Osteoporosis classification using VGG-16 and logistic regression: A radiograph and BMD data approach

Osteoporosis is an ailment associated with the bones, in which the bone resorption takes place at a much faster pace as compared to the formation of bones, eventually leading to the deterioration of bone mineral density (BMD). Ultimately, it adversely affects the strength of bones. To determine different diseases, deep learning is used in almost every sector of healthcare. In the context of Osteoporosis, there are numerous machine learning technologies that have been utilized for early detection of the disease. Certainly, these techniques provided great accuracy, but their scope of study was limited exclusively to individual factors. This paper proposes a model which studies multiple aspects leading to the early prognosis of disease, thus increasing the reliability. The aspects are Bone Density Measure, the X-rays of affected bone, lifestyle of the patient which may include medical history if any, fracture status and the specific bone. The dataset used for the research contains 2000 X-rays in total and 500 BMD reports of 500 distinct patients. in Logistic regression is used for the BMD based classification, where the accuracy achieved is 98.66%, with a recall of 97%, precision of 100% and f1-score of 98% for osteoporotic category. The VGG16 model used for the classification based on image dataset achieves the accuracy of 97.19% which is acceptable comparative to existing methods.