Laser Microbeam and Medical Program    »     People  |   Education & Dissemination  |   Research & Resources  |   Publications
LAMMP Seminar Video
Multispectral Optical Imaging for the Detection and Delineation of Oral Neoplasia.
Darren Roblyer, PhD

Oral cancer is a significant world health problem. The current standard of care for diagnosis relies on visual inspection of the oral cavity followed by confirmatory biopsies of suspicious areas. Unfortunately, many neoplastic lesions are clinically occult and not visible under white light inspection. Optical imaging has the ability to reveal morphological and biochemical information which may able to aid in oral cancer detection and lesion delineation. Fluorescence imaging at specific excitations can identify locations where extracellular matrix components have been degraded. Reflectance and polarized imaging provide a simple means to probe varying tissue depths and may help visualize new blood vessel growth. We have designed and constructed a widefield microscope with fluorescence, reflectance, and polarized reflectance imaging capabilities for the detection of oral neoplasia. Multispectral image data sets were obtained from 56 patients being treated at the University of Texas MD Anderson Cancer Center. These images show significant differences in contrast between abnormal and normal tissue regions. A classification algorithm was created using features such as pixel ratios, texture metrics, and spatial frequency information from image data. Data was separated into independent training and validation sets to test the accuracy of the algorithm. Sensitivity of 95% and specificity of 95% were obtained in the training set and 94% and 87% were obtaining in a validation set using the single feature of red-to-green pixel ratio from fluorescence images. A visualization technique was developed which maps the probability of disease, obtained from the classification algorithm, onto images of tissue. This technique was validated using carefully constructed pathology maps. This type of analysis promises to aid physicians during tissue resection and screening of cancer.

LAMMP Seminars

View selected seminars online

Supported by


P41EB015890