Development and validation of statistical models of femur geometry for use with parametric finite element models

Abstract

Statistical models from a previous study that predict male and female femur geometry as functions of age, body mass index (BMI), and femur length were updated as part of an effort to develop lower-extremity finite element models with geometries that are parametric with subject characteristics. The process for updating these models involved extracting femur geometry from clinical CT scans of an additional 8 men and 36 women (previous models used CT scans from 62 men and 36 women for a new total of 70 men and 72 women), using all of the scans for fitting a template finite element femur mesh to the surface geometry of each patient, and then programmatically determining thickness at each nodal location. Principal component analysis was then performed on the thickness and geometry nodal coordinates, and linear regression models were developed to predict principal component scores as functions of age, BMI, and femur length. The results from the updated models were compared to the previous study, and the only improvement was in the R2 value for the female models (0.74 to 0.82). The largest differences between the original models and the previous models occurred in the ends of the femur, where the largest errors in model predictions occurred.