Development of Quantitative Multi-spectral Fluorescence Endoscopic Imaging for Early Cancer Detection

Abstract

The incidence of esophageal adenocarcinoma has been rising in the western world with a low 5 year survival rate of less than 15%, but it can be treated when diagnosed early. However, conventional white light endoscopy screening has significant limitations because early cancerous lesions usually lie flat on the tissue surface and do not differ in contrast relative to the surrounding area. In an effort to detect early stages of cancer, fluorescence molecular imaging was developed since it improves diagnostic sensitivity and specificity through targeted visualization of multiple biomarkers at the cellular and sub-cellular level. This dissertation presents technologies and algorithms developed for real-time wide-field and quantitative multispectral fluorescence endoscopy for image-guided biopsies and resections. The body of work includes a realistic Barrett's esophagus phantom model development with calibrated fluorescence targets, enhanced visualization and documentation of diagnostic endoscopic records through image mosaicking, real-time compensation for fluorophore emission cross-talk and mitigation of background tissue autofluorescence for high contrast and quantitative molecular endoscopy. Although the scanning fiber endoscope technology forms the basis of the multispectral imaging that was implemented, the concepts, algorithms, and techniques can be applied to a wide range of in vivo and in vitro imaging technologies. In the long term, this work provides the foundation and framework of computer-aided quantitative endoscopy for personalized diagnosis and treatment.