Registration of 2D X-ray and 3D CT Images for Robot-Assisted Pelvic Fracture Reduction Surgery

Abstract

Minimally invasive pelvic fracture reduction surgery (PFRS) is one of the most challenging procedures in trauma orthopedics, in which accurate navigation is pivotal to surgical success. This paper introduces a two-stage 2D-3D registration algorithm, encompassing coarse registration and multi-view refinement module, to accurately predict the fragment's pose. Based on the 2D-3D registration algorithm, a navigation system for real-time fragment tracking is established and deployed in a master-slave pelvic fracture reduction robot system. Validation on a publicly available dataset demonstrates the algorithm achieves an average registration accuracy of 0.56 mm, with a processing time of 4.6 s, and a navigation error of 1.47 mm on a pelvic model. In vitro simulated experiments yield an average reduction accuracy of 2.58 mm and 2.33°, rated as Excellent according to clinically used criteria. These findings substantiate the superiority of the proposed algorithm and navigation system, providing intraoperative guidance for pelvic fracture reduction surgery.