Fluorescence spectroscopy using excitation and emission matrix for quantification of tissue native fluorophores and cancer diagnosis

Abstract

Native fluorescence spectrum of normal and cancerous human prostate tissues is studied to distinguish between normal and cancerous tissues, and cancerous tissues at different cancer grade. The tissue samples were obtained from Cooperative Human Tissue Network (CHTN) and National Disease Research Interchange(NDRI). An excitation and emission matrix (EEM) was generated for each tissue sample by acquiring native fluorescence spectrum of the sample using multiple excitation wavelengths. The non-negative matrix factorization algorithm was used to generate fluorescence EEMs that correspond to the fluorophores in biological tissues, including tryptophan, collagen, elastin, nicotinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD) and the background paraffin. We hypothesize that, as a consequence of metabolic changes associated with the development of cancer, the concentrations of NADH and FAD are different in normal and cancerous tissues, and also different for different cancer grades. We used the ratio of the abundances of FAD and NADH to distinguish between normal and cancerous tissues, and the tissue cancer grade. The FAD-to-NADH ratio was found to be the highest for normal tissue and decreased as the cancer grade increased.

Full text

Fluorescence spectroscopy using excitation and emission matrix for
quantification of tissue native fluorophores and cancer diagnosis
Binlin Wua, S. K. Gayenb, and M. Xu*c
aBiochemistry Department, Weill Cornell Medical College, New York, NY, USA 10065; bPhysics
department, the City College and the Graduate Centre of The City University of New York, New
York, NY, USA 10031; cPhysics Department, Fairfield University, Fairfield, CT, USA 06824
ABSTRACT
Native fluorescence spectrum of normal and cancerous human prostate tissues is studied to distinguish between normal
and cancerous tissues, and cancerous tissues at different cancer grade. The tissue samples were obtained from
Cooperative Human Tissue Network (CHTN) and National Disease Research Interchange(NDRI). An excitation and
emission matrix (EEM) was generated for each tissue sample by acquiring native fluorescence spectrum of the sample
using multiple excitation wavelengths. The non-negative matrix factorization algorithm was used to generate
fluorescence EEMs that correspond to the fluorophores in biological tissues, including tryptophan, collagen, elastin,
nicotinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD) and the background paraffin. We
hypothesize that, as a consequence of metabolic changes associated with the development of cancer, the concentrations
of NADH and FAD are different in normal and cancerous tissues, and also different for different cancer grades. We used
the ratio of the abundances of FAD and NADH to distinguish between normal and cancerous tissues, and the tissue
cancer grade. The FAD-to-NADH ratio was found to be the highest for normal tissue and decreased as the cancer grade
increased.
Keywords: Optical biopsy, non-negative matrix factorization, fluorescence, spectroscopy, excitation and emission
matrix, cancer diagnosis, prostate cancer, near-infrared
1. INTRODUCTION
Fluorescence spectroscopy has been studied for developing modalities for cancer diagnosis in the past decades [1-15].
The fluorescence spectra of untreated human tissues, sometimes referred to as autofluorescence, is a convolution of
spectra of native fluorophores, which include tryptophan, collagen, elastin, reduced nicotinamide adenine dinucleotide
(NADH), flavin adenine dinucleotide (FAD) etc. The changes in the fluorescence signal due to native fluorophores are
believed to be related to the biochemical and molecular changes in the tissue, which occur during the development of
cancer.
Prostate cancer is the most common cancer in American men other than skin cancer [16]. Gleason score is a well-known
grading system that describes the evolution of malignant prostate tumor and reveal the change of prostate tissue [17]. A
Gleason score is the sum of two Gleason grades (1~5) of the two most common tumor patterns. For the prostate tissue
with Gleason grade 1 (corresponding to early stage), the pattern of the tissue consists of evenly placed uniform gland
cells supported by an extracellular matrix of collagen fiber [17, 18]. As the grade advances, the cancer cells proliferate,
the cell density increases, the cellular nuclei become non-uniform [17] and content of collagen decreases [17, 18].
NADH and FAD are involved in the oxidation of fuel molecules. Redox fluorometry based on autofluorescence of
NADH and FAD has been a useful tool for studying cellular energy metabolism [1, 19-22]. Reduced NADH and
oxidized FAD are fluorescent, while oxidized NADH and reduced FAD are non-fluorescent [1]. Redox ratio
([FAD]/[NADH]) decreases as cancer develops [1, 5, 23-26].
Even though differences in autofluorescence spectra measured from normal, benign and malignant tissues have been
demonstrated, it is difficult to interpret the sources of the differences directly, because the fluorescence signal is
significantly distorted by the intrinsic absorber and scatterers in tissues [6, 27]. Differences exist in the spectra acquired
from different tissue samples of the same type, such as cancer or normal.
* mxu@fairfield.edu
Photonic Therapeutics and Diagnostics X, edited by Bernard Choi, et al., Proc. of SPIE
Vol. 8926, 89261M · © 2014 SPIE · CCC code: 1605-7422/14/$18 · doi: 10.1117/12.2040985
Proc. of SPIE Vol. 8926 89261M-1
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

In this manuscript, we retrieve the concentrations of NADH and FAD using non-negative matrix factorization (NMF)
[28] from fluorescence excitation and emission matrix (EEM) [5, 6, 29], and correlate the redox ratio ([FAD]/[NADH])
with tumor grade (using Gleason score) of prostate tissue. An EEM is formed by putting together emission spectra at
successive excitation wavelengths. Each row in the EEM corresponds to an emission spectrum at a specific excitation
wavelength. Each column in the EEM corresponds to an excitation spectrum.
The rest of the manuscript is organized as follows. Section 2 introduces the NMF algorithm. Section 3 shows the
experiments and results. Section 4 includes a summary and discussions.
2. METHODOLOGY
Non-negative matrix factorization (NMF) is a statistical method to solve the problem of blind source separation (BSS),
which is also known as blind signal separation. BSS is a general problem in information theory that seeks to separate
constituent signals and their abundances from the weighted mixtures of the individual constituent signals. In matrix
notation, it may be expressed as follows. The mixed signals X = WH, where columns of W are the individual signals or
emission spectra in this study, H is a mixing or weighting matrix, columns of X are the mixture signals. In fluorescence
spectral study, if M excitation wavelengths are used, and the emission spectra which are linear combinations of J basis
spectra, are measured at N wavelengths, NJ
WR
×
∈, JM
H
R×
∈
, NM
X
R×
∈
, and J < min (M, N). The objective of BSS is
to factorize the matrix X into W and H, where W and H are non-negative [30].
NMF does not imply any relationship between the retrieved basis signals; instead, it just enforces non-negativity of the
basis signals and their weights [31] and find solutions through an iterative process using different methods such as
multiplicative update method [28] or alternating least squares method [32, 33]. NMF learns parts-based representation of
the signal and find the hidden parts which may be more recognizable [28].
NMF has been extensively used in such diverse applications as, facial image recognition [28], genetic and molecular
pattern discovery [34], spectral data analysis [35], cancer class discovery [36] and diffuse optical imaging [31].
The rationale of using NMF in spectral analysis is based on the following considerations: (a) the fluorescence signal and
fluorophore concentrations are positive, therefore it is natural to use non-negativity constraints; (2) the unknown
constituents in the mixed compounds of complicated environment such as biological cells and tissues may show
different spectra from pure individual chemicals due to the influence of the complex surroundings, and NMF may resolve
the “real” spectra in the mixed environment [6, 28].
3. EXPERIMENTS AND RESULTS
Four normal tissue samples and 4 cancerous tissue samples with different grades fixed in paraffin wax were used in this
initial investigation of the efficacy of NMF approach to optical biopsy. The tissue specimens were obtained from
National Disease Research Interchange (NDRI) and Cooperative Human Tissue Network (CHTN) under Institutional
Review Board (IRB) approvals at the City College of New York. Perkin-Elmer LS-50 spectrometer was used to measure
the Fluorescence spectra of the samples were measured for eleven (11) excitation wavelengths (300, 325, 340, 380, 400,
440, 460, 460, 480, 510, and 532 nm). Narrow-band (NB) interference filters (FWHM ~ 10 nm) were used on the
excitation side, and long-pass filters were used on the detection side to reduce the stray light and improve the signal-to-
noise ratio. The transmission spectra of the long-pass filters were measured and used to correct for the distortion in the
fluorescence spectra due to long-pass filters. The fluorescence spectra of pure fluorophores (tryptophan, collagen,
elastin, NADH and FAD) and paraffin wax were measured under the same conditions for comparison.
Fluorescence spectrum of Rhodamine-6G in methanol was also measured under the same conditions to calibrate the
tissue fluorescence spectra. The distortion in the spectra due to non-uniformity of light source at different wavelengths
was corrected, by comparing the Rhodamine-6G spectra measured using our system and the standard Rhodamine spectra
obtained from literature.
The calibrated fluorescence spectra recorded at different excitation wavelengths were put together to form excitation and
emission matrix (EEM). EEMs at smaller incremental steps of excitation wavelength than the recorded data were
generated using interpolation. The contours are plotted for demonstration.
Proc. of SPIE Vol. 8926 89261M-2
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

Wavelength (nm)
Wavelengt h (nm)
300 350 400 450 500 550 600 650 700 750
300
350
400
450
500
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
(a)
Wavelength (nm)
Wavelength (nm)
300 350 400 450 500 550 600 650 700 750
300
350
400
450
500
200
400
600
800
1000
1200
1400
1600
1800
(b)
Emission wavelength (nm)
Emission wavelength (nm)
Excitation wavelength (nm)Excitation wavelength (nm)
Fig. 1. Typical fluorescence EEM of (a) normal prostate tissue and (b) cancerous prostate tissue.
The contours of typical fluorescence EEMs of normal and cancerous cancer tissues are shown in Fig. 1. The EEMs of
fluorescence spectra show that at the peak around 460nm (corresponding to NADH) relative to the peak around 525 nm
(corresponding to FAD) is lower in normal prostate tissue than that in malignant tissue.
The EEM data were analyzed using NMF. Six (6) basis fluorescence EEMs were retrieved by NMF, and shown in Fig.
2. The six fluorescence EEMs measured from the pure chemicals are shown in Fig. 3 for comparison. The weights of the
six basis EEMs represent the contributions to the recorded tissue fluorescence signal due to the fluorophores, and are
considered to be proportional to the concentrations of them in the specimens.
Proc. of SPIE Vol. 8926 89261M-3
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
2
4
6
8
x 10
-4
2
4
6
8
x 10
-4
2
4
6
8
10
12
x 10
-3
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
2
4
6
8
10
12
x 10
-4
5
10
15
x 10
-4
2
4
6
8
x 10
-3
(a) (b)
(c) (d)
(e) (f)
Fig. 2. NMF-retrieved basis fluorescence spectra EEM (Ex.: Excitation, Em.: Emission).
Comparison of Fig. 2 and Fig. 3 suggests that the first NMF-retrieved component [Fig. 2(a)] corresponded to
NADH. The third and fourth basis components shown in Fig. 2(c) and 2(d) were both similar to collagen and elastin.
Since the peak positions of the fluorescence spectra due to collagen and elastin were very close, it is not easy to
separate these two components. However, according to the chemical spectra in Fig. 3, the collagen fluorescence
spectrum is slightly broader than that of elastin. The correlation coefficients were also calculated to identify the
retrieved components. The correlation coefficients indicate Fig. 2(c) and Fig. 2(d) may be associated with collagen
and elastin, respectively. According to the (excitation) peak position, the 2nd basis EEM shown in Fig. 2(b)
correspond to paraffin. The 5th and 6th basis EEMs [Fig. 2(e) and Fig. 2(f)] are very similar. The 5th basis EEM
clearly corresponds to FAD. However, the corresponding scores found in matrix H were almost 0 for all specimens.
Even though the 6th retrieved component seems to be superposition of contributions from FAD and tryptophan,
tryptophan is expected to be much weaker using the excitation wavelengths that are far from its excitation peak (~
280 nm). Therefore, we associate the 6th basis EEM with FAD, with tryptophan contribution neglected. So the
redox ratio would be the retrieved weight for the 6th basis EEM (FAD) divided by the weight for the 1st basis EEM
(NADH).
Proc. of SPIE Vol. 8926 89261M-4
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
Em. Wavelength (nm)
Ex. Wavelength (nm)
300 400 500 600 700
300
350
400
450
500
0.5
1
1.5
2
x 10
4
2000
4000
6000
200
400
600
800
1000
200
400
600
800
1000
1200
100
200
300
1
2
3
x 10
4
(a) (b)
(c) (d)
(e) (f)
Fig. 3. Measured intrinsic fluorescence EEMs of the biochemicals. (a) – (f) are EEMs acquired from tryptophan, collagen,
elastin, NADH, FAD and paraffin, respectively.
Table 1. Redox ratios calculated using the concentrations of NADH and FAD extracted from the fluorescence EEM by
NMF. (N: normal, C: cancerous)
Sample # 1 2 3 4 5 6 7 8
Gleason
score
N N N N 6 (C) 8 (C) 8 (C) 9 (C)
Redox
Ratio
25.01 3.07 4.28 3.96 2.18 0.46 0.48 0.36
The redox ratios were calculated for all the 8 specimens using relative concentrations of NADH and FAD, which
were retrieved as the weights of the first and sixth basis spectra. The redox ratios are shown in Table 1 along with
the Gleason scores obtained from the providers (NDRI and CHTN). The redox ratio due to the first specimen was
particularly high. Normal tissues were obtained near the sites of tumors. Heterogeneities existed in the specimens of
Proc. of SPIE Vol. 8926 89261M-5
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

the same tissue type. It is possible there were slightly more biochemical changes in sample 2-4 than in sample 1,
which may be detected by fluorescence spectroscopy. The relative concentrations of NADH and FAD retrieved by
NMF are scatter plotted in Fig. 4. The normal and cancerous tissues are clearly separated.
00.5 11.5 22.5
x 10
6
0
2
4
6
8
10
12
14 x 10
5
Concentration of NADH (arb. units )
Concent ration of FAD (arb. units )
Fig. 4. Scatter plot of concentration of FAD vs. NADH. The straight line separate normal and cancerous specimens (o:
normal, *: cancer).
4. SUMMARY AND DISCUSSIONS
NMF is shown to be an efficient method for unmixing different basis spectra of tissue native fluorophores. The Redox
ratios obtained using NMF retrieved relative concentrations of basis fluorophores are well correlated with the Gleason
scores. This indicate that optical biopsy using fluorescence EEM may be used not only to differentiate normal and
cancerous tissues, but also to estimate the aggressiveness of the tumor malignancy for prognosis.
The conventional spectroscopic methods used in the studies of tissue optics are absorption (optical density, i.e. O.D.),
emission and excitation measurements. In the absorption measurements, only a few chromophores are detectable in
tissues. Even though the fluorescence emission (excitation) spectroscopy can detect fluorophores at very low level,
because the pumping (detecting) wavelength is fixed, the optimal emission (excitation) signal can only be acquired for
one or two fluorophores [37].
In the contrast, EEM provides the most spectral information. It ensures the coverage of all endogenous fluorophores. As
can be seen in the EEM, the most information-rich line is not necessarily a line with slope 1. The redundancy in the
EEM can make the analysis much more robust, which is particularly important because fluorescence signal of some
native fluorophores is weak and subtle, and the fluorescence is altered by the local complex environment. The subtle
information can be extracted from the background using NMF.
In our future study, fresh tissues will be used instead of the fixed tissues so that the biology of the tissue is better
preserved, and the paraffin background is removed. Eventually in vivo study will be conducted. More excitation
wavelengths and better chromatic light source may be used if available, therefore higher-resolution EEM data can be
measured. At the same time, spectrograph system may be used to measure spectral data in a two-dimensional area, which
may be used to detect cancer margin, 2D cancer prognosis, and help make decisions in surgery. In particular, optical
biopsy may be combined with diffuse optical imaging technology [31, 38, 39], so that spectral data can be taken from an
accurate suspicious location is already identified by the imaging technology. Therefore, multiple biopsies are avoided,
which is of great clinical interest.
Proc. of SPIE Vol. 8926 89261M-6
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

ACKNOWLEDGEMENT
B. Wu and S. K. Gayen acknowledge the support by USAMRMC under Contract Number W81XWH-07-1-0454. M. Xu
acknowledges Research Corporation, NIH (1R15EB009224), and DOD (W81XWH-10-1-0526) for their support.
REFERENCES
[1] J. R. Lakowicz, Principles of Fluorescence Spectroscopy (Plenum Press, New York, 1983).
[2] R. R. Alfano, D. Tata, J. Cordero, P. Tomashefsky, F. Longo, and M. Alfano, "Laser induced fluorescence
spectroscopy from native cancerous and normal tissue," IEEE J. Quantum Electron. 20, 1507–1511 (1984).
[3] R. R. Alfano, G. C. Tang, A. Pradhan, W. Lam, D. S. J. Choy, and E. Opher, "Fluorescence spectra from
cancerous and normal human breast and lung tissues," IEEE J. Quantum Electron. 23, 1806 (1987).
[4] N. Ramanujam, M. F. Mitchell, A. Mahadevan, S. Warren, S. Thomsen, E. Silva, and R. Richards-Kortum, "In
vivo diagnosis of cervical intraepithelial neoplasia using 337-nm-excited laser-induced fluorescence," PNAS 91,
10193–10197 (1994 ).
[5] D. L. Heintzelman, R. Lotan, and R. R. Richards-Kortum, "Characterization of the Autofluorescence of
Polymorphonuclear Leukocytes, Mononuclear Leukocytes and Cervical Epithelial Cancer Cells for Improved
Spectroscopic Discrimination of Inflammation from Dysplasia," Photochem. and Photobiol. 71, 327–332 (2000).
[6] I. Georgakoudi, B. C. Jacobson, M. G. Muller, E. E. Sheets, K. Badizadegan, D. L. Carr-Locke, C. P. Crum, C.
W. Boone, R. R. Dasari, J. V. Dam, and M. S. Feld, "NAD(P)H and Collagen as in vivo Quantitative Fluorescent
Biomarkers of Epithelial Precancerous Changes," Cancer Res. 62, 682–687 (2002).
[7] S. K. Chang, Y. N. Mirabal, E. N. Atkinson, D. Cox, A. Malpica, M. Follen, and R. R. Richards-Kortum,
"Combined reflectance and fluorescence spectroscopy for in vivo detection of cervical pre-cancer," J. Biomed.
Opt. 10, 024031 (2005).
[8] K. Harbig, B. Chance, A. G. B. Kovac, and M. Reivich, "In vivo measurement of pyridine nucleotide fluorescence
from cat brain cortex," J. Appl. Physiol. 41, 480-488 (1976).
[9] B. Chance, B. Schoener, R. Oshino, F. Itshak, and Y. Nakase, "Oxidation-reduction ratio studies of mitochondria
in freeze-trapped samples. NADH and flavoprotein fluorescence signals.," J. Biol. Chem. 254, 4764-4771 (1979).
[10] R. M. Cothren, R. R. Richards-Kortum, M. V. Sivak, M. Fitzmaurice, R. P. Rava, G. A. Boyce, G. B. Hayes, M.
Doxtader, R. Blackman, T. Ivanc, M. S. Feld, and R. E. Petras, "Gastrointestinal tissue diagnosis by laser-induced
fluorescence spectroscopy at endoscopy," Gastrointest. Endosc. 36, 105-111 (1990).
[11] R. Richards-Kortum, R. P. Rava, R. E. Petras, M. Fitzmaurice, M. Sivak, and M. S. Feld, "Spectroscopic
diagnosis of colonic dysplasia," Photochem. Photobiol. 53, 777–786 (1991).
[12] R. R. Alfano and Y. Yang, "Stokes shift emission spectroscopy of human tissue and key biomolecules," IEEE J.
Select. Topics Quantum Electron. 9, 148-153 (2003).
[13] X. Shao, W. Zheng, and Z. Huang, "Polarized near-infrared autofluorescence imaging combined with near-
infrared diffuse reflectance imaging for improving colonic cancer detection," Opt. Express 18, 24293-24300
(2010).
[14] A. J. Thompson, S. Coda, M. B. Sørensen, G. Kennedy, R. Patalay, U. Waitong-Brämming, P. A. A. D. Beule, M.
A. A. Neil, S. Andersson-Engels, N. Bendsøe, P. M. W. French, K. Svanberg, and C. Dunsby, "In vivo
measurements of diffuse reflectance and time-resolved autofluorescence emission spectra of basal cell
carcinomas," J. Biophotonics 5, 240-254 (2012).
[15] S. Coda, A. J. Thompson, G. T. Kennedy, K. L. Roche, L. Ayaru, D. S. Bansi, G. W. Stamp, A. V.
Thillainayagam, P. M. W. French, and C. Dunsby, "Fluorescence lifetime spectroscopy of tissue autofluorescence
in normal and diseased colon measured ex vivo using a fiber-optic probe," Biomed. Opt. Express 5, 515-538
(2014).
[16] http://www.cancer.org/cancer/prostatecancer/detailedguide/prostate-cancer-key-statistics.
[17] D. F. Gleason and G. T. Mellinger, "Prediction of prognosis for prostatic adenocarcinoma by combined
histological grading and clinical staging," J. Urol. (Baltimore) 111, 58–64 (1974).
[18] C. Morrison, J. Thornhill, and E. Gaffney, "The connective tissue framework in the normal prostate, BPH and
prostate cancer: analysis by scanning electron microscopy after cellular digestion," Urol. Res. 28, 304–307
(2000).
Proc. of SPIE Vol. 8926 89261M-7
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms

[19] R. S. Balaban and L. J. Mandel, "Optical methods for the study of metabolism in intact cells," in Noninvasive
Techniques in Cell Biology, J. K. Foskett and S. Grinstein, eds. (Wiley-Liss, New York, 1990).
[20] B. Chance, "Optical Method," Annu. Rev. Biophys. Biophys. Chem. 20, 1-28 (1991).
[21] B. R. Masters, "Noninvasive corneal redox fluorometry," Curr. Top. Eye Res. 4, 139-200 (1984).
[22] B. R. Masters and B. Chance, "Redox confocal imaging: intrinsic fluorescent probes of cellular metabolism," in
Fluorescent and Luminescent Probes for Biological Activity, W. T. Mason, ed. (Academic Press, New York,
1993), pp. 44-57.
[23] B. Chance, B. Schoener, R. Oshino, F. Itshak, and Y. Nakase, "Oxidation-reduction ratio studies of mitochondria
in freeze-trapped samples. NADH and flavoprotein fluorescence signals.," J. Biol. Chem. 254, 4764-4771 (1979).
[24] B. Chance, "Metabolic heterogeneities in rapidly metabolizing tissues," J. Appl. Cardiol. 4, 207–221 (1989).
[25] M. C. Skala, K. M. Riching, A. Gendron-Fitzpatrick, J. Eickhoff, K. W. Eliceiri, J. G. White, and N. Ramanujam,
"In vivo multiphoton microscopy of NADH and FAD redox states, fluorescence lifetimes, and cellular
morphology in precancerous epithelia," PNAS 104, 19494–19499 (2007).
[26] M. C. Skala and N. Ramanujam, "Multiphoton Redox Ratio Imaging for Metabolic Monitoring in vivo," Methods
Mol. Biol. 594, 155-162 (2010).
[27] G. M. Palmer, P. J. Keely, T. M. Breslin, and N. Ramanujam, "Autofluorescence spectroscopy of normal and
malignant human breast cell lines," Photochem. and Photobiol. 78, 462-469 (2003).
[28] D. D. Lee and H. S. Seung, "Learning the parts of objects by non-negative matrix factorization," Nature 401, 788-
791 (1999).
[29] C. N. Ho, G. D. Christian, and E. R. Davidson, "Application of the method of rank annihilation of quantitative
analyses of multicomponent fluorescent data from the video fluorometer," Anal. Chem. 50, 1108-1113 (1978).
[30] M. W. Berry, M. Browne, A. N. Langville, V. P. Pauca, and R. J. Plemmons, "Algorithms and applications for
approximate nonnegative matrix factorization," Comp. Stat. Data Anal. 52, 155-173 (2007).
[31] B. Wu, M. Alrubaiee, W. Cai, M. Xu, and S. K. Gayen, "Diffuse optical Imaging using decomposition methods,"
Int. J. Opt. 2012, 185435 (2012).
[32] P. Paatero, "The multilinear engine: a table-driven least squares program for solving multilinear problems,
including the n-way parallel factor analysis model," J. Comp. Graph. Stat. 8, 854–888 (1999).
[33] P. Paatero and U. Tapper, "Positive matrix factorization: a non-negative factor model with optimal utilization of
error estimates of data values," Environmetrics 5, 111–126 (1994).
[34] J.-P. Brunet, P. Tamayo, T. R. Golub, and J. P. Mesirov, "Metagenes and molecular pattern discovery using
matrix factorization," PNAS 101, 4164-4169 (2004).
[35] V. P. Pauca, J. Piper, and R. J. Plemmons, "Nonnegative matrix factorization for spectral data analysis," Lin. Alg.
Appl. 416, 29-47 (2006).
[36] Y. Gao and G. Church, "Improving molecular cancer class discovery through sparse non-negative matrix
factorization," Bioinformatics 21, 3970-3975 (2005).
[37] Y. Pu, W. Wang, and R. R. Alfano, "Optical detection of meat spoilage using fluorescence spectroscopy with
selective excitation wavelength," Appl. Spectrosc. 67, 210-213 (2013).
[38] B. Wu, W. Cai, M. Alrubaiee, M. Xu, and S. K. Gayen, "Time reversal optical tomography: locating targets in a
highly scattering turbid medium," Opt. Express 19, 21956–21976 (2011).
[39] B. Wu, W. Cai, and S. K. Gayen, "Three-dimensional localization of fluorescent targets in turbid media using
time reversal optical tomography," Appl. Phys. Lett. 101, 251103 (2012).
Proc. of SPIE Vol. 8926 89261M-8
Downloaded From: http://proceedings.spiedigitallibrary.org/ on 03/04/2014 Terms of Use: http://spiedl.org/terms