3D-Scan hole detection for robot-assisted laparoscopic surgery

Minimally invasive laparoscopic surgery, where endoscopes and instruments are inserted through small incisions, has advanced to the next stage of development: robot-assisted minimally invasive surgery. These systems use a camera and robotic manipulators operated by human surgeons through human-in-the-loop control. To further improve surgical precision and autonomy, data-driven assistance must be expanded. One promising approach is 3D reconstruction based on endoscopic images. A 30°endoscope tip enhances the field of view by enabling rotational motion around the instrument’s axis. However, when performing a 3D scan with such an endoscope, a blind spot inherently forms along the shaft axis, creating a region that cannot be captured during rotation. Additional missing data may arise due to occlusions from anatomical geometry and the specific endoscope pose during a scan. These limitations result in incomplete 3D reconstructions, which can negatively impact surgical navigation and decision-making. This paper presents a method tailored to medical applications for detecting and characterizing holes in laparoscopic 3D scans. The proposed method uses geometric analysis of the point cloud to identify regions of sparse or missing data and correlates these gaps with endoscope positioning and anatomical visibility. It is designed to operate robustly on high-density point clouds generated by advanced laparoscopic 3D reconstruction systems. By integrating robotic control, our method provides a foundation for adaptive endoscope repositioning to recover missing views and improve reconstruction completeness. The proposed method paves the way towards fast ( ∼ 5 s) feedback for optimized 3D scanning in laparoscopic environments.