Artificial intelligence (AI) in endoscopy refers to the capacity of computer
algorithms using “machine learning” to aid in the detection and characterization of
lesions in the digestive tract. The field of AI in endoscopy is expanding at a very rapid
pace and, while the potential for development is enormous, the only validated
applications currently available in everyday practice are computer-assisted detection
and characterization of colonic polyps. The main advantage of machine learning is the
capability of analyzing vast quantities of data to detect patterns that are not readily
available to the endoscopist, thus theoretically increasing the accuracy of detection and
diagnosis of the predefined lesion. However, the current technology is still heavily
reliant on adequate image databases which have to be appraised by expert endoscopists
before the algorithms can be trained on these datasets. Furthermore, each individual
algorithm is trained to answer very specific questions, usually in a binary fashion (i.e. –
is the polyp neoplastic or hyperplastic?).
Endoscopists need to be aware of the developments in the field, because in the near
future such applications as detection and characterization of early esophageal and
gastric cancer might also be included in their diagnostic armamentarium. Finally,
several ethical and practical questions regarding the implementation of AI-based
diagnosis and treatment in everyday practice need to be addressed by the academic and
medical community before the large-scale adoption of AI in endoscopy becomes a
reality.
Keywords: Algorithms, Artificial intelligence, Cancer, Colonoscopy, Computerassisted detection, Computer-assisted diagnosis, Deep learning, Endoscopy.
