Image-guided surgery in the treatment of rectal cancer: the impact of virtual and augmented reality in clinical practice.

Aim: To enhance minimally invasive surgery for rectal cancer using Augmented Reality (AR) to overlay preoperative 3D models onto intraoperative endoscopic images. Methods: Magnetic resonance neurography and computer tomography images were segmented to create a 3D virtual pelvic model, including rectum, nerves, vessels, and ureters. The pelvic structures were reconstructed as separate units, with the option to make them transparent. To address the persistent challenge of real-time instrument occlusion during intraoperative AR overlays, a binary segmentation ConvNext U-Net architecture was utilized. Initially, the neural network was pre-trained on a publicly available dataset of labelled laparoscopic surgical procedures. Subsequently, it was fine-tuned using an in-house dataset, specifically collected and labelled to enhance robustness during live surgeries. By identifying the pixel corresponding to the surgical instruments, the overlay was masked, ensuring that the instruments remained visible above the virtual model during the intraoperative phase. The abdominal aorta at the level of the bifurcation of the iliac arteries and the right iliac artery which intersects the ureter were used as anchor points to accurately superimpose the 3D model onto the real-time surgical image. Application of AR to retroperitoneal structures were preferred because they are fixed and less susceptible to deformation due to pneumoperitoneum. The critical and high-risk points of anterior rectum resection were identified to tailor the use of AR technology and restrict its application to the specific surgical steps where it is most needed. Results: The AR-assisted procedure was performed successfully in the operating room on two cases of laparoscopic rectal resection. During the AR-assisted procedures, the surgeon's ability to identify key pelvic anatomical structures, such as vessels, ureters, and nerves, was assessed. Conclusion: The study confirms the feasibility of using AR during rectal surgery and suggests that it could improve surgical safety by improving the identification of anatomical structures.