Reverse Engineering the Fracture: Three-Dimensional Fragment Reconstruction and Reduction Planning in a Synthetic Tibial Plateau Fracture Phantom

Background Restoration of pre-injury anatomy is central to the surgical management of comminuted intra-articular fractures, yet current computed tomography (CT)-based and 3D-printed planning tools mainly visualize fragment anatomy rather than computationally determine how fragments should be reduced. Objective To develop and pilot-test, in a synthetic, in-silico phantom, a computational framework that reverse-engineers pre-fracture anatomy from segmented fracture fragments and generates a quantitative, case-specific virtual reduction plan. Methods We propose a seven-stage pipeline centred on a sequential Iterative Closest Point (ICP) registration algorithm, in which fragments are reduced against a template representing the mirrored contralateral anatomy in decreasing order of size, so that already-reduced fragments constrain registration of subsequent, smaller fragments via fracture-surface contact. As no public dataset of segmented fracture fragments with known pre-injury ground-truth geometry exists, the algorithm was validated on a synthetic geometric phantom of the proximal tibia, computationally fractured into 3–5 fragments under mild, moderate, and severe displacement conditions (n = 15 cases per severity; 45 total), with known ground-truth transforms. Reduction quality was quantified using root-mean-square (RMS) surface distance, rotational error, and maximum articular step-off, referenced against a 2 mm threshold associated with post-traumatic osteoarthritis risk in the clinical literature. Results Across all 45 synthetic cases, virtual reduction decreased mean RMS surface distance from 3.85 mm to 0.97 mm and mean maximum articular step-off from 9.55 mm to 3.73 mm. The proportion of cases achieving the clinically-referenced ≤2 mm articular step-off threshold increased from 0% pre-reduction to 53.3% (mild), 20.0% (moderate), and 0% (severe) post-reduction (24.4% overall), with mean rotational error of 7.4°, 7.2°, and 24.6° respectively. Conclusion Sequential, contralateral-template-anchored virtual reduction is technically feasible and substantially improves fragment alignment on a synthetic fracture phantom, with performance degrading predictably as displacement severity increases. Fragment-segmentation integration with real CT data and cadaveric or clinical validation are necessary before clinical translation. No real patient CT, cadaveric, or clinical data were used in this pilot study, and no claim of clinical readiness is made or implied.