Robotic Registration of Patient-Specific 3D Elbow Models from Sparse Probe Contacts for Arthroscopic Debridement

Mohammad Reza Pournader, Armin Badre, Mahdi Tavakoli

Keywords: arthroscopic elbow debridement, bone registration, computer-aided orthopedic surgery, Robot-assisted surgery, iterative closest point, gradient-SDF

Abstract: Arthroscopic elbow debridement is challenging due to the endoscopic camera’s limited field of view and insufficient depth perception. To address these limitations, we aimed to develop a 3D visual context to assist surgeons intraoperatively. We align a patient-specific 3D elbow model to the physical scene by performing virtual model-to-patient registration from sparse contact points collected with a robot-held probe, as bone exposure and long-term fiducials are impractical. The pipeline uses PCA+RANSAC for coarse alignment, followed by ICP with a complementary signed-distance-field objective for refinement. On elbow phantoms, this contact-only approach achieved 2.71–3.84 mm RMS error without fiducials or extensive bone exposure, indicating workable alignment under the constraints of the problem.

Submission Number: 59