Statistical Shape Modeling for Virtual Bone Reconstruction: An Innovative Approach to Allograft Selection

The treatment of malignant bone tumors often involves resection and reconstruction using a prosthesis or donor bone (allograft). Accurate donor-recipient bone matching is crucial, but tumor-related deformities complicate this process. This study explores using a Statistical Shape Model (SSM) to reconstruct the femur's premorbid anatomy for better allograft selection, comparing it with the contralateral mirroring method. The SSM has been developed from 78 healthy femurs obtained after segmentation of post-mortem CT scans. The evaluation of the SSM led to a Generalization of 0.990 mm, a Specificity of 4.622 mm, and a Compactness of 98% with 13 modes of variation. The reconstruction capability of the SSM was assessed on 980 simulated resections. Results showed how the SSM was able to achieve a Chamfer Distance of 1.066 mm and a Hausdorff Distance of 3.229 mm with respect to the original anatomy of the bones. The comparison with the contralateral method showed how this approach offers a promising alternative, particularly when bilateral imaging is not feasible or the contralateral limb is compromised. Furthermore, the retrospective application to a real case of a femur affected by diaphyseal osteosarcoma showed the effectiveness of the proposed SSM in a real case scenario.