A statistical shape and density model can accurately predict bone morphology and regional femoral bone mineral density variation in children.

Abstract

Finite element analysis (FEA) is a widely used tool to predict bone biomechanics in orthopaedics for prevention, treatment, and implant design. Subject-specific FEA models are more accurate than generic adult-scaled models, especially for a paediatric population, due to significant differences in bone geometry and bone mineral density. However, creating these models can be time-consuming, costly and requires medical imaging. To address these limitations, population-based models have been successful in characterizing bone shape and density variation in adults. However, children are not small adults and need their own population-based model to generate accurate and accessible musculoskeletal geometry and bone mineral density in a paediatric population. Therefore, this study aimed to create a biomechanical research tool to predict the personalized shape and density of the paediatric femur using a statistical shape and density model for a population of children aged from 4 to 18 years old. Femur morphology and bone mineral density were extracted from 330 CT scans of children. Variations in shape and density were captured using Principal Component Analysis (PCA). Principal components were correlated to demographic and linear bone measurements to create a predictive statistical shape-density model, which was used to predict femoral shape and density. A leave-one-out analysis showed that the shape-density model can predict the femur geometry with a root mean square error (RMSE) of 1.78 ± 0.46 mm and the bone mineral density with a normalized RMSE ranging from 8.9 % to 13.5 % across various femoral regions. These results underscore the model's potential to reflect real-world physiological variations in the paediatric femur. This statistical shape and density model has the potential for clinical application in rapidly generating personalized computational models using partial or no medical imaging data.