Automated virtual reduction of pseudo-fractured mandibles using sex-specific statistical shape models: a proof-of-concept study.

PURPOSE: This proof-of-concept study developed an automated virtual mandibular fracture reduction framework using statistical shape models (SSMs), evaluated it in ten controlled pseudo-fracture cases, and compared sex-specific and combined-sex SSMs.

METHODS

Male, female, and combined-sex SSMs were constructed from CT-derived mandibular meshes using the same correspondence estimation, atlas updating, and principal component analysis pipeline. Ten independent mandibles (five male and five female) were divided into three fragments and displaced. Reduction jointly optimized SSM coefficients and fragment transformations using the covariance matrix adaptation evolution strategy. Paired results were compared using exact two-sided Wilcoxon signed-rank tests.

RESULTS

The male, female, and combined-sex SSMs required 28, 32, and 29 principal components, respectively, to reach 95% cumulative explained variance. Using the sex-specific SSMs, the mean symmetric chamfer distance and symmetric root-mean-square error were 0.712 ± 0.210 mm and 0.855 ± 0.231 mm, respectively, compared with 0.778 ± 0.309 mm and 0.937 ± 0.358 mm using the combined-sex SSM. Although the sex-specific SSMs yielded numerically lower mean errors, the differences were not statistically significant (p = 0.232 and p = 0.193, respectively). Computation time was approximately 23 min per reduction run.

CONCLUSION

The framework achieved near-millimetric accuracy, supporting the technical feasibility of automated virtual mandibular reduction under controlled conditions. Sex-specific SSMs yielded numerically lower errors than the combined-sex SSM, without a statistically significant advantage. Larger validation cohorts and real-fracture studies are required before clinical application.