Comparative analysis of two different deep convolutional neural network models in the detection and anatomical classification of mandibular fractures

Mandibular fractures are among the most common conditions encountered in maxillofacial trauma and can affect several critical functions such as mastication, speech, and aesthetics. Accurate detection and anatomical classification of these fractures are crucial for the success of the treatment process. Panoramic radiographs are commonly used imaging tools for the evaluation of mandibular fractures. With recent technological advancements, the integration of artificial intelligence-based systems into medical imaging has increased and these systems have demonstrated potential contributions to clinical decision-making processes. In this study, the diagnostic performances of two deep learning algorithms, YOLOv5 and YOLOv8, were compared for the automatic detection of mandibular fractures on panoramic radiographs. Both YOLOv5 and YOLOv8, which are single-stage object detection models, were trained to identify fractures in specific anatomical regions of the mandible (corpus, angulus, ramus, condyle, symphysis). A total of 400 panoramic radiographs were used in the study, and the dataset was expanded to 980 images using data augmentation techniques. The dataset was split into 80% training, 10% validation, and 10% testing. The performances of the models were evaluated using metrics such as precision, recall, F1-score, mean Average Precision (mAP), and Intersection over Union (IoU). According to the test results, the YOLOv8 model achieved 0.85 precision, 0.83 recall, 0.84 F1-score, 0.89 mAP, and 0.82 IoU, demonstrating overall higher performance. The YOLOv5 model, on the other hand, achieved 0.81 precision, 0.78 recall, 0.79 F1-score, 0.84 mAP, and 0.77 IoU. Although both models showed high accuracy in the detection of mandibular fractures, YOLOv8 exhibited a more stable and superior performance. The findings of this study suggest that deep learning-based object detection algorithms can serve as effective decision support tools in the evaluation of mandibular fractures on panoramic radiographs.