JACC. Clinical electrophysiology · 2025
Predicting Sudden Cardiac Death in Patients With Sarcoidosis Using a Multimodal Artificial Intelligence Model.
СтатьяПоддержка NIH: внешняя программаНазвание 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.
Lai Changxin, Yin Minglang, Kholmovski Eugene G, Sani Maryam M, Gilotra Nisha A, Chrispin Jonathan, Trayanova Natalia A
Аннотация
BACKGROUND: Patients with cardiac sarcoidosis (CS) are at high risk of sudden cardiac death (SCD), but existing guidelines relying primarily on left ventricular ejection fraction (LVEF) in recommending implantable cardioverter-defibrillator therapy have limited accuracy.
OBJECTIVES: This study sought to develop and evaluate a multimodal artificial intelligence for ventricular risk stratification in CS (MAARS-CS), integrating raw late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) images and clinical covariates to predict SCD risk in patients with sarcoidosis.
METHODS: A retrospective cohort of 317 patients with sarcoidosis was used to train and test MAARS-CS using 5-fold cross-validation. The model comprised a CMR branch using a 3-dimensional convolutional neural network for LGE-CMR image analysis, an electronic health records branch using a feedforward neural network for clinical covariates, and a classifier integrating multimodal information to predict personalized SCD risk.
RESULTS: MAARS-CS achieved an area under the receiver-operating characteristic curve (AUROC) of 0.86 (95% CI: 0.80-0.91), significantly outperforming the LVEF ≤35% criterion (AUROC: 0.59; 95% CI: 0.53-0.66; P < 0.0001 compared with MAARS-CS) and continuous LVEF (AUROC: 0.77; 95% CI: 0.67-0.85; P = 0.019 compared with MAARS-CS). The model also showed higher area under the precision-recall curve (0.54 vs 0.43) and balanced accuracy (0.83 vs 0.74) compared with continuous LVEF. MAARS-CS maintained robust performance across varying image qualities and magnetic resonance sequences. Interpretation analysis identified key clinical covariates and important image regions contributing to SCD risk predictions.
CONCLUSIONS: MAARS-CS offers superior predictive accuracy and consistency over LVEF in predicting SCD risk among patients with sarcoidosis. The model's robustness and interpretability enhance its potential as a reliable clinical decision support tool. With further validation, MAARS-CS may improve personalized patient care in sarcoidosis.
OBJECTIVES: This study sought to develop and evaluate a multimodal artificial intelligence for ventricular risk stratification in CS (MAARS-CS), integrating raw late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) images and clinical covariates to predict SCD risk in patients with sarcoidosis.
METHODS: A retrospective cohort of 317 patients with sarcoidosis was used to train and test MAARS-CS using 5-fold cross-validation. The model comprised a CMR branch using a 3-dimensional convolutional neural network for LGE-CMR image analysis, an electronic health records branch using a feedforward neural network for clinical covariates, and a classifier integrating multimodal information to predict personalized SCD risk.
RESULTS: MAARS-CS achieved an area under the receiver-operating characteristic curve (AUROC) of 0.86 (95% CI: 0.80-0.91), significantly outperforming the LVEF ≤35% criterion (AUROC: 0.59; 95% CI: 0.53-0.66; P < 0.0001 compared with MAARS-CS) and continuous LVEF (AUROC: 0.77; 95% CI: 0.67-0.85; P = 0.019 compared with MAARS-CS). The model also showed higher area under the precision-recall curve (0.54 vs 0.43) and balanced accuracy (0.83 vs 0.74) compared with continuous LVEF. MAARS-CS maintained robust performance across varying image qualities and magnetic resonance sequences. Interpretation analysis identified key clinical covariates and important image regions contributing to SCD risk predictions.
CONCLUSIONS: MAARS-CS offers superior predictive accuracy and consistency over LVEF in predicting SCD risk among patients with sarcoidosis. The model's robustness and interpretability enhance its potential as a reliable clinical decision support tool. With further validation, MAARS-CS may improve personalized patient care in sarcoidosis.