Detection of external root resorption in periapical radiographs using YOLO-based deep learning model

External root resorption is a destructive process that usually develops without any symptoms and, when diagnosed, can lead to tooth extraction because it causes serious tooth tissue loss. Therefore, it is aimed to develop artificial intelligence algorithms that can assist in the diagnosis of external root resorption.

Totally, 110 extracted teeth were demineralized by applying 40% nitric acid solution for 8 hours, 8% sodium hypochlorite for 10 minutes, and then a distilled water washing procedure. The prepared teeth were placed on a radioconjugate phantom model and imaged. The data set obtained from the teeth used in the study consists of a total of 584 periapical radiographs. YOLOv5x-cls and YOLOv5x-seg models were used to detect external root resorption.

The F1 score value of the YOLOv5x-cls model used for calcification of external root resorption was found to be 1.0, indicating that the model has a high success rate during the testing phase. In the YOLOv5x-seg model used for the segmentation of external root resorption, the F1 score values were found to be 0.8593. This value is an indication that the model is working effectively during the testing phase. It has also been determined that the classification is more successful than the segmentation model.

In this study, artificial intelligence algorithms were used in the radiological evaluation of teeth with chemical external root resorption using a phantom model compatible with jawbone radiopacity. High success rates have been achieved in the detection of external root resorption areas with artificial intelligence.

This study presents an innovative approach to detecting external root resorption using artificial intelligence. In addition, the reliability of the study was increased by using the radioconjugate phantom model.