Journal of the American Heart Association · 2026
Automated Classification of Mitral and Tricuspid Regurgitation With Explainability and Real-World Practice Experience.
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.
Lin Wei-Chen, Li Yi-Ting, Chen Yu-De, Hsiao Ting-An, Li Wen-Yao, Tsai Wei-Chuan, Liu Ping-Yen, Huang Mu-Shiang
Аннотация
BACKGROUND: Mitral regurgitation and tricuspid regurgitation frequently coexist and are evaluated using overlapping echocardiographic views. Although artificial intelligence-based approaches have shown promise, current existing models lack explainability and physiologic constraints, limiting their reliability and adoption in real-world echocardiographic workflows.
METHODS: We developed an end-to-end, explainable, flow-aware deep learning pipeline that automatically selects relevant echocardiographic views from routine Digital Imaging and Communications in Medicine files, performs valve morphology assessment via leaflet pose estimation, incorporates systolic phase awareness for Doppler interpretation, and jointly grades mitral regurgitation and tricuspid regurgitation severity from transthoracic echocardiography. The model was trained and internally tested using 5086 outpatient studies from a tertiary center and externally validated in independent cohorts from 2 additional institutions. Model performance was evaluated using area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, and negative predictive value, with subgroup analyses across key clinical strata.
RESULTS: In the internal test set, the model demonstrated excellent discrimination for clinically significant (moderate or greater) mitral regurgitation and tricuspid regurgitation (area under the receiver operating characteristic curve >0.980), with high negative predictive values (>0.970). Performance remained robust across age, sex, and valvular comorbidities, with modest attenuation observed in patients with atrial fibrillation or reduced left ventricular ejection fraction. In the external cohort, areas under the receiver operating characteristic curve were modestly lower, particularly for severe regurgitation; however, sensitivity and negative predictive value remained high.
CONCLUSIONS: This explainable, unified artificial intelligence framework enables accurate and efficient screening of mitral regurgitation and tricuspid regurgitation in routine practice. By maintaining strong rule-out capability and robust external performance, the proposed system supports automated triage and prioritization of echocardiographic studies in both specialist and noncardiology settings.
METHODS: We developed an end-to-end, explainable, flow-aware deep learning pipeline that automatically selects relevant echocardiographic views from routine Digital Imaging and Communications in Medicine files, performs valve morphology assessment via leaflet pose estimation, incorporates systolic phase awareness for Doppler interpretation, and jointly grades mitral regurgitation and tricuspid regurgitation severity from transthoracic echocardiography. The model was trained and internally tested using 5086 outpatient studies from a tertiary center and externally validated in independent cohorts from 2 additional institutions. Model performance was evaluated using area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, and negative predictive value, with subgroup analyses across key clinical strata.
RESULTS: In the internal test set, the model demonstrated excellent discrimination for clinically significant (moderate or greater) mitral regurgitation and tricuspid regurgitation (area under the receiver operating characteristic curve >0.980), with high negative predictive values (>0.970). Performance remained robust across age, sex, and valvular comorbidities, with modest attenuation observed in patients with atrial fibrillation or reduced left ventricular ejection fraction. In the external cohort, areas under the receiver operating characteristic curve were modestly lower, particularly for severe regurgitation; however, sensitivity and negative predictive value remained high.
CONCLUSIONS: This explainable, unified artificial intelligence framework enables accurate and efficient screening of mitral regurgitation and tricuspid regurgitation in routine practice. By maintaining strong rule-out capability and robust external performance, the proposed system supports automated triage and prioritization of echocardiographic studies in both specialist and noncardiology settings.