As an emerging optical modality, hyperspectral imaging (HSI) holds great promise for early cancer detection and image-guided surgery. The major advantage of HSI is that it is a noninvasive technology that doesn't require any contrast agent, and it combines wide-field imaging and spectroscopy to simultaneously attain both spatial and spectral information from an object in a non-contact way. Light delivered to the tissue surface undergoes multiple elastic scattering and absorption interactions, and part of it returns as diffuse reflectance carrying diagnostic information about the underlying tissue structure and composition. The biochemical and morphological properties of the tissue change during disease progression. Therefore hyperspectral images, which contain high-dimensional spectral information at each image point, can be analyzed for visualization, characterization, and quantification of the disease state in biological tissue.
The overall goal of this dissertation was to investigate the potential of label-free HSI technology combined with machine learning methods as a noninvasive diagnostic tool for quantitative detection and delineation of head and neck cancer. More specifically, this dissertation work has two applications: the early detection of cancer, and surgical guidance. To achieve this, we had four different aims. The first two aims evaluated the diagnostic performance of HSI and machine learning algorithms at differentiating cancer from normal tissue in preclinical animal models, including a subcutaneous cancer model (Aim 1), and a chemically-induced tongue carcinogenesis model (Aim 2). The last two aims investigated the detection and delineation of head and neck cancer in a surgical animal model (Aim 3) and fresh surgical specimens of human patients (Aim 4).