Improving the endoscopic detection of early oesophageal neoplasia

The endoscopic detection of oesophageal cancer is complex; largely owing to the subtle appearances of early oesophageal lesions on endoscopy, as well as clinician experience. Early detection is vital, since lesions confined to the mucosal or superficial layers of the submucosa can be treated with endoscopic eradication therapies to good effect. Conversely, patients presenting with late stage oesophageal cancer have very poor outcomes. Improving the detection of oesophageal cancer requires a multifaceted approach. Since the symptoms patients present with are often vague until the disease has progressed beyond the point that it is curable, developing a way to risk stratify or rationalise patient access to endoscopy, based on objective markers of the presence of serious underlying pathology, is vital to allow adequate resource provision in the modern UK endoscopy unit. In patients who do undergo endoscopy there remains a significant mis-rate of cancers in those with de-novo oesophageal cancer as well as those enrolled in Barrett’s oesophagus surveillance programs. We postulate that advanced imaging technologies, in combination with artificial intelligence systems, may improve the diagnostic performance of endoscopists assessing for oesophageal cancers. This body of work presents a comprehensive review of the literature surrounding the epidemiology, detection, classification and endoscopic treatment modalities for both squamous cell and adenocarcinomas of the oesophagus. It also presents four studies undertaken with the overarching aim of improving the endoscopic detection of oesophageal cancer. The first study presents a methodology for the quantification of a biomarker from gastric aspirate samples and an assessment of whether differences in expression levels can be used to predict the presence of neoplasia in patients with or without Barrett’s oesophagus. The second study investigates the role of a novel, advanced endoscopic imaging technology and whether it improves the diagnostic performance of expert and trainee endoscopists assessing Barrett’s oesophagus for the presence of dysplasia or adenocarcinoma. The final two studies present a significant body of work assessing the feasibility and diagnostic performance of a novel artificial intelligence system designed as part of this thesis, for the detection and characterisation of squamous cell cancer of the oesophagus based on microvascular patterns