Nevin Manimala Statistics

Generation of hemipelvis surface geometry based on statistical shape modelling and contralateral mirroring

Biomech Model Mechanobiol. 2022 Jun 17. doi: 10.1007/s10237-022-01594-1. Online ahead of print.


Personalised fracture plates manufactured using 3D printing offer an improved treatment option for unstable pelvic ring fractures that may not be adequately secured using off-the-shelf components. To design fracture plates that secure the bone fragments in their pre-fracture positions, the fractures must be reduced virtually using medical imaging-based reconstructions, a time-consuming process involving segmentation and repositioning of fragments until surface congruency is achieved. This study compared statistical shape models (SSMs) and contralateral mirroring as automated methods to reconstruct the hemipelvis using varying amounts of bone surface geometry. The training set for the geometries was obtained from pelvis CT scans of 33 females. The root-mean-squared error (RMSE) was quantified across the entire surface of the hemipelvis and within specific regions, and deviations of pelvic landmarks were computed from their positions in the intact hemipelvis. The reconstruction of the entire hemipelvis surfaced based on contralateral mirroring had an RMSE of 1.21 ± 0.29 mm, whereas for SSMs based on the entire hemipelvis surface, the RMSE was 1.11 ± 0.29 mm, a difference that was not significant (p = 0.32). Moreover, all hemipelvis reconstructions based on the full or partial bone geometries had RMSEs and landmark deviations from contralateral mirroring that were significantly lower (p < 0.05) or statistically equivalent to the SSMs. These results indicate that contralateral mirroring tends to be more accurate than SSMs for reconstructing unilateral pelvic fractures. SSMs may still be a viable method for hemipelvis fracture reconstruction in situations where contralateral geometries are not available, such as bilateral pelvic factures, or for highly asymmetric pelvic anatomies.

PMID:35713823 | DOI:10.1007/s10237-022-01594-1

By Nevin Manimala

Portfolio Website for Nevin Manimala