Study of the propensity to suffer an osteoporotic hip fracture based on biomechanical parameters and the automatic selection of the region of interest

Abstract

Osteoporotic hip fractures represent a high social and economic burden. As such, the identification of the fracture risk for a patient is of high interest to clinicians. Femur finite elements 3D models based on advanced DXA imaging allowed us to discriminate between fracture and non-fracture cases. However, the effectiveness of such methods is very sensible to the selection of the analysis region. This project aims to evaluate the power of classification of biomechanical parameters obtained through those models, focused on automatized ROI selection using advanced statistical methods. A cohort of 180 patients, 90 control and 90 fracture, were used, considering an equal balance between men and women. Proximal femur 3D models were obtained from advanced DXA acquisitions, having the same correspondence of nodes and elements. Lateral fall was simulated by the application of a patient-specific force at the top of the femoral head, the trochanter was fixed in the direction of the force and the distal area was fully constrained. Five parameters were evaluated: volumetric bone mineral density (vBMD), maximum principal stress/strain and major principal stress/strain. The critical areas were selected by using a statistical parametric map based on random field theory to evaluate the power of classification of the parameters. The results showed 100% accuracy when predicting trochanteric fractures and a 93.3% for neck fractures. By gender separation, 100% prediction accuracy for the neck fractures. When considering only vBMD, a 100% and 81.7% accuracy on the predictions for trochanter and neck fractures respectively was reached. Although the prediction using the outcomes of the FEM give perfect prediction, this tool is far from the clinics. However, by only considering vBMD, which can be obtained from de DXA based model without the need of lateral fall simulation, good prediction can be obtained. This opens us the possibility to provide a preventive treatment to those patients with a prediction of risk of fracture. This way the social and economic burden of this medical condition would be reduced.