Clinical oral investigations · 2026
Evaluating upper airway in orthodontics via 3D UX-Net model on CBCT scans.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
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.
Liu Yisi, Zhan Jiaqi, Zhang Jiaqi, Hong Wentao, Fan Wenjie, Shihabi Ramy, Lan Zhicong, Gu Shuhan, Pan Linqiang, Hu Li
Аннотация
INTRODUCTION AND AIMS: The relationship between orthodontic treatment and upper airway morphology is increasingly recognized. Artificial intelligence (AI) now supports airway analysis, but traditional 3D U-Net models show limited accuracy, particularly in the laryngopharynx. This study proposes a deep learning model to accurately and efficiently extract 3D upper airway structures from CBCT scans, facilitating improved orthodontic monitoring.
METHODS: The 3D UX-Net was employed for airway segmentation. Biased pharyngeal interface information from the network output enabled precise localization of boundary landmarks on the midsagittal plane, enhancing interface delineation.
RESULTS: On internal 5-fold cross-validation, 3D UX-Net achieved a mean Dice similarity coefficient (DSC) of 0.953 ± 0.007 for total airway segmentation, outperforming existing methods. External validation across three geographic datasets confirmed strong generalization. After refining the pharyngeal interface via midsagittal landmarks, mean DSC improved to 0.963 ± 0.006.
CONCLUSION: The proposed model enables high-precision upper airway segmentation, supporting more efficient and comprehensive clinical image analysis.
CLINICAL RELEVANCE: This study addresses the insufficient segmentation accuracy of prior 3D U-Net models, especially in the laryngeal region, offering enhanced reliability for orthodontic airway assessment.
METHODS: The 3D UX-Net was employed for airway segmentation. Biased pharyngeal interface information from the network output enabled precise localization of boundary landmarks on the midsagittal plane, enhancing interface delineation.
RESULTS: On internal 5-fold cross-validation, 3D UX-Net achieved a mean Dice similarity coefficient (DSC) of 0.953 ± 0.007 for total airway segmentation, outperforming existing methods. External validation across three geographic datasets confirmed strong generalization. After refining the pharyngeal interface via midsagittal landmarks, mean DSC improved to 0.963 ± 0.006.
CONCLUSION: The proposed model enables high-precision upper airway segmentation, supporting more efficient and comprehensive clinical image analysis.
CLINICAL RELEVANCE: This study addresses the insufficient segmentation accuracy of prior 3D U-Net models, especially in the laryngeal region, offering enhanced reliability for orthodontic airway assessment.