To explore the role of artificial intelligence based augmented reality intraoperative real-time navigation in minimally invasive partial nephrectomy to standardize renal hilum dissection procedures and improve operative efficiency.

A three-dimensional (3D) model by intelligent/interactive qualitative and quantitative analysis was created from patient-specific cross-sectional computed tomography imaging, and the semi-transparent picture were holographically projected on the surgeon console in Da Vinci or monitor in laparoscopy. We developed an artificial intelligence auto-matching technique to generate real-time augmented reality (AR) images by one or two points indicated by the surgeon. The size, rotation, and transparency of the three-dimensional model were manipulated automatically to overlap anatomy in the operative field. We performed laparoscopic nephrectomy or robotic-assisted partial nephrectomy utilizing this technique and evaluated the accuracy of renal vascular localization. We developed a renal hilum difficulty score based on this technique and assigned scores to anatomical relationship between arteries and veins.

A total of 105 patients were finally included in this study, with 46 patients underwent partial nephrectomy with AR real-time navigation. In group AR-navigation, a significantly higher rate of patients had lower renal hilum exposure time. Operation time, WIT, estimated blood loss, and complications rates were comparable in both groups. No differences were found in hospital stay, pathological subtype, and stage. Patients were classified into complex and simple renal hilum score groups. Renal hilum exposure time was significantly shorter in the AR real-time navigation group compared to no navigation group, especially in the complex renal hilum score group.

AI based intraoperative navigation renders a surgical roadmap, provides a standard renal hilum dissecting procedures, and avoid time consuming or unnecessary renal hilum dissection with relatively low complications rates.