Evaluating Artificial Intelligence in Dental Radiography

Sammanfattning

The integration of Artificial Intelligence (AI) in dental radiography not only presents an opportunity but also holds immense potential to enhance diagnostic accuracy and efficiency. This study addresses the exciting challenge of leveraging AI, specifically a generative pre-trained transformer model, to interpret dental panoramic X-rays, a task traditionally reliant on human expertise. The central purpose of the study is to evaluate the diagnostic capabilities of this AI model compared to professional dental evaluations, focusing on its accuracy and consistency, thereby paving the way for a promising future in dental diagnostics. The research involved a sample of 35 dental panoramic X-rays obtained from Flexident AB, anonymized and annotated by a panel of dental professionals. The study was conducted in two stages: Stage One tested the AI model in three different methods: 1- without any annotations, 2- with numbered teeth, and 3- with colored circles highlighting areas of interest. Stage Two involved training a specialized GPT model with domain-specific knowledge. Key findings indicate that the AI model, when provided with detailed visual annotations, achieved diagnostic accuracy comparable to that of dental professionals, as statistical analysis showed no significant differences between the golden standard (dentist group) and the visually annotated group (P>0.05). However, the model struggled with unannotated images, highlighting the importance of structured input. The research underscores the potential of language-based AI in medical imaging while emphasizing the need for detailed input to optimize performance. This study is pioneering in applying a generative pre-trained transformer model for dental diagnostics, opening new avenues for AI integration in healthcare.