Osteoarthritis (OA), particularly knee osteoarthritis, is the most prevalent form of arthritis, resulting in severe dis-ability for sufferers throughout the world. A Manual diagnosis, segmentation, and annotating joints of knee continue to be the most used procedure for diagnosing osteoarthritis (OA) in clinical settings, despite being laborious and highly susceptible to user variation. Several prediction models displayed prognostic ability in ways of predicting the possible onset of OA, the potential aggravation of OA, the prospective progression of pain and structural deterioration as well as the potential occurrence of total knee replacement (TKR). Apart from research gaps, techniques of machine learning continue to demonstrate enormous potential for challenging tasks e.g., initial knee OA detection and recognition of further disease events, also basicthings such as identifying innovative imaging features and establishing a novel measure of OA status. Future OA treatment discoveries may be aided by the continuous improvement of machine learning models.