Artificial Intelligence for Surgical Training Towards Safer and Effective Sub-mucosal Dissection

Endoscopic submucosal dissection (ESD) is a widely adapted, minimally invasive procedure for treating superficial cancers. However, ESD is particularly challenging due to the thin muscular layer and narrow lumen of the oesophagus, which is surrounded by vital organs, including the lungs.

In the UK, a handful of centres offer ESD, and ESD training is available at even fewer centres with systematic supervision. This can be even more challenging in low-middle-income countries and countries with higher health inequalities.

Empowering clinical fellows through effective training can minimise the current gap and help scale up centres’ potential to safely and effectively treat patients with cancer at an early stage. ESD involves various phases, including lesion marking, submucosal injection, incision and dissection. These sub-tasks can be non-trivial for trainee clinicians and hence developing an AI system, which provides feedback for each process, including any missing phases and the amount of time taken on each phase, would make the treatment much more accessible.

As part of this project, a phase recognition AI system with interactive feedback will be developed. This will include training the system for detection and boundary identification of the lesion area through to guidance for dissection. All associated risks will be flagged as feedback for the clinician.

Project website

https://wun.ac.uk/wun/research/view/artificial-intelligence-for-surgical-training-towards-safer-and-effective-sub-mucosal-dissection/