Osteoporosis Identification Based on Computed Tomography Scan Image and Machine Learning

Osteoporosis, characterized by a reduction in bone density, is a common condition among the elderly, leading to increased fracture risks. Early detection is critical for effective medical intervention to prevent severe complications. This study explores the viability of using machine learning-based technologies for detecting osteoporosis through computerized tomography (CT) scan images and enhanced image attributes. The machine learning model was trained on a dataset of 520 CT scan images from patients with normal and osteoporotic bone conditions. Novel image attributes – phase, contrast, roughness, and grayscale – were derived from the original CT scan images. These attributes were tested in multiple input scenarios (single, double, and multi-attribute) to assess their contribution to the accuracy of the model. The results demonstrated that incorporating these image attributes into the machine learning model significantly enhanced the detection accuracy of osteoporosis, showcasing the potential of this method for automated, non-invasive diagnosis. Unlike conventional methods, this approach introduces a novel set of image attributes for bone quality evaluation, which improves the prediction of osteoporosis in CT scan images and reduces false negatives. However, further validation on a larger dataset is required before clinical application.