The prevalence of Temporomandibular Joint Disorder (TMD) is currently substantial, with symptoms encompassing auditory and pain sensations in the vicinity of the temporomandibular joint. If left untreated, the condition will progressively worse, leading to the manifestation of joint restrictions and restricted mouth opening. The diagnostic rate of TMD patients remains low due to constraints imposed by the limited availability of clinical technology. While AI-assisted medical interventions have proven beneficial in diagnosing various ailments, there remains a dearth of research specifically addressing the diagnosis of TMD. The objective of this study is to utilize radiomics technology in order to create and evaluate an artificial intelligence analysis model for temporomandibular joint disease. This will be achieved by conducting feature extraction and selection on CBCT images of the temporomandibular joint (TMJ) to identify features that exhibit substantial disparities in distinguishing normal and abnormal TMJ attributes. Consequently, a machine learning model will be trained to accurately identify TMD. This methodology has demonstrated enhanced accuracy and efficiency in obtaining diagnostic outcomes, thereby enhancing the detection rate of TMD. The diagnostic performance of five machine learning models, namely Support Vector Machine (SVM), Random Forest (RF), XGBoost, AdaBoost, and GBDT, was evaluated. Among these models, GBDT achieved the most favorable results, with a diagnostic accuracy of 0.985, an AUC of 0.927, a sensitivity of 0.969, and a specificity of 1.0. These findings underscore the efficacy of employing machine learning techniques to enhance the TMJ model.