fluorescence molecular tomography for the past five years and he expect s to obtain his Ph.D. degree in the end of 2010.