Quantification of risk factors for osteoporosis using thresholding technique based on trabecular microarchitectures of the proximal femur

Abstract

The purpose of this study was to investigate how the threshold condition used in a bone density scan influences the morphological parameters and to find an optimal threshold for predicting osteoporosis. Experimental subjects comprised a total of 54 post menopausal women aged over 40 years, who were classified into two groups, 20 normal and 34 osteoporosis patients, according to their T-scores. Bone mineral density was measured on the femoral neck, greater trochanter, and Ward's triangle by dual-energy X-ray absorptiometry. To set the threshold conditions, we employed a newly proposed thresholding method using thresholds ranging from 0 to 95% based on the trabecular bone area. We also selected seven evaluation parameters that composed of structural parameters, skeletonized parameters, and fractal dimension to evaluate the osteoporosis predictability. Based on the experimental results, we found that a threshold condition of 20% yields the most reliable predictability for osteoporosis. Quantitative evaluation also demonstrated that the statistical significances are weakened, as the threshold excessively increases.