Advanced imaging and artificial intelligence for upper gastrointestinal endoscopy

The endoscopic recognition of early-stage Barrett's carcinoma is challenging due to several factors. Firstly, these abnormalities are subtle in nature, and secondly, the average endoscopist encounters them infrequently. Consequently, early-stage Barrett's carcinomas are frequently overlooked during surveillance endoscopies. The same holds true for early forms of gastric cancer.
The doctoral dissertation titled "Advanced Imaging and Artificial Intelligence for Upper Gastrointestinal Endoscopy" outlines various approaches to enhance the detection of early carcinomas. In the first part of the dissertation (Chapters 1 and 2), an evaluation is conducted to determine whether the use of optical chromoscopy techniques, specifically blue light imaging and linked color imaging, improves the endoscopic recognition of early carcinomas when employed by both expert and non-expert endoscopists.
The second part of the dissertation (Chapters 3 through 5) details various studies that investigate whether computer algorithms, utilizing machine learning techniques, can enhance the endoscopic detection of early-stage Barrett's carcinomas. These computer algorithms can provide assistance to the endoscopist during endoscopic procedures, aiding in the recognition of early carcinomas. The computer-aided detection system described in this thesis has been developed to offer real-time feedback to the endoscopist, identifying any abnormal areas on the endoscopy screen and suggesting preferred locations for biopsy.
In the third part of this dissertation (Chapter 6), an exploration is conducted to determine whether the available images can be utilized more efficiently in the development of deep learning systems for gastrointestinal endoscopy.