Evaluating ChatGPT's Diagnostic Accuracy in Oral Mucosal Lesions: A Comparative Study with a Maxillofacial Surgeon

Objective: Artificial intelligence (AI) and profound learning algorithms have been increasingly used for computerized decision-making in various complex tasks in recent years. This study aimed to compare ChatGPT (OpenAI, San Francisco, California, U.S.) with a maxillofacial surgeon to diagnose and find differential diagnoses of oral mucosal lesions and evaluate their usefulness. Material and Methods: A maxillofacial surgeon with five years of experience and ChatGPT answered questions about twenty-three oral mucosal lesions. The lesion diagnosis is labeled as diagnosed and incapable of providing a diagnosis, and one point is awarded for each accurate differential diagnosis. Results: While the clinician correctly diagnosed all twenty-three oral mucosal lesions included in the study, ChatGPT correctly diagnosed nineteen, and there was no statistically significant difference (P = 0.109). When the differential diagnosis results of the clinician and ChatGPT were compared, no statistically significant difference was found (P = 0.500). Conclusion: Our study showed that a maxillofacial surgeon with five years of experience and ChatGPT showed similar results in the diagnosis and differential diagnosis of oral mucosal lesions. It will be speculated that ChatGPT can act as a new tool that provides information for patients with oral mucosal lesions. Hence, it possesses the capacity to function as a supplementary apparatus, thereby mitigating the workload encountered within the healthcare domain and enabling patients to reach preliminary evaluation from home.