Improving precision in bone tumor resection surgery: automated allograft selection using contralateral reconstruction

In tumor resection surgery, an approach to the reconstruction of the bone anatomy, after removal of the tumor, is represented by allografts. A good match between the donor and recipient is crucial in enhancing integration and minimizing failure rate. However, the recipient bone is often altered by the tumor, making the comparison unreliable. Moreover, evaluating multiple donors is often a complex and time-consuming process. Thus, in this work, an algorithm is proposed that is able to automatically reconstruct the virtual healthy bone anatomy of the recipient and to use this reconstruction to select the most suitable donor. The proposed algorithm first reconstructs the healthy anatomy by aligning the bone to the mirrored contralateral using the Iterative Closest Point (ICP) algorithm. Then, to select the optimal donor, the algorithm queries a digital bone bank by aligning the reconstructed template to donor candidates using a localized search strategy. Finally, L1-Chamfer Distance and a newly devised metric called Surface Distance are computed to evaluate the similarity and assess the optimal donor. The algorithm was validated on a dataset of 49 healthy femurs by creating 980 simulated resections. A retrospective study on two real cases was also conducted. The algorithm achieved a median L1-Chamfer Distance of 0.7217 mm and a Surface Distance of 0.3225 mm for the reconstruction of the healthy anatomy and 0.8621 mm and 0.5338 mm, respectively, for the best donor. The retrospective studies showed how the algorithm was able to outperform the manual procedure in both time and accuracy of the allograft. These case studies highlighted how the algorithm performs efficiently even in real case scenarios, enhancing the shape matching between the recipient and donor bone and thus possibly improving post-operative outcomes.