An approach for osteoporosis diagnosis based on support vector machine using micro-CT images

Introduction: The early diagnosis of osteoporosis is of great meaning for reducing osteoporotic fracture risk. Non-invasive osteoporosis diagnosis based on image analysis has been focused on in recent years. Materials and Methods: In this paper, we propose a new approach for osteoporosis diagnosis based on support vector machine (SVM) using 3D micro-CT images. This approach combines the parameters related to osteoporosis and obtains an optimal classifier for distinguishing osteoporosis and normal groups. Bone mineral density (BMD) and four parameters relevant to trabecular bone (TB) structure have been used as features for osteoporosis recognition, which are obtained from micro-CT image data through Volumetric Topological Analysis (VTA). Combining the five parameters, we utilize SVM method to train these data due to its excellent performance on classification. 50 micro-CT image data are used in our experiment, in which, 25 are from osteoporosis group and 25 are from normal group. Results: Precision, recall and F-measure are used to evaluate the performance of the classifier obtained, which are 90.9%, 100% and 95.2% in the experiment. Discussion: The good performance demonstrates that SVM method performed on the five parameters can effectively distinguish osteoporosis and normal groups. It is convinced that SVM is a promising method for osteoporosis diagnosis.