Fig. 3

Performance of ChatGPT 4.0 in identifying benign/negative, leukoplakia, and squamous cell carcinoma (SCC) cases based on different scenarios: image-only, image and clinical history, and clinical history only; (A) The overall performance of ChatGPT 4.0; (B) The results are further stratified by lesion location (oral cavity [OC] vs. oropharynx [OP]) and (C) tumor stage (early stage [T1/T2] vs. advanced stage [T3/T4]). Each table summarizes the diagnostic outcomes and highlights the impact of clinical history and lesion characteristics on the accuracy of ChatGPT 4.0’s predictions.