Abstract
Summary
A retrospective study on 98 patients shows that FE-based bone strength from CT data (using validated FE models) is a suitable candidate to discriminate fractured versus controls within a clinical cohort.
Introduction
Subject-specific Finite element models (FEM) from CT data are a promising tool to non-invasively assess the bone strength and the risk of fracture of bones in vivo in individual patients. The current clinical indicators, based on the epidemiological models like the FRAX tool, give limitation estimation of the risk of femoral neck fracture and they do not account for the mechanical determinants of the fracture. Aim of the present study is to prove the better predictive accuracy of individualised computer models based a CT-FEM protocol, with the accuracy of a widely used standard of care, the FRAX risk indicator.
Patients and Methods
This retrospective cohort is individually-matched case control study composed by 98 Caucasian women who were at least 5 years post menopause. The case group consisted of 49 patients who had sustained a hip fracture (36 intra-capsular and 13 extra-capsular fractures) within the previous 90 days due to low-energy trauma. The CT datasets were segmented (using the ITK-Snap software) in order to extract the periosteal bone surface. Unstructured meshes (10-node tetrahedral elements) were generated using ANSYS mesh morphing software. Each CT dataset was calibrated using the European Spine Phantom. The inhomogeneous material properties were mapped from CT datasets into the FEM with the BoneMat_V3 software. Bone strength was evaluated in quasi-axial loading conditions, for a set of 12 different configurations sampling the cone of recorded in vivo hip joint reactions, and was defined as the minimum load inducing on the femoral neck surface an elastic principal strain value greater than a limit value.
Results
There were no statistically significant difference between the fracture and the control groups for age, height and weight (p<0.05). All indices of areal bone mineral density (aBMD) and the volumetric mineral density (vBMD) between fractured and controls showed on average a lower value for fractured respect of the controls, with similar mean difference (14% for aBMD and 13% for the vBMD). FEM-predicted strength differed between fractured and non-fractured on average for 20%. To evaluate its ability to identify patients at risk of hip fracture, FEM-based strength was compared to the FRAX predictor by computing for each predictor the Receiver Operating Characteristic (ROC) curve, and the Area Under the Curve (AUC). The individualised risk predictor based on FEM bone strength was found to perform significantly better (AUC=0.76) than FRAX (AUC=0.66). When the FEM-based strength indicator was combined with available clinical information in a logistic regression, the resulting predictor achieved in this retrospective study an excellent accuracy (AUC=0.82).
Discussion
This study confirms that individualised, CT- FEM, when generated using to the state-of-the-art protocols, can provide a predictor of the risk of hip fracture more accurate than those based on clinical data alone. In the integrated workflow developed in the VPHOP Project (FP7-ICT-223865) CT-based risk prediction is requested only for those patients for whom the clinical decision is uncertain.
