Oral Presentation ANZBMS-MEPSA-ANZORS 2022

Fracture plates are typically used to secure unstable pelvic ring fractures and facilitate bone healing. Due to the highly varied and often complex nature of such fractures, personalised fracture plates may offer benefits over off-the-shelf fracture plates. However, in order to design personalised fracture plates, the fracture must be virtually reduced, a time-consuming and subjective process involving manual segmentation of the fractured pelvis and repositioning of bone segments. This study compared two methods to reconstruct the hemipelvis from varying amounts of bone geometry as alternatives to manual fracture reduction: statistical shape models (SSMs) and contralateral mirroring. 33 pelvis CT scans were used to train SSMs using a leave-one-out approach. The root-mean-squared error (RMSE) of the SSM reconstruction as well as that of the mirrored hemipelvis was quantified over the entire surface of the hemipelvis and within specific regions. Deviations of pelvic landmarks from their original positions were also computed. The reconstruction of the entire hemipelvis surface based on contralateral mirroring (Figure 1a) produced an RMSE value of 1.21±0.29 mm which was not significantly different to the reconstruction from SSMs using the entire hemipelvis surface (Figure 1b, RMSE value of 1.11±0.29 mm). Landmark deviations generated from contralateral mirroring were found to be significantly lower or statistically equivalent to the deviations generated from SSMs. These results indicate that contralateral mirroring tends to be more accurate for reduction of unilateral pelvis fractures. However, in cases of bilateral pelvic fractures or for pelvic anatomies that are known to be highly asymmetric, SSMs may still be a viable method for rapidly reducing fractures.