Exploring inter-subject anatomic variability using a population of patient-specific femurs and a statistical shape and intensity model.

This paper is motivated by the need to accurately and efficiently measure key periosteal and endosteal parameters of the femur, known to critically influence hip biomechanics following arthroplasty. The proposed approach uses statistical shape and intensity models (SSIMs) to represent the variability across a wide range of patients, in terms of femoral shape and bone density. The approach feasibility is demonstrated by using a training dataset of computer tomography scans from British subjects aged 25-106 years (75 male and 34 female). For each gender, a thousand new virtual femur geometries were generated using a subset of principal components required to capture 95% of the variance in both female and male training datasets. Significant differences were found in basic anatomic parameters between females and males: anteversion, CCD angle, femur and neck lengths, head offsets and radius, cortical thickness, densities in both Gruen and neck zones. The measured anteversion for female subjects was found to be twice as high as that for male subjects: 13 ± 6.4° vs. 6.3 ± 7.8° using the training datasets compared to 12.96 ± 6.68 vs. 5.83 ± 9.2 using the thousand virtual femurs. No significant differences were found in canal flare indexes. The proposed methodology is a valuable tool for automatically generating a large specific population of femurs, targeting specific patients, supporting implant design and femoral reconstructive surgery.