A comprehensive computational pipeline for foramen magnum shape analysis integrating elliptic fourier transform, polygon approximation, and procrustes heatmaps from CT images: Application to forensic sex estimation

Aim: The foramen magnum (FM) shows size-based sexual dimorphism, but whether its shape differs independently between sexes is unresolved, as previous studies relied on subjective classification. This study applied computational geometry to computed tomography (CT)-derived FM contours to objectively quantify shape variation and evaluate its value for sex estimation.

Materials and Methods: FM boundaries were automatically extracted from axial CT images of 473 adults (234 males, 239 females) from an Eastern Turkish population and screened by an automated topological quality check. Seven geometric analyses, including normalized elliptic Fourier transform and Procrustes-based heatmaps, were applied; six feature-based analyses yielded 209 descriptors per specimen, classified a priori as size measures (n = 30), scale-dependent contour descriptors (n = 63), or scale-invariant shape descriptors (n = 116). Sex differences were assessed with Mann–Whitney U tests (Bonferroni-corrected) and Cohen's d. Classification used linear discriminant analysis (LDA), support vector machines, and random forest with stratified 10-fold cross-validation and nested forward feature selection.

Results: All 30 size descriptors differed significantly between sexes (Cohen's d = 0.71–1.33), whereas only 1 of 116 scale-invariant shape descriptors did (boundary roughness, d = 0.62). Size alone yielded an LDA accuracy of 77.1% (95% confidence interval 73.2 – 80.7%). Scale-invariant shape descriptors reached 64.0% (58.3% after nested forward selection) and Procrustes heatmaps 59.9–60.9%; adding forward-selected shape descriptors to size did not improve classification (74.6%; McNemar p = .08). A six-sided polygon was optimal in both sexes, and the Bayesian information criterion favoured a single mixture component, indicating no discrete within-sex subtypes.

Conclusion: FM sexual dimorphism is almost entirely size-mediated. Once scale is removed, the FM outline carries only weak sex information that does not add to size-based estimation. The automated, openly available pipeline replaces subjective shape categorization with objective, continuous measurements and may be transferable to other skeletal structures.