Multispectral Fluorescence Imaging of Ovarian Surface for ...

Multispectral Fluorescence Imaging of Ovarian Surface for
Oncologic Tissue Characterization
Timothy Renkoski, MS 1 , Urs Utzinger, PhD 2
1 University of Arizona, College of Optical Sciences, Tucson, Arizona 85721, USA
2 University of Arizona, Biomedical Engineering, Obstetrics & Gynecology, Electrical & Computer Engineering, College of Optical Sciences,
Tucson, Arizona 85721, USA
Email: renkoski@email.arizona.edu
1. Problem
Ovarian cancer causes more deaths than any other gynecological cancer. When ovarian cancer is diagnosed at the
localized stage, it is highly treatable with 5-year survival rate of patients at 92%. However, only 19% of ovarian
cancer cases are discovered at the localized stage, leading to a 5-year survival rate of only 45% [1]. At this time, no
sufficiently accurate screening test exists for ovarian cancer. A reliable screening method for ovarian cancer would
likely yield a significant reduction in fatalities associated with the disease. Multispectral imaging holds promise for
distinguishing cancerous and benign growth tissues from healthy tissue because of the fluorescence signatures of
cellular metabolic cofactors and extracellular matrix components. Ratios of endogenous fluorescent intensities at
different wavelengths have been used to differentiate cancerous and normal ovarian epithelium[2]. Ramanujam and
Wagnières have produced useful reviews on the application of fluorescence spectroscopy and imaging to oncologic
tissue evaluation [3, 4].
2. Methods
In this study, we make use of a particular multispectral imaging system in attempt to differentiate between human
ovarian tissues of normal, benign and cancerous pathologies. The images are of whole human ovaries obtained ex
vivo from thirty consented study patients undergoing oophorectomy. In cases of women at high risk for ovarian
cancer, ovary removal was prophylactic. In other cases removal was deemed necessary due to discovery of ovarian
abnormality. Each ovary was imaged 5 to 15 minutes after disconnection of blood supply in an unlighted room.
The imaging system used a fixed focal distance of 30 cm and a field of view of approximately 8 cm, and the whole
ovaries ranged in size from 2 cm to 12 cm. Histopathology results were obtained on each ovary and used to classify
particular ovaries in 1 of 4 categories (Normal, Cancer, Benign, or Endometriosis)
3. Equipment
The system used to acquire multispectral images of ovarian tissue was provided by Apogen Technologies (a
subsidiary of QinetiQ North America). This system was produced for diagnostic imaging of surface epithelium. It
captures two different sets of images, both fluorescence and reflectance, in rapid succession.
For fluorescence images, illumination is accomplished via a filtered short arc mercury lamp providing excitation
in a narrow band around 365 nm. Bandpass filters are used to allow capture of fluorescence images in 8 slightlyoverlapping
wavelength bands. Together, four of these bands cover the spectrum from 400 nm to 500 nm. The four
remaining filter bands cover the range from 500 nm to 630 nm. The imager produces simultaneous capture of four
spectral images thanks to a novel optical system which splits and filters the collected light in four spatially-separate
but identical images [5]. Each filtered image projects onto one of four quadrants of the camera CCD. (See Figures 1
and 2.)
For reflectance images, illumination is provided by a halogen lamp. This white light is polarized vertically prior
to reaching the sample. Again, light is collected and split into four separate beams for independent filtering. In this
case, the four filters used are wide, bandpass, and centered in the near-infrared (NIR), red, green, and blue spectral
regions. These filters incorporate a polarizer along a fixed direction. Two sets of reflectance images are captured in
succession. One with the collection polarization direction oriented at 90° (crossed) to the illumination polarization
direction and a second with these polarization directions parallel.

Figure 1. Layout of adjacent images on a single CCD
enabling simultaneous multispectral image capture
Figure 2. Rotating filter wheel for independent filtering of
each image quadrant
4. Analysis
Each set of raw image data from the CCD is 1024 by 1280 pixels and contains 4 images--one 512 by 640
pixel image in each quadrant. A corresponding set of dark image data is captured for each image set. The
first step involves separating the quad-image into 4 individual images. Then registration of the images and
correction of distortion are performed. Next, dark correction and flat field correction are applied. Scale
factors correct for wavelength response and integration time.
The preprocessed images can then be compared and analyzed both visually and quantitatively at certain
regions of interest (ROI’s). ROI’s can be identified through multiple methods. The crossed polarizer
reflectance images, viewed in RGB color, assist this process by reducing surface glare and allowing one to
see just beneath the first tissue layer. The images at 8 fluorescence wavelengths can be compared, noting
the relative fluorescence intensities in spatially coincident image regions. These fluorescence images can
also be loaded in the RGB channels to produce a false color image with channels weighted to maximize
contrast. (See Figure 3) Fluorescence image regions with comparatively low signal can be enhanced by
multiplication of the image with (1-Refectance Image) mask.
Figure 3. Sample reflectance and fluorescence ovary images
If no perceptible visual difference in these images can be elucidated for tissue characterization, then
quantitative measures must be implemented. Levels of fluorescent cellular components such as NADH and
FAD are indicators of cellular metabolic rate and have been tied to onset of abnormal tissue growth and
cancer [6-8]. Their spectra have been measured, and an applicable parallel-factor analysis for tissue

fluorescence spectroscopy results has been described [9]. The fluorescence images of this study give us
relative fluorescent intensities for eight different narrow wavelength bands for all imaged regions of tissue.
We can apply parallel factor analysis to the fluorescence image data or simply compose ratios of
fluorescent intensity in different bands in order to tie relative fluorescence levels to relative NADH and
FAD levels. Then we group tissues and attempt to correlate the data to the normal, benign, and cancerous
designations of the imaged tissue as determined by histopathology.
5. Results and conclusions
The results of this spectral image analysis will be presented at conference and include statistics such as the
sensitivity and specificity achieved using histopathology results as the reference standard. Results will also
include how a ROI is best chosen to yield reliable diagnostic results. Discussion would include how the
results compare with what was expected to be observed based on current knowledge of endogenous
fluorescence spectra such as that of collagen, NADH and FAD. Future work would be described, including
how one might construct a minimally invasive device for in vivo diagnostic use.
6. References
[1] Cancer Facts & Figures 2008. Atlanta: American Cancer Society, 2008.
[2] M. Brewer, U. Utzinger, E. Silva, D. Gershenson, R. C. Bast, Jr., M. Follen, and R. Richards-Kortum, "Fluorescence
spectroscopy for in vivo characterization of ovarian tissue," Lasers in Surgery & Medicine, vol. 29, pp. 128-35, 2001.
[3] N. Ramanujam, "Fluorescence spectroscopy of neoplastic and non-neoplastic tissues," Neoplasia (New York), vol. 2, pp.
89-117, 2000.
[4] G. A. Wagnieres, W. M. Star, and B. C. Wilson, "In vivo fluorescence spectroscopy and imaging for oncological
applications," Photochemistry & Photobiology, vol. 68, pp. 603-32, 1998.
[5] M. D. Tocci, "US Patent 7,177,085: Multiple Imaging System and Method for Designing Same," USPTO, Ed., 2007.
[6] R. Drezek, C. Brookner, I. Pavlova, I. Boiko, A. Malpica, R. Lotan, M. Follen, and R. Richards-Kortum,
"Autofluorescence microscopy of fresh cervical-tissue sections reveals alterations in tissue biochemistry with dysplasia,"
Photochemistry & Photobiology, vol. 73, pp. 636-41, 2001.
[7] R. Drezek, K. Sokolov, U. Utzinger, I. Boiko, A. Malpica, M. Follen, and R. Richards-Kortum, "Understanding the
contributions of NADH and collagen to cervical tissue fluorescence spectra: modeling, measurements, and implications,"
Journal of Biomedical Optics, vol. 6, pp. 385-96, 2001.
[8] M. Brewer, U. Utzinger, Y. Li, E. N. Atkinson, W. Satterfield, N. Auersperg, R. Richards-Kortum, M. Follen, and R. Bast,
"Fluorescence spectroscopy as a biomarker in a cell culture and in a nonhuman primate model for ovarian cancer
chemopreventive agents," Journal of Biomedical Optics, vol. 7, pp. 20-6, 2002.
[9] M. Michaelides, "Early Detection of Ovarian Cancer Using Fluorescence Spectroscopy and Parallel Factor Analysis," in
Electrical and Computer Engineering. vol. Master of Science: University of Arizona, 2007.