Statistical shape modelling of the lumbar spine with reference to gender and principal component analysis

Numerical simulation of spine mechanics relies on anatomical features of the vertebrae. Statistical shape models aid in these simulations by including shape variability and differences between groups, facilitating finite element simulations for medical procedures. This study aims to develop a statistical shape model of the lumbar vertebrae and to determine shape differences between genders. This study used CT images and 3D segmentations from the Vertebrae Segmentation challenge (VerSe) with a total of 138 vertebrae from 46 patients consisting of L3, L4, and L5 vertebrae. From these segmentations, statistical shape models were created, using ShapeWorks software, for the overall and gender-specific means. Moreover, gender differences, principal component analysis and modes of variations were illustrated. Major differences in posterior elements between overall, male, and female means were identified. Regarding the principal component analysis, the first principal component accounted for 39.4% of the variation, which was distributed between the vertebrae in dissimilar patterns. The second principal component, accounting for 20.4% of the variation, was concentrated in L5. In conclusion, this study specifies locations in the lumbar vertebrae with significant differences between genders and identifies modes of variations in the sample using principal component analysis.