Deep learning approach for critical exposure during division of the inferior mesenteric artery in colorectal surgery.

With growth of minimally invasive surgery, there is an opportunity to analyze and learn from intraoperative videos. These videos contain a wealth of data which can inform surgical techniques. In colorectal surgery, a common critical step is inferior mesenteric artery (IMA) division. However, optimal retraction and relative tissue-relationship metrics for critical exposure during IMA division are not defined. We leverage intraoperative robotic surgery videos and artificial intelligence (AI) models to define and quantify key angles and surgical approaches during IMA division. Videos of robotic-assisted low anterior resections and sigmoid colectomies were annotated for the moments during and preceding IMA division. Deep learning algorithms and three-dimensional mesh modeling quantified variables including three-dimensional and two-dimensional angles of stapler and IMA, surface area and width of IMA before division, and visualized area under the IMA during tissue exposure. We derived an AI-object detection model to capture optimal angles, relative instrument relationships, and tissue metrics during IMA division. Results showed consistency of intraoperative angles and intersection vectors of the IMA and stapler between surgeons. This model also successfully determined IMA tissue characteristics including surface area, width, and exposure area using three-dimensional modeling. Despite variable operative environments, our model consistently quantified steps and approaches during key operative steps. This establishes feasibility of automating quantitative analysis of intraoperative maneuvers. After multicenter validation, this model has the potential to be used for optimizing intraoperative maneuvers, real-time teaching, retrospective review/surgeon credentialing, and improving patient outcomes with applicability to a wide-range of robotic-assisted surgeries.