JTCVS open · 2026
MVRepairAI: A machine learning-based system to predict surgical methods in mitral valve repair.
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.
AlGhamdi Mohammed, Bata Kemal Bori, Lellouche Nicolas, Vitiello Damien, Leprince Pascal, Saiydoun Gabriel
Аннотация
OBJECTIVE: This study aimed to develop a machine learning model, MVRepairAI, that predicts appropriate surgical repair techniques for mitral valve pathology using preoperative echocardiographic data.
METHODS: A retrospective cohort study was conducted on 180 patients who underwent primary mitral valve repair between 2017 and 2019. Preoperative transthoracic and transesophageal echocardiography reports were documented, which detailed segmental pathology, etiologic determinants, and morphologic features. The MVRepairAI model used a hierarchical clinical decision tree to predict surgical techniques on the basis of these echocardiographic data. Predicted techniques were compared with documented operative techniques using multiclass accuracy metrics, precision, recall, and F1 scores. Subgroup validation assessed resection-type precision and technique disagreement across etiological strata.
RESULTS: MVRepairAI achieved 92.22% overall accuracy (95% CI, 89.1-94.7%; P < .001) in matching the intraoperative approach. Precision and recall were 89.5% (86.2-92.1%) and 91.7% (88.8-94.0%), respectively, with an F1 score of 90.6% (87.6-92.9%). Resection-specific predictions were 94.3% accurate (91.5-96.3%). Agreement between predicted and actual techniques exceeded 93% for all major interventions. Endocarditis etiology, pulmonary hypertension, and leaflet calcification were independent negative predictors of accuracy, whereas the presence of Barlow disease enhanced accuracy. Valve complexity did not impair performance.
CONCLUSIONS: MVRepairAI demonstrates the potential of artificial intelligence to convert preoperative imaging into surgically pertinent plans for mitral valve repair. The hierarchical model structure showed substantial concordance with operative approaches across diverse pathologic presentations. Future refinements require rigorous multicenter validation, integration of dynamic intraoperative data, and longitudinal outcomes analysis to further advance this foundational platform for standardized, patient-specific mitral valve restoration.
METHODS: A retrospective cohort study was conducted on 180 patients who underwent primary mitral valve repair between 2017 and 2019. Preoperative transthoracic and transesophageal echocardiography reports were documented, which detailed segmental pathology, etiologic determinants, and morphologic features. The MVRepairAI model used a hierarchical clinical decision tree to predict surgical techniques on the basis of these echocardiographic data. Predicted techniques were compared with documented operative techniques using multiclass accuracy metrics, precision, recall, and F1 scores. Subgroup validation assessed resection-type precision and technique disagreement across etiological strata.
RESULTS: MVRepairAI achieved 92.22% overall accuracy (95% CI, 89.1-94.7%; P < .001) in matching the intraoperative approach. Precision and recall were 89.5% (86.2-92.1%) and 91.7% (88.8-94.0%), respectively, with an F1 score of 90.6% (87.6-92.9%). Resection-specific predictions were 94.3% accurate (91.5-96.3%). Agreement between predicted and actual techniques exceeded 93% for all major interventions. Endocarditis etiology, pulmonary hypertension, and leaflet calcification were independent negative predictors of accuracy, whereas the presence of Barlow disease enhanced accuracy. Valve complexity did not impair performance.
CONCLUSIONS: MVRepairAI demonstrates the potential of artificial intelligence to convert preoperative imaging into surgically pertinent plans for mitral valve repair. The hierarchical model structure showed substantial concordance with operative approaches across diverse pathologic presentations. Future refinements require rigorous multicenter validation, integration of dynamic intraoperative data, and longitudinal outcomes analysis to further advance this foundational platform for standardized, patient-specific mitral valve restoration.