Inas Karawia, Aya Mahmoud, Rewan Ali, Marie Kamal, Aya Gaber, Mohamed Ashraf Hall
Teledentistry demonstrated higher diagnostic accuracy and agreement with the clinical reference standard than ChatGPT in this preliminary study and may serve as a reliable diagnostic adjunct, particularly in remote settings. However, further refinement and validation of AI-based tools such as ChatGPT are necessary to improve diagnostic performance, with future research focusing on optimizing image quality and expanding datasets to enhance AI reliability.
INTRODUCTION: Dental caries is a common oral health problem globally, necessitating accurate and accessible diagnostic approaches. This study aimed to evaluate and compare the diagnostic accuracy of ChatGPT (AI-based assessment) and teledentistry with the clinical dental examination as the clinical reference standard for detecting dental caries.
METHODOLOGY: A diagnostic accuracy study was conducted on a sample of 40 tooth images, determined through power analysis. Standardized smartphone images were obtained and assessed independently using two approaches: ChatGPT-based analysis and teledentistry evaluation by dental professionals. The findings from both methods were compared to clinical examination results. Diagnostic performance was evaluated using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and area under the curve (AUC). Agreement with the clinical reference standard was assessed using Cohen's kappa coefficient, while the McNemar test was used to assess systematic differences between paired classifications.
RESULTS: Teledentistry showed numerically higher diagnostic performance than ChatGPT across all evaluated parameters. It achieved higher sensitivity (89.5% vs. 78.9%), specificity (85.7% vs. 66.7%), PPV (85% vs. 68.2%), NPV (90% vs. 77.8%), and overall accuracy (87.5% vs. 72.5%). Additionally, teledentistry showed a higher AUC (0.876) compared to ChatGPT (0.728), suggesting better discriminative ability. Both methods showed no statistically significant differences from the clinical reference standard; however, ChatGPT exhibited greater inconsistency.
CONCLUSION: Teledentistry demonstrated higher diagnostic accuracy and agreement with the clinical reference standard than ChatGPT in this preliminary study and may serve as a reliable diagnostic adjunct, particularly in remote settings. However, further refinement and validation of AI-based tools such as ChatGPT are necessary to improve diagnostic performance, with future research focusing on optimizing image quality and expanding datasets to enhance AI reliability.