Comparison of diagnosis and treatment planning based on clinical and radiographic findings in traumatized teeth between pedodontists and GPT-4.1 mini: An evaluation in light of AAPD Guidelines

Aim

The aim of this study was to compare the clinical performance of pedodontists with that of the GPT-4.1 mini model, operating on the ChatGPT platform developed by OpenAI, in the diagnosis and treatment planning of traumatic dental injuries in pediatric patients.

Materials And Methods

This study was conducted using clinical and radiographic records of 18 pediatric patients aged 8–14 years who presented to the clinic with dental trauma, completed treatment successfully, and were followed for at least 12 months. Excluding avulsion cases, a total of 19 maxillary central incisors with different types of trauma were evaluated. Clinical findings and radiographic images for each tooth were digitized. The questionnaire used in this study was developed by expert pedodontists based on trauma types, diagnostic criteria, and treatment approaches outlined in the current American Academy of Pediatric Dentistry (AAPD) trauma guidelines, as well as the clinical procedures routinely applied in practice. Accordingly, a total of 36 multiple-choice questions were created (18 related to diagnosis and 18 to treatment planning). A pilot study involving five pediatric dentists was conducted to assess questionnaire clarity and administration time, and minor revisions were made accordingly. The prepared questions were individually submitted to GPT-4.1 mini, and the same questions were distributed to 17 pedodontists via Google Forms. The most recent AAPD dental trauma guideline (2020) was accepted as the gold standard for diagnosis and treatment comparisons. Statistical analysis was performed using IBM SPSS Statistics version 23. The level of agreement between GPT-4.1 mini and pedodontists was assessed using Cohen’s Kappa statistic, and 95% confidence intervals were calculated using the asymptotic (Wald) method. Categorical data were presented as frequencies and percentages, and the level of statistical significance was set at p < 0.05.

Results

A substantial level of agreement was observed between GPT-4.1 mini and pedodontists in diagnostic decisions (Kappa = 0.75; p < 0.001). Diagnostic agreement ranged from 50% in extrusion cases to 100% in intrusion, root fracture, lateral luxation, enamel fracture, and enamel-dentin fracture cases. Agreement for complicated crown-root fractures was 67%. Similarly, a substantial agreement was found between GPT-4.1 mini and pedodontists in treatment planning decisions (Kappa = 0.75; p < 0.001).

Conclusion

In cases of traumatic dental injuries, when diagnosis and treatment processes are evaluated in accordance with current guidelines and individualized approaches, GPT-4.1 mini demonstrates a considerable level of agreement with pedodontists.