Abstract:
Femur shape and cortical thickness are vital in accurate models of lower limb function and device performance. However, in any population, femur shape and cortical thickness can vary greatly with respect to demographic and anthropometric variables. We show that a linear partial least-squares regression model can capture these variations and accurately generate femoral shape and cortical thickness distribution given demographic and anthropometric inputs. In K-fold cross validation experiment on 204 femora, average shape prediction error was 2.3 mm RMS, and 0.5 mm RMS for cortical thickness. This method allows the efficient generation of population-specific femur geometry for downstream modeling.
