Trackerless Augmented Reality Navigation for Robotic Partial Nephrectomy using Simultaneous Localization and Mapping

To evaluate the performance of a novel trackerless surgical navigation system for partial nephrectomy using 3D laparoscopy integrated with the simultaneous localization and mapping (SLAM) algorithm.

Patients undergoing robotic partial nephrectomy were retrospectively analyzed. From the preoperative CT/MR images, three-dimensional (3D) models of the kidney, tumor, and vasculature were generated using a deep learning-based approach. A calibrated stereo laparoscope acquired stereo video, which was processed using the SLAM algorithm to reconstruct high-fidelity 3D surface models of the kidney. This reconstruction was registered to the preoperative kidney model using anatomical landmarks and surface-based alignment. Augmented overlays of anatomy, including the tumor, were projected onto the surgical view. Accuracy was assessed using surface target registration error (STRE), landmark registration error (LRE), and localization error. A multi-institutional surgeon survey evaluated the trackerless navigation system using 10 representative cases.

Twenty patients undergoing robotic partial nephrectomy were analyzed. Dense 3D surface reconstruction and laparoscope calibration were successful in all cases. Mean (SD; range) STRE, LRE, and localization errors were 1.47 (0.5; 0.6-2.5) mm, 4.3 (1.6; 1.1-6.5) mm, and 2.6 (1.2; 0.7-5.2) mm, respectively. Nine surgeons highly rated the 3D and 2D overlays (3.99 and 4.41/5). Reported cognitive load was moderate (2.38), with high scores for anatomical accuracy (4.44), trust in the overlay for intraoperative decision-making (3.89), and likelihood of adoption (4.56) on a 5-point scale.

A trackerless SLAM-based navigation system, evaluated retrospectively on recorded operative video, demonstrates the feasibility of accurate augmented visualization during partial nephrectomy.