MedElevator
Войти в исследования Oral diseases · 2026
Multimodal Deep Learning-Based Screening of Degenerative Temporomandibular Joint Disease Using 2D Radiography: A Cost-Effective and Low-Radiation Approach.
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.
Xiong Xin, Yang Ming, Zheng Yunhao, Zhao Liang, Li Kang, Wang Jun
Аннотация
OBJECTIVES: This study aimed to develop and preliminarily validate a multimodal deep learning model based on two-dimensional maxillofacial imaging for screening temporomandibular joint (TMJ) degenerative joint disease (DJD).
METHODS: A total of 1000 TMJs from 500 orthodontic patients were retrospectively analyzed. TMJ DJD diagnoses based on cone-beam computed tomography served as the reference standard, with substantial interrater agreement. A fine-tuned YOLOv8 model was used to localize condyles on panoramic radiographs. EfficientNetV2 networks extracted features from panoramic radiographs and lateral cephalograms, which were fused with patient age and sex for final classification.
RESULTS: The YOLOv8 model achieved an mAP50 of 0.995 for condylar localization. The multimodal model demonstrated strong screening performance, achieving an AUC of 0.898, with notable sensitivity and specificity on the test set. Grad-CAM analyses indicated attention to clinically relevant structures, and panoramic radiographs and lateral cephalograms provided complementary diagnostic information. Integrating both 2D modalities improved performance compared with panoramic radiographs alone.
CONCLUSIONS: This multimodal deep learning model enables effective, low-radiation, and cost-efficient screening for TMJ DJD using widely available 2D radiographs. Combining panoramic and lateral cephalometric imaging represents a promising approach for enhancing clinical screening accuracy. Larger multicenter studies are warranted for further validation.
METHODS: A total of 1000 TMJs from 500 orthodontic patients were retrospectively analyzed. TMJ DJD diagnoses based on cone-beam computed tomography served as the reference standard, with substantial interrater agreement. A fine-tuned YOLOv8 model was used to localize condyles on panoramic radiographs. EfficientNetV2 networks extracted features from panoramic radiographs and lateral cephalograms, which were fused with patient age and sex for final classification.
RESULTS: The YOLOv8 model achieved an mAP50 of 0.995 for condylar localization. The multimodal model demonstrated strong screening performance, achieving an AUC of 0.898, with notable sensitivity and specificity on the test set. Grad-CAM analyses indicated attention to clinically relevant structures, and panoramic radiographs and lateral cephalograms provided complementary diagnostic information. Integrating both 2D modalities improved performance compared with panoramic radiographs alone.
CONCLUSIONS: This multimodal deep learning model enables effective, low-radiation, and cost-efficient screening for TMJ DJD using widely available 2D radiographs. Combining panoramic and lateral cephalometric imaging represents a promising approach for enhancing clinical screening accuracy. Larger multicenter studies are warranted for further validation.