BMC oral health · 2025
Evaluation of the effectiveness of panoramic radiography in maxillary 3rd molars on an artificial intelligence model developed with findings obtained with cone beam computed tomography.
СтатьяНазвание 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.
Kadan Esra Aydemir, Tiryaki Burcu, Miloğlu Özkan
Аннотация
BACKGROUND: Panoramic radiography (PR) is accessible for determining the contact between third maxillary molar teeth (tMMT) and the maxillary sinus floor (MSF). However, this method does not provide clear and detailed anatomical information, so more advanced imaging techniques may be required.
AIM: This study aims to evaluate the positional relationship between tMMT and the MSF using PR images analyzed by deep learning (DL) models trained with cone-beam computed tomography (CBCT) data. The study also compares the classification performance of different DL architectures.
MATERIALS AND METHODS: A total of 1,054 PR images of tMMT were analyzed. The relationship between the tMMT and MSF was categorized based on CBCT findings into three classes: no relation, contact, and sinus-related. Five DL models (VGG16, VGG19, ResNet50, ResNet101, and GoogleNet) were trained and tested across four classification problems. Performance metrics, including accuracy and confusion matrices, were evaluated. Final results were aggregated using a majority voting-based fusion strategy.
RESULTS: For binary classification (relation vs. no relation), accuracies were 89.34% for right tMMTs and 91.24% for left tMMTs. For the three-class problems (relation, contact, no relation), the accuracies were 68.72% and 69.2%, respectively. The highest classification success was achieved in the "no relation" class. Depending on the problem, the most successful models were VGG16, VGG19, and ResNet101.
CONCLUSION: DL models can effectively identify the anatomical relationship between tMMTs and MSF on PR images, especially in cases that are challenging to interpret visually. This approach has the potential to reduce reliance on CBCT imaging, providing objective diagnostic support and saving time for clinicians.
AIM: This study aims to evaluate the positional relationship between tMMT and the MSF using PR images analyzed by deep learning (DL) models trained with cone-beam computed tomography (CBCT) data. The study also compares the classification performance of different DL architectures.
MATERIALS AND METHODS: A total of 1,054 PR images of tMMT were analyzed. The relationship between the tMMT and MSF was categorized based on CBCT findings into three classes: no relation, contact, and sinus-related. Five DL models (VGG16, VGG19, ResNet50, ResNet101, and GoogleNet) were trained and tested across four classification problems. Performance metrics, including accuracy and confusion matrices, were evaluated. Final results were aggregated using a majority voting-based fusion strategy.
RESULTS: For binary classification (relation vs. no relation), accuracies were 89.34% for right tMMTs and 91.24% for left tMMTs. For the three-class problems (relation, contact, no relation), the accuracies were 68.72% and 69.2%, respectively. The highest classification success was achieved in the "no relation" class. Depending on the problem, the most successful models were VGG16, VGG19, and ResNet101.
CONCLUSION: DL models can effectively identify the anatomical relationship between tMMTs and MSF on PR images, especially in cases that are challenging to interpret visually. This approach has the potential to reduce reliance on CBCT imaging, providing objective diagnostic support and saving time for clinicians.