This innovative study proposes a revolutionary strategy to enhance the efficacy and customization of hip arthroplasty, focusing on a short prosthesis design meticulously tailored to individual patient needs, with special emphasis on the femoral stem. The implemented methodology involves data collection through a meshless software, which quantifies the stress/strain shielding resulting from femoral insertion. Automatic pre-processing is performed using clustering algorithms, allowing the data to be divided into anatomical areas of the body. Subsequently, training of machine learning algorithms, such as random forests and stacking, is carried out to accurately predict the shielding caused by the stem. These predictions are based on design parameters, specifically linked to the patient's femoral cavity dimensions and their relationship to stem size, which are quantified through dimensionless parameters and serve as model inputs. The reliability of the models is confirmed by evaluation of data from patient with different femoral morphology than those used during the training and validation phases. This approach not only provides fast and cost-effective analyses for medical experts and engineers, but the described methodology can also be successfully extended to the development of alternative arthroplasty prostheses, such as shoulder and knee implants.
Read full abstract