Predicting osteoporosis by analyzing fracture risk factors and trabecular microarchitectures of the proximal femur from DXA images

Abstract

This study aimed to identify the optimal threshold ranges for predicting osteoporosis and osteoporotic fractures by analyzing the correlations between trabecular patterns and fracture risk factors. We selected 85 post-menopausal women as experimental subjects and classified them into 29 normal and 56 osteoporotic groups. We proposed a novel thresholding algorithm that divides the threshold ranges from 0 to 95% based on trabecular bone area and assessed osteoporosis predictability for each range. Evaluation parameters were categorized into morphological parameters (Tb.Area, Sk.Length and fractal dimension) and fracture risk factors (MSCT, LSCT, FNW, TW, FNAL and HAL). Consequently, we found the clinical usefulness of our algorithm for discriminating patients with osteoporosis from those with normal bone. The significances between the morphological parameters and the fracture risk factors improved as bone mineral density decreased. Based on these results, we selected the optimal threshold conditions for predicting osteoporosis and osteoporotic fractures at thresholds of 40-80%. © 2014 Springer-Verlag Berlin Heidelberg.