Quantification of Multimodal Fluorescence Imaging for Cancer Detection, Guiding Biopsy and Guiding Surgery

Abstract

Fluorescence imaging is emerging for guiding biopsy, early cancer detection and guiding surgery, with the development of new molecularly selective fluorescence agents and fluorescence imaging systems. Accurate quantification of fluorescence imaging is needed to provide diagnostic information, such as tumor stage, tumor location etc. Several fluorescence quantification methods have been addressed in this dissertation, to correlate fluorescence with histology, which is the gold-standard for cancer detection. First, an automatic ratiometric algorithm, target to background (T/B) ratio is developed for esophageal cancer classification and guiding biopsy using single fluorescence labeled peptide. This method may be generally used for other cancer detection in GI tract. Second, a registration pipeline is established to correlate near infrared fluorescence images to histological images, which may contribute to evaluating the sensitivity and specificity of new fluorescence labeled contrast agents in cancer detection. Then T/B ratio is applied to quantify multimodal fluorescence imaging with multiple fluorescence labeled peptides for guiding biopsy and early neoplasia detection in esophagus. Lastly, advanced deep learning algorithms are applied to achieve real-time, automatic frame selection and T/B ratio calculation, which allows computer-aided quantitative fluorescence endoscopy for cancer surveillance. The methodology developed in this work has been used for quantifying multimodal fluorescence imaging, which can assist clinicians in making decisions.