Identification of Image-based Markers for Oral Pre-cancers and Cancers using in-vivo Multimodal Imaging
MetadataShow full item record
The five-year survival rate for oral cancer is 57% and highly correlated to the stage at diagnosis. Early diagnosis is challenging with current clinical approach of detection based on visual conventional oral examination of oral tissue surface with palpation to guide biopsy as subtle alterations often occur in early neoplasia, despite subsurface cellular and molecular alterations that are known to occur. Robust noninvasive methods are needed with improved performance to detect neoplasia and delineate benign tissue from treatable high-risk precancers and cancers for biopsy guidance. Nonlinear optical microscopy (NLOM) has shown potential to identify optical signatures associated with microstructural and biochemical alterations in cancer, providing in-vivo three-dimensional imaging with contrast from endogenous fluorophores, including tissue metabolites, NADH and FAD. Features revealed by NLOM have shown to parallel histopathology, however, studies are few and have focused on epithelial cells or stromal collagen but rarely together. While NLOM demonstrates potential in detection of neoplasia, image areas are spatially small. Practical implementation will require strategies for assessing larger areas of mucosal surface. NLOM encompassing epithelial and stromal-based metrics was investigated to delineate neoplasia from normal and benign tissue. To address limited image views, a strategy for multimodal multiscale imaging, combining large area fluorescence imaging with subsurface microscopic evaluation of neoplastic features by NLOM, was devised. In-vivo NLOM revealed high sensitivity/specificity for neoplasia. Novel quantitative features from multiple layers were defined, including alterations in epithelial-connective tissue interface (ECTI) and associated with neoplastic cell clusters in 3D. Two approaches for wide-field fluorescence imaging (endogenous and using contrast for glucose metabolism), resulted in high sensitivity, an attribute desired in large area imaging to guide NLOM to suspicious sites. A study was conducted combining WF imaging with NLOM- results indicating feasibility for multiscale imaging with high sensitivity/specificity. This research has resulted in new knowledge regarding the use of label-free NLOM in epithelial neoplasia and established a proof-of-concept multiscale approach for detection of oral neoplasms. Concepts developed in this dissertation are expected to be beneficial in further development of multimodal image based cancer detection method as well as future research on understanding how neoplastic changes occur in epithelium.