A Biophysics and Fluorescence Guided Surgery Inspired Artificial Intelligence Classifier for the Identification and Classification of Colorectal Cancer

Cancer detection within pathology remains one of the biggest challenges facing clinicians with its presence, or absence, significantly impacting the subsequent patient journey and experience. This dichotomy is demonstrated in the management of significant colorectal lesions with the presence of malignancy, in general, mandating major surgery in the form of a formal bowel resection and lymph node harvesting (at times with the formation of a stoma) whereas benign tumours can be adequately managed via a more limited local excision or even in situ ablation. This thesis introduction begins by exploring the current state of the art with respect to cancer detection and treatment in significant colorectal lesions, including a review of newly emerging Artificial Intelligence (AI) technologies to help clinicians in such endeavours. The second component comprises two parts. The first assesses clinician ability to accurately identify the presence of malignancy within significant rectal lesions in a manner that resembles current diagnostic pathways. As this thesis ultimately revolves around the development of an AI and fluorescence augmented tissue classifier, the second component analyses physician ability to assess and interpret ICG fluorescence signals using the naked eye. The development and validation of a fluorescence augmented AI classifier for the identification and classification of cancer represents the majority of this work and was carried out in the form of a clinical trial (NCT04220242). This thesis details the development and application of a novel, biophysics inspired method of ICG fluorescence angiography assessment to the characterisation of lesions of the colorectum. The expansion of the biophysics inspired fluorescence classifier developed to other areas of the body, namely the liver is also explored. As the principles exploited are based on differentials between tissues of differing nature (cancer vs benign) and are agnostic of tissue location such methodology can be deployed for the characterization and delineation of liver metastases with high levels of accuracy also. Study at a microscopic level was also performed to provide insights in the distribution and trapping of ICG within human tissues which has not previously been done in human colorectal cancer tissue. This thesis concludes with a discussion of the implications of this work on the field of fluorescent guided surgery and details the next steps to ensure continued development and application of the established methodologies.