The Effectiveness of ChatGPT in Temporomandibular Joint Pathologies and Sequence Determination in Magnetic Resonance Imaging

Objectives: Temporomandibular joint (TMJ) pathologies is a general term for pain and dysfunction of the TMJ complex and masticatory muscles. Magnetic resonance imaging (MRI) is the gold standard for evaluating the TMJ complex, disc-condylar relationship and disc displacement. The aim of this study was to evaluate the effectiveness of ChatGPT version 4.0 (ChatGPT-V4) in identifying TMJ pathologies, slice plane and sequence in MR images. Materials and Methods: One hundred MR images of patients with bilateral TMJ pathology (200 TMJs) were recorded. TMJ pathologies, slice plane, sequence of the MR images and ChatGPT responses were evaluated. The subheadings in the ChatGPT-V4 answers were recorded as true if present in the image and false if not. The slice plane was categorized as true-false and the sequence as true-false-missing. Results: In all images, ChatGPT-V4 correctly recognized the slice plane. The correct MR sequence recognition rate was 81.5%. Missing sequence recognition rate was 12% on both sides. It failed to identify the fat-suppressed sequences. ChatGPT-V4 misidentified the sequence in 6.5% of the images. The accuracy of ChatGPT-V4 in identifying TMJ pathologies remained at 50.7%. ChatGPT-V4 achieved the highest accuracy in the diagnosis of anterior disc dislocation and osteoarthritis. Conclusions: ChatGPT-V4 can be relied upon to control the slice plane and sequence of MR images. The results show that ChatGPT-V4 is currently limited in its ability to produce responses related to TMJ pathologies on MR images. Therefore, ChatGPT cannot replace a dentist and physicians should be aware of this limitation during their use of chatGPT to check for TMJ pathologies.