Prediction of Vertebral Fracture Risk Using Support Vector Machine Algorithm

Since osteoporosis is a widespread disease characterized by low bone density and bone tissue structural deterioration, early diagnosis of this pathology is then essential. The purpose of this work is to provide a support vector machine (SVM) multi descriptor model for fracture risk prediction based on CT Scan and Dual energy X-ray absorptiometry (DXA) input of cadaveric lumbar vertebrae. More specifically, the aim is to test the SVM learning by combining biomechanical characteristics such as “fracture stresses” and microarchitectural features with Bone Mineral Density (BMD). The contribution of microarchitectural parameters allowed the model to better capture essential aspects of bone quality overlooked by conventional techniques like DXA. The experimental results show that the proposed method can detect vertebral fracture risk with high accuracy.