Related Experiment Video
Updated: Sep 5, 2026
Pedicle Screw Placement Using an Augmented Reality Head-Mounted Display in a Porcine Model
Published on: May 24, 2024
Development and Validation of a Computer-Assisted Screw Trajectory Planning Model Based on Iterative Closest Point
Abudusalamu Tuoheti1, Gufuding2, Musitapa Mijiti3
1Department of Physician Services, Xinjiang International Medical Center (Xinjiang International Hospital, Xinjiang Academy of Medical Sciences), Urumqi, China.
Abstract:
Study DesignRetrospective study.ObjectivesTo develop and validate a computer-assisted model for planning screw trajectories to support modified cortical bone trajectory (MCBT), cortical bone trajectory (CBT), and pedicle screw (PS) techniques, based on iterative closest point (ICP) registration and weighted k-nearest neighbors (kNN) algorithms.MethodsCT data from 110 patients undergoing lumbar surgery were analyzed, comprising an internal set of 50 younger patients with normal bone density and an external set of 60 patients including younger and older individuals with or without bone loss. L4-L5 segments were reconstructed using Mimics 21.0. Two surgeons manually planned MCBT, CBT, and PS trajectories bilaterally in the internal set to serve as reference standards. A personalized computer-assisted screw planning model was developed using ICP registration and weighted kNN, where the template library consisted of manually generated screw plans. Accuracy was evaluated by comparing algorithm-generated screw trajectories against expert manually planned trajectories to calculate deviations in sagittal inclination (α), axial inclination (β), screw head, pedicle entry point, pedicle crossing point, and screw tip. Hounsfield unit (HU) values along screw trajectories were also measured.ResultsIn the external set (total of 720 screws: 240 PS, 240 CBT, 240 MCBT), deviations for CBT weresagittal inclination (α) 4.117 (2.028, 7.251)°, axial inclination (β) 3.714 (1.901, 6.316)°, screw head 3.382 (2.514, 4.803) mm, pedicle entry point 3.0 (1.999, 4.149) mm, pedicle crossing point 1.511 (1.008, 2.443) mm, screw tip 3.586 (2.465, 4.542) mm. Those for MCBT were 4.865 (2.126, 7.547)°, 3.801 (1.658, 6.147)°, 4.153 (3.14, 5.658) mm, 3.818 (2.43, 6.209) mm, 1.545 (1.061, 2.222) mm, 3.777 (2.768, 5.241) mm, respectively. Computational acceptance rates for PS were 97.06% (HU: 238.4±67.71) in younger normal bone group, 92.71% (173.6±53.83) in older normal bone group, and 97.37% (113.3±58.68) in older bone loss group. For CBT, rates were 100% (477.2±168.6 HU), 100% (338.5±125.8 HU), and 100% (178.1±99.75 HU), respectively. For MCBT, rates were 100% (475.8±131 HU), 98.96% (376.8±104.9 HU), and 97.37% (213.4±106 HU), respectively. MCBT and CBT achieved significantly higher HU values than PS (P < 0.05).ConclusionThe ICP registration and weighted kNN-based planning model demonstrates high computational acceptance rates and excels in planning cortical bone screw trajectories with low breach rates and high HU values.
