Detection of ectopic eruption in permanent first molars using deep learning methods

Objective: Ectopic eruption refers to the abnormal positioning of a tooth due to growth limitations in the jaw and jaw segments. This study aims to evaluate the effectiveness of artificial intelligence in detecting ectopic eruption of permanent first molars and determining its severity in panoramic radiographs using a deep learning method. Materials and Methods: The study used panoramic images of 12,934 pediatric patients, aged 5-10, who visited the Pediatric Dentistry Clinic at the Faculty of Dentistry of Ordu University between January 2015 and September 2023 for various reasons. A total of 876 images included in the study were classified based on the presence/absence of ectopic eruption and severity of it. For the detection and classification of ectopic eruption, a CNN-based model, Mask R-CNN architecture, and algorithms such as Data Efficient Image Transformer (DeiT) or BERT Pre-Training of Image Transformers (BeiT) were used. The performance of the artificial intelligence model was evaluated using the K-Fold Cross Validation method and a confusion matrix. Results: The AI model demonstrated high performance in detecting ectopic eruption in panoramic radiographs. However, in classifying the severity of ectopic eruption, the results were lower compared to detection accuracy. Specifically, the detection of ectopic eruption having moderate severity showed lower success rates. Using a confusion matrix, values of accuracy, precision, recall, and F1-score were calculated for each tooth and these values were used for severity classification. Conclusion: Early detection of ectopic eruption tendencies in permanent first molar teeth and comprehensive evaluation to plan appropriate treatment can help prevent potential problems and reduce the need for lengthy and costly treatments in the future. Keywords: Ectopic Eruption, Permanent First Molar, Deep Learning Method, Artificial Intelligence