Determination of osteoporosis using deep learning methods

Osteoporosis is the most common chronic bone disease, which is characterized by low bone mineral density. Dual Energy X-Ray Absorptiometry (DEXA) scan is the most used method for measuring bone mineral density and diagnosing osteoporosis. Unfortunately, this method has certain limitations, such as the size of the device and it's high cost. Other screening methods like standard X-rays and computed tomography (CT) can't detect osteoporosis until it's fully accrued. In this study, a non-invasive method for osteoporosis classification using X-ray images (plain radiographs) of the heel is proposed. Convolutional Neural Networks along with Data Augmentation techniques and Transfer Learning Architectures are combined to classify X-ray images of healthy and osteoporotic patients. With the proposed approach, diagnosis of osteoporosis has been achieved with high accuracy.