Application of artificial intelligence in oncologic surgery of the upper gastrointestinal tract
Background: The application of artificial intelligence (AI) to minimally invasive surgery has the potential to improve surgical safety, support intraoperative decision making, and reduce operative complications. Computer vision and machine learning are subfields or AI, focused on making statistical inferences and generating predictive calculations about relevant patterns in data. These patterns are extracted from the data based on annotations of clinically relevant target features. While there are numerous real-time applications of AI to medicine, particularly in diagnostic specialties, surgical AI is currently predominantly limited to pre- and postoperative analysis of patient data. Intraoperative deployment of AI, based on retrospective video and image analysis of minimally invasive procedures, requires the fundamental comprehension of surgical workflow. Therefore, the automated detection and prediction of operative phases, tracking of surgical instruments, differentiation of tissue and analysis of tool–tissue interactions has been the central research focus. While these tasks promise tremendous clinical value, surgical AI is still predominantly limited to highly standardized, routine procedures such as laparoscopic cholecystectomy. To reveal its true risk mitigation potential, AI must be generalizable to more complex, and specifically oncological procedures. Objective: This article provides as a review of the existing applications of AI to oncological foregut surgery and illustrates the technology’s current limitations and future potential.
