Local topological analysis of densitometer-generated scan images of the proximal femur for differentiation between patients with hip fracture and age-matched controls

Abstract

We evaluate densitometer-generated scan images of the proximal femur with respect to topological properties of bone mineral distribution patterns in selected regions of interest. In a population of 100 post-menopausal women, the method has a highly discriminative potential with a performance superior to standard densitometry. Results vary with anatomical location within the proximal femur. The objectives of the study were to evaluate densitometer-generated scan images of the proximal femur with respect to topological properties of bone mineral distribution patterns in selected regions of interest, to test the ability for differentiation between post-menopausal women hip fracture and controls, and to compare results with standard bone densitometry. We used dual-energy X-ray absorptiometry (DXA) to measure the femoral bone mineral density (BMD) of 100 post-menopausal women (73.4 +/- 12.2), 50 of whom had a recent hip fracture. Local bone mineral distribution in the scanner-generated images was analyzed in the standard DXA-regions of interest (ROIs; femoral neck, the shaft, the trochanteric area; and the total hip) using an optimized, local topological parameter MF2D. Performance of topological analysis and BMD was tested by receiver-operator characteristic and discriminant analysis. Area under the curve (AUC) for correct differentiation between patients with and without fractures by BMD in the different ROIs ranged from 0.64 to 0.71; AUC of regional density-pattern analysis varied between 0.79 and 0.84. Using multivariate statistical models, between 71% and 84% of patients were correctly identified as fracture/non-fracture cases by regional topological analysis, whereas BMD reached levels from 58% to 68%. Our analysis indicates that identification of patients with hip fracture by regional evaluation of density patterns varies with anatomical location within the proximal femur. In our study population, performance of the novel parameter was superior to densitometry.