Achieving the Automation of Knee Joint Functional Scoring Through Based on Improved YOLOX

Abstract Objective This study aims to assess the feasibility of computer model-based evaluation of knee joint functional capacity in comparison to manual assessments. Methods The study consists of two phases: (1) developing the automatic knee joint action recognition and classification systems based on improved YOLOX; (2) analyzing the feasibility between the software systems and doctors, identifying knee joint function of patients, and determining the accuracy of the software systems. Results In this study, the items of knee joint assessment included the stair climbing, walking on uneven surfaces, and knee joint function in the life. The weighted Kappa coefficients between doctors and the software systems are 0.76180, 0.64148, 0.68792, respectively, indicating good consistency of software systems. Conclusion This paper introduces an automatic knee joint action recognition and classification method based on improved YOLOX. Comparing the results between the orthopedic doctors and software system, the effectiveness of this software system is validated.