Integrating Artificial Intelligence in Periodontal Diagnosis: A Comparative Evaluation of ChatGPT-4 and Dental Educators.

OBJECTIVE

As artificial intelligence (AI) tools gain traction in healthcare, their role in periodontal diagnostics remains largely unexplored. This study evaluated the diagnostic accuracy of ChatGPT-4 in applying the 2017 Classification of Periodontal and Peri-Implant Diseases and Conditions compared to US dental school faculty, with the goal of exploring its potential in dental education and clinical decision-making.

METHODS

Twenty clinical periodontal cases were assessed. Five full-time, non-periodontal dental school faculty members reviewed the 2017 Classification System and completed a Qualtrics-based survey, diagnosing each case by extent, stage, and grade. Diagnoses were compared to reference standards established by two board-certified periodontists. ChatGPT-4 was independently prompted with the same clinical information. Diagnostic accuracy was analyzed using McNemar's test and hierarchical binary logistic regression.

RESULTS

Two faculty members significantly outperformed ChatGPT-4 in overall diagnostic accuracy (p < 0.01), with one also outperforming the model in grading. However, there were no statistically significant differences between the group performance of dentists and ChatGPT-4 across diagnostic categories. The group of dentists achieved an overall accuracy of 75%, while ChatGPT-4 achieved 63.3%.

CONCLUSIONS

ChatGPT-4 demonstrated comparable diagnostic performance to dentists in applying the 2017 periodontal classification system. While not a domain-specific tool, its accessibility and moderate accuracy suggest potential as a supplemental educational and clinical support resource in dentistry.