Machine Learning and Deep Learning for Knee Osteoarthritis Diagnosis: A Survey on Classification and Severity Grading

Osteoarthritis (OA) is a pervasive degenerative joint disease where there is a gradual breakdown of the cartilage in joints causing pain, stiffness, reduced mobility and swelling. Osteoarthritis is a debilitating condition affecting millions globally and mainly occurs because of the unhealthy and improper diet rising among the urban population. Osteoporosis can have a major impact on the tibiofemoral joint (Weight bearing joint of knee) because of its reduced bone density and increased risk of fractures. These joints are the most commonly occurring pressure points in the entire body as they bear majority of the body weight. These hinge joints are critical for many activities and Osteoarthritis in these areas significantly impact the mobility and movement of the entire body causing pain and stiffness. In extreme cases, requirement of knee replacement surgery which have complex procedures and time for recovery. Planning an appropriate diagnosis and course of treatment for osteoporosis requires an expertise understanding of the degree of joint degeneration. Machine Learning methods and deep learning algorithms can be implemented to enable AI for image recognition through which we can reduce reliance and interpretation by Radiologist and clinical experts. This research provides an extensive literature review on the use of X-ray imaging, GAN models for Synthetic image generation and various deep learning models and ML algorithms implemented to identify and classify the degree of tibiofemoral joint degeneration in osteoporosis. The review summarises various Machine learning models and algorithms previously implemented for Knee Osteoarthritis Detection and classification (including both binary and multi class classification) of osteoporosis on the tibiofemoral joint and also covers the difficulties, developments, and potential paths in this field.