Detection and classification of enhanced periapical lesion images with YOLO algorithms

OpenAlex İngilizce

In recent years, artificial intelligence has become a reliable technology in clinical decision support systems with the solutions it offers in the dental field. This success also indicates a significant potential in detecting five different types of lesions on the tooth root through radiographic images. Because the periapical X-ray-taking process may vary depending on the individual's physical, psychological, and mental conditions. Also, environmental parameters may negatively affect the image acquisition process. It is possible to tolerate these disadvantageous situations with artificial intelligence-based algorithms and image processing approaches. In this study, it was planned to in-depth analysis of the periapical lesion data types in the original dataset called Periapical X-rays provided from the Kaggle public database. For this, the original adaptive image processing approach developed by integrating the ABC optimisation algorithm was applied to the dataset for five different lesion types. Then, the enhanced images enriched with the data augmentation approach were trained with YOLOv7, YOLOv8, YOLOv9 and YOLOv10 algorithms. As a result of the training, the enhanced images compared to the original images reached 96% F Criterion thanks to the network architecture of YOLOv8 algorithm. This shows that the YOLOv8 network architecture in enhanced lesion images is more successful.