Developing an algorithm for virtual fracture reduction (VFR) of mandibular fractures using CT scans: a preliminary technical study

Aim: Mandibular fractures are common and challenging to reduce accurately, particularly in multi-fragment or edentulous settings. Virtual fracture reduction (VFR) could standardize preoperative planning and provide millimetric targets for open reduction and internal fixation but remains limited by manual workflows and variable imaging quality. Methods: We developed a two-stage VFR algorithm using computed tomography (CT)-derived 3D models. Coarse alignment was achieved via a genetic optimization algorithm, followed by fine alignment with the Iterative Closest Point (ICP) algorithm. A weighted objective function optimized (i) condylar positioning, (ii) dental occlusion, and (iii) fracture gap minimization. We tuned weights over 14 experiments and optimized 6 pose parameters per segment with a Genetic Algorithm. Accuracy was evaluated on a synthetic fracture model and applied to three clinical cases. Outcomes included surface-distance maps, occlusal landmark deviation, interfragmentary gaps, and overlap volume and direct contact area. Results: On the synthetic model, maximum surface deviation was 1.58 mm, with fracture-surface errors predominantly < 1 mm. Occlusal reconstruction reproduced molar relationships within 0.07-0.15 mm of the pre-fracture model. In three clinical cases (simple and comminuted patterns), mean interfragmentary gaps were 0.22-0.70 mm, overlap volumes ranged from 28.54 to 119.26 mm3, and direct contact areas spanned 216.00-533.17 mm2, indicating minimal interpenetration and substantial bone-to-bone fit. The pipeline demonstrated stable convergence and reproducible alignment without intraoperative data. Conclusion: These preliminary technical findings support the proposed VFR algorithm as a practical proof-of-concept adjunct to trauma planning. Larger multicenter clinical validations with statistical analyses and functional endpoints are warranted to validate its clinical impact and scalability.