Determination of diagnostic accuracy through artificial intelligence applications developed using deep learning method

Aim: This study aims to evaluate the diagnostic performance of an artificial intelligence (AI) model trained with a deep learning algorithm in detecting interproximal, buccal, and occlusal carious lesions in the primary molars of children aged 5 to 12 on panoramic radiographs. The diagnostic results of the AI model were compared with those obtained by three dentists with different levels of clinical experience, based on their assessments of the same panoramic radiographs. Materials and Methods: Panoramic radiographs of children aged 5–12, who received dental treatment at Recep Tayyip Erdoğan University between 2018 and 2024, were retrospectively analyzed. Interproximal, buccal, and occlusal carious lesions in the primary molars were annotated using CranioCatch annotation software (Eskişehir, Türkiye) by three dentists with varying levels of expertise. These labeled datasets were used to train an AI model based on the YOLOv8x algorithm. To compare the diagnostic accuracy of the AI model with the dentists, a set of 100 panoramic radiographs was randomly selected and re-annotated. The ground truth for these images was determined by a pediatric dentist with 10 years of clinical experience. Results: The AI model demonstrated moderate overall performance in caries detection. Specifically, the detection accuracy was moderate for interproximal caries, weak for occlusal caries, and low for buccal caries. Both the AI model and the human examiners achieved the highest performance in detecting interproximal lesions, whereas the poorest results were observed in the identification of buccal caries. Among all evaluators, the experienced dental assistant achieved the highest diagnostic accuracy. Conclusion: Our study highlights the notable success in detecting interproximal caries and the superior performance of the experienced assistant. The findings suggest that AI models can enhance the clinical accuracy and efficiency of diagnostic systems, providing a strong basis for future research in this field. Key words: Artificial intelligence, Deep learning, Dental caries, Panoramic radiography, YOLOv8x