Biophysical basis of skin cancer detection using Raman spectroscopy

The goal of this dissertation is to study the potential of Raman spectroscopy in improving the clinical diagnosis of skin cancer, including two main applications: noninvasive screening of melanoma skin cancer and surgical margin detection of nonmelanoma skin cancer. Skin cancer is the most common type of malignancy, accounting for over 5.4 million cases and 10 thousand deaths per year in the United States alone. Like most cancers, the current “gold standard” diagnosis relies on biopsy and histopathology, which is invasive, time-consuming, and costly. Moreover, large numbers of benign lesions are biopsied for melanoma diagnosis, resulting in substantial financial burden and patient discomfort. Therefore, an urgent need exists to develop a noninvasive, fast, and accurate method for skin cancer detection. The first part of the dissertation focuses on exploring the biophysical origin of in vivo melanoma detection. Our group has previously reported on the development of a clinical Raman spectroscopy system towards spectral biopsy of skin; however, the biochemical changes that Raman spectroscopy relies on for accurate melanoma diagnosis remained unclear. As a result, we proposed a biophysical inverse model to address this issue. To build the model, we established a custom confocal Raman microscope to extract in situ human skin constituents spanning normal and various diseased states. Our results indicate collagen, elastin, keratin, cell nucleus, triolein, ceramide, melanin, and water are the most important model components. Furthermore, collagen and triolein are the most relevant markers to discriminate malignant melanoma from benign nevi. The second part of the dissertation discusses the biophysical basis of nonmelanoma skin cancer margin delineation. We discovered the diagnostic markers to accurately differentiate tumor from normal skin, which is critical to maximize positive patient outcomes in skin cancer surgery. The biochemical changes derived from our model were highly correlated with histopathological diagnosis. We further demonstrated the feasibility of a superpixel acquisition approach for rapid classification of tumor boundaries in skin biopsies. Our results suggest Raman spectroscopy will be a powerful tool for intraoperative surgical guidance