Date of Award
Doctor of Philosophy (PhD)
Tissue optical scattering has recently emerged as an important diagnosis parameter associated with early tumor development and progression. To characterize the differences between benign and malignant colorectal tissues, we have created an automated optical scattering coefficient mapping algorithm using an optical coherence tomography (OCT) system. A novel feature called the angular spectrum index quantifies the scattering coefficient distribution. In addition to scattering, subsurface morphological changes are also associated with the development of colorectal cancer. We have observed a specific mucosa structure indicating normal human colorectal tissue, and have developed a real-time pattern recognition neural network to localize this specific structure in OCT images, enabling classification of the morphological changes associated with the progression of human colon cancer. Differentiating normal from malignant tissues is critically important, however, identifying different subtypes of abnormalities is also useful in clinical diagnosis. We have designed a feature extraction method using texture features and computer-vision related features to characterize different types of colorectal tissues. We first ranked these features according to their importance, then trained two classifiers: one for normal vs. abnormal, and the other one for cancer vs. polyp, where polyp is a pre-cancer marker. In assessing tissue abnormalities, optical absorption reveals contrast related to tumor microvasculature and tumor angiogenesis. Spatial frequency domain imaging (SFDI), a powerful wide field, label-free imaging modality, is sensitive to both absorption and scattering. We designed a computer-aided diagnostic algorithm, AdaBoost, to use multispectral SFDI imaging for ex vivo assessment of different types of colorectal tissues, including normal and cancerous tissue and adenomatous polyps. For diagnosis of human ovarian cancer, we first designed a histogram-based feature extraction algorithm. Then we trained and tested traditional machine learning methods utilizing these histogram features for ovarian cancer diagnosis. We also explored the use of these features in characterizing human fallopian tubes, which are believed to be the origin of the most lethal subtype of human ovarian cancers.
Hong Chen, Ian S. Hagemann, Matthew Lew, Yuan-Chuan Tai,