Computational efficient method for assessing the influence of surgical variability on primary stability of a contemporary femoral stem in a cohort of subjects

Finite element (FE) modelling can provide detailed information on implant stability; however, its computational cost limits the possibility of completing large numerical analyses into the effect of surgical variability in a cohort of patients. The aim of this study was to develop an efficient surrogate model for a cohort of patients implanted using a common cementless hip stem. FE models of implanted femora were generated from computed tomography images for 20 femora (11 males, 9 females; 50–80 years; 52–116 kg). An automated pipeline generated FE models for 61 different unique scenarios that span the femur-specific range of implant positions. Peak hip contact and muscle forces for stair climbing were scaled to the donors’ body weight and applied to the models. A cohort-specific surrogate for implant micromotion was constructed from Gaussian process models trained using data from FE simulations representing the median and extreme implant positions for each femur. A convergence study confirmed suitability of the sampling method for cohorts with 10+ femora. The final model was trained using data from the 20 femora. Results showed very good agreement between the FE and the surrogate predictions for a total of 1036 alignment scenarios [root mean squared error (RMSE) < 20 µm; \(R_{\text{validation}}^{2}\) = 0.81]. The total time required for the surrogate model to predict the micromotion range associated with surgical variability was approximately one-eighth of the corresponding full FE analysis. This confirms that the developed model is an accurate yet computationally cheaper alternative to full FE analysis when studying the implant robustness in a cohort of 10+ femora.