Automatic diagnostic system for identifying individuals with osteoporosis on panoramic radiographs

点击翻译按钮获取中文摘要

Recent studies suggest mandibular inferior cortical shape on dental panoramic radiographs may be the useful indicator for the identification of individuals with osteoporosis. However, the diagnostic performances of this index largely depend on the experience of observers because the classification of cortical shape on panoramic radiographs is determined subjectively. In order to diagnose the cortical shape objectively, we developed computer-aided diagnosis (CAD) system that automatically can determine the classification of cortical shape on panoramic radiographs. Two hundred panoramic radiographs of postmenopausal women who had bone mineral density (BMD) assessment at the lumbar spine and the femoral neck by dual energy x-ray absorptiometry were used for this purpose. CAD system was constructed by mathematical morphology. One hundred panoramic radiographs were used for the development of CAD system in order to identify women with low BMD, and another 100 panoramic radiographs were used to estimate the validation of CAD system. Low skeletal BMD was defined as a BMD T score of -1.0 or less. The sensitivity, specificity, predictive values, and accuracy of CAD for the identification of low skeletal BMD were calculated in dichotomous 2 by 2 tables. The likelihood ratio for a positive risk index was also calculated. The sensitivity and specificity were 73.8% (95% CI, 63.1 to 84.5%) and 68.6% (95% CI, 53.2 to 84.0%), respectively, when BMD of the lumbar spine was used as a standard. Those were 66.2% (95% CI, 54.9 to 77.4%) and 56.7% (95% CI, 38.9 to 74.4%), respectively, when BMD of the femoral neck was used as a standard. The likelihood ratio was 2.35 (95% CI, 1.41 to 3.91) at the lumbar spine and 1.53 (95% CI, 0.98 to 2.38) at the femoral neck, respectively. These results suggest CAD system applied for panoramic radiographs may be useful for the identification of low skeletal BMD