Temporomandibular joint (TMJ) dysfunction is a condition that can lead to inflammation, chronic pain, and difficulty performing daily activities. It affects the joint connecting the jawbone to the skull, which enables essential movements such as speaking, chewing, and yawning. TMJ disorders (TMD) may arise from factors including tooth wear, improper bite, arthritis, trauma, teeth grinding, and stress, and are commonly associated with symptoms such as earaches, headaches, jaw tenderness, and facial pain. Traditionally, clinicians rely heavily on cone-beam computed tomography (CBCT) reports, complemented by physical examinations and dental assessments, to identify and assess TMJ dysfunction. The severity of the disorder is determined by examining CBCT images of the joint structures alongside clinical symptoms. This research aims to leverage patient datasets to develop a deep-learning model capable of classifying the degree of TMJ dysfunction. Among the models evaluated, DenseNet201 demonstrated the highest performance in predicting severity. The classification workflow is automated through a graphical user interface (GUI), providing a transparent system that supports healthcare professionals and researchers in detecting, analyzing, and understanding TMJ cases, thereby strengthening the link between medical diagnosis and artificial intelligence.