Texture analysis in the assessment of bone structure and prediction of osteoporotic hip fractures

Purpose. The development of algorithms for bone texture analysis in osteoporosis with digital X-ray. Materials and methods. The study involved 132 patients with various levels of the osteoporosis, and osteoporotic hip fractures. Texture analysis of radiographs was performed using the developed computer program. We analyzed correlations between indices of bone mineral density, visual assessment of the data, the fact of hip fracture and the results of the texture analysis. Results. Established the level of convergence of views of experts of various levels of training. Revealed correlations between measures of bone mineral density, visual assessment of the data, the fact of osteoporotic fracture and the results of the texture analysis. Assess the diagnostic efficacy of the mathematical modeling of probability osteoporotic hip fracture on the basis of neural networks. Conclusion: ROC-analysis of the effectiveness of the model based on artificial neural networks (ANN) showed prognostic value of textural changes as additional predictors of osteoporotic fracture of the femoral neck; ANN AUC value reaches 0.92.