Deep learning approach for tooth numbering and restoration detection on pediatric periapical radiographs in mixed dentition.

objectivesAccurate tooth numbering and restoration detection on periapical radiographs in mixed dentition are critical to the treatment planning process. They also improve the speed and accuracy of treatment processes by automating the early diagnosis process. This study aims to evaluate the numbering of teeth on periapical radiographs taken during the mixed dentition period and the detection efficiency of dental restorations using the YOLOv8 model. MATERIALS AND

methodsThis study included 1,504 periapical radiographs to number primary And permanent teeth taken for diagnostic purposes for patients between the ages of 6 And 12 And 1,599 periapical radiographs to evaluate the automatic detection efficiency of restorations. Python programming language and YOLOv8 model were used. Model performance success, IoU (Intersection over Union) confusion matrix, accuracy, precision, F1-score, and recall values ​​were used to evaluate the performance of the models.

resultsRecall, precision, F1-score, and accuracy values ​​calculated using the confusion matrix for numbering teeth were determined as 0.915, 0.979, 0.946, And 0.89, respectively. Recall, precision, F1-score, and accuracy values ​​for the detection of restorations were calculated as 0.954, 0.871, 0.911, And 0.83, respectively.

conclusionYOLOv8-based deep learning models offer a promising model for accurate numbering of primary and permanent teeth and detection of restorations in periapical radiographs of pediatric patients with mixed dentition, leading to future studies with high accuracy rates. CLINICAL RELEVANCE: Automating the process of identifying and numbering teeth in dental radiographs, the first stage of clinical evaluation in dentistry, with artificial intelligence-based algorithms saves physicians significant time and labor in diagnosis and treatment planning, increasing the efficiency and accuracy of clinical practices.