Clinical and experimental dental research · 2025
Deep Learning Models for Evaluating the Anatomical Relationship Between Posterior Maxillary Teeth and Maxillary Sinus in Panoramic Radiographs.
СтатьяПоддержка исследования: вне правительства СШАНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Fallah Akram, Soltani Parisa, Mehdizadeh Mojdeh, Riahi Farsani Mostafa, Moannaei Mahsa, Ourang Seyed Amir Hossein, Hossaini Maryam, Spagnuolo Gianrico, Rengo Carlo
Аннотация
OBJECTIVES: Accurate evaluation of the anatomical relationship between posterior maxillary teeth and the maxillary sinus is critical in dental treatments such as orthodontics, surgery, and implantology. This study investigates the efficacy of deep learning models applied to panoramic images for predicting this anatomical relationship.
MATERIALS AND METHODS: A total of 300 panoramic images and 1760 cropped slices were collected and converted to PNG format at a resolution of 512 × 512 pixels. Three convolutional neural network architectures-VGG, ResNet, and ResNeXt-were trained and evaluated. The dataset was split into training (80%), validation (10%), and testing (10%) sets. Performance metrics included accuracy, precision, recall, F1 score, ROC-AUC, and confusion matrices.
RESULTS: The VGG model achieved an accuracy, precision, recall, and F1 score of 0.84, with an ROC-AUC of 0.89. ResNet and ResNeXt demonstrated superior performance with all metrics at 0.88 and ROC-AUC values of 0.93 and 0.94, respectively. Both ResNet and ResNeXt showed signs of overfitting after 30-50 epochs, suggesting the need for early stopping or regularization. False positives mainly occurred in second molars falsely classified as contacting the sinus.
CONCLUSION: Deep learning models, particularly ResNet and ResNeXt, provide reliable assessment of the anatomical relationship between posterior maxillary teeth and the maxillary sinus using panoramic radiographs. These models can serve as efficient diagnostic aids when CBCT is unavailable, improving clinical decision-making in panoramic images.
MATERIALS AND METHODS: A total of 300 panoramic images and 1760 cropped slices were collected and converted to PNG format at a resolution of 512 × 512 pixels. Three convolutional neural network architectures-VGG, ResNet, and ResNeXt-were trained and evaluated. The dataset was split into training (80%), validation (10%), and testing (10%) sets. Performance metrics included accuracy, precision, recall, F1 score, ROC-AUC, and confusion matrices.
RESULTS: The VGG model achieved an accuracy, precision, recall, and F1 score of 0.84, with an ROC-AUC of 0.89. ResNet and ResNeXt demonstrated superior performance with all metrics at 0.88 and ROC-AUC values of 0.93 and 0.94, respectively. Both ResNet and ResNeXt showed signs of overfitting after 30-50 epochs, suggesting the need for early stopping or regularization. False positives mainly occurred in second molars falsely classified as contacting the sinus.
CONCLUSION: Deep learning models, particularly ResNet and ResNeXt, provide reliable assessment of the anatomical relationship between posterior maxillary teeth and the maxillary sinus using panoramic radiographs. These models can serve as efficient diagnostic aids when CBCT is unavailable, improving clinical decision-making in panoramic images.