I read with great interest the recent article titled “Performance of Chat Generative Pretrained Transformer-4.0 in Determining Labiolingual Localization of Maxillary Impacted Canine and Presence of Resorption in Incisors through Panoramic Radiographs” by Salmanpour and Akpınar ( Am J Orthod Dentofacial Orthop 2025;168:220–31). The authors provide timely insight into the limitations of using a general-purpose large language model (LLM) for highly specialized diagnostic tasks in orthodontics. Their work highlights an important and often overlooked distinction between human-like reasoning in text generation and true diagnostic competency in image-based clinical decision-making.
The study’s findings—namely, the model’s low accuracy in localizing impacted maxillary canines (37.1%) and detecting incisor resorption (46.0%)—underscore the need for cautious integration of LLMs into radiographic interpretation workflows. Although ChatGPT-4.0 possesses image-input capabilities, it was not trained as a radiologic diagnostic model, which likely contributes to the observed performance limitations. These results reinforce that LLMs cannot yet replace domain-specific imaging algorithms, particularly convolutional neural networks explicitly designed for radiologic pattern recognition.
Importantly, the authors also demonstrate the value of structured and optimized prompting. Despite the thoughtful use of iterative prompt calibration and multiple-choice formats to reduce hallucinations, the model still failed to achieve clinically acceptable thresholds. This highlights that prompt engineering alone cannot compensate for fundamental training limitations when LLMs are applied outside their intended scope.
The study makes a valuable contribution by setting realistic expectations regarding the current capabilities of LLMs in dental radiology. It also reinforces the need for developing specialty-specific artificial intelligence models, stronger regulatory frameworks, and clearer lines of responsibility as artificial intelligence tools become increasingly accessible to clinicians.
Declaration of AI and AI-Assisted Technologies in the Writing Process
Stay updated, free dental videos. Join our Telegram channel
Leave a Reply
VIDEdental - Online dental courses
