Deep Neural Networks for Diagnosis of Osteoporosis: A Review

Abstract

Osteoporosis, a pathological disorder of bones affects millions of individuals worldwide and is the most common disease of bones after arthritis. It is caused due to a decrease in mineral density of bones leading to pain, morbidity, fractures and even mortality in some cases. It is diagnosed with DXA, but its high-cost, low-availability and inconsistent BMD measurements do not make it a promising tool for diagnosis of osteoporosis. The computer-aided diagnosis has improved the diagnostics to a large extent. Deep learning-based artificial neural networks have shown state-of-the-art results in the diagnostic field leading to an accurate diagnosis of the disease. This paper reviews the major neural network architectures used for diagnosis of osteoporosis. We reviewed the neural network architectures based on the questionnaires and the deep neural architectures based on image data implemented for diagnosis of osteoporosis and have summarized the future directions which could help in better diagnosis and prognosis of osteoporosis.