Sensitivity of femoral strain calculations to anatomical scaling errors in musculoskeletal models of movement

The determination of femoral strain in post-menopausal women is important for studying bone fragility. Femoral strain can be calculated using a reference musculoskeletal model scaled to participant anatomies (referred to as scaled-generic) combined with finite-element models. However, anthropometric errors committed while scaling affect the calculation of femoral strains. We assessed the sensitivity of femoral strain calculations to scaled-generic anthropometric errors. We obtained CT images of the pelves and femora of 10 healthy post-menopausal women and collected gait data from each participant during six weight-bearing tasks. Scaled-generic musculoskeletal models were generated using skin-mounted marker distances. Image-based models were created by modifying the scaled-generic models using muscle and joint parameters obtained from the CT data. Scaled-generic and image-based muscle and hip joint forces were determined by optimisation. A finite-element model of each femur was generated from the CT images, and both image-based and scaled-generic principal strains were computed in 32 regions throughout the femur. The intra-participant regional RMS error increased from 380με (R²=0.92, p<0.001) to 4064με (R²=0.48, p<0.001), representing 5.2% and 55.6% of the tensile yield strain in bone, respectively. The peak strain difference increased from 2821με in the proximal region to 34,166με at the distal end of the femur. The inter-participant RMS error throughout the 32 femoral regions was 430με (R²=0.95, p<0.001), representing 5.9% of bone tensile yield strain. We conclude that scaled-generic models can be used for determining cohort-based averages of femoral strain whereas image-based models are better suited for calculating participant-specific strains throughout the femur.