2 July 2025 in Nature Cardiovascular Research
Sudden cardiac death from ventricular arrhythmias is a leading cause of mortality worldwide. Arrhythmic death prognostication is challenging in patients with hypertrophic cardiomyopathy (HCM), a setting where current clinical guidelines show low performance and inconsistent accuracy. Here, we present a deep learning approach, MAARS (Multimodal Artificial intelligence for ventricular Arrhythmia Risk Stratification), to forecast lethal arrhythmia events in patients with HCM by analyzing multimodal medical data. MAARS’ transformer-based neural networks learn from electronic health records, echocardiogram and radiology reports, and contrast-enhanced cardiac magnetic resonance images, the latter being a unique feature of this model. MAARS achieves an area under the curve of 0.89 (95% confidence interval (CI) 0.79–0.94) and 0.81 (95% CI 0.69–0.93) in internal and external cohorts and outperforms current clinical guidelines by 0.27–0.35 (internal) and 0.22–0.30 (external). In contrast to clinical guidelines, it demonstrates fairness across demographic subgroups. We interpret MAARS’ predictions on multiple levels to promote artificial intelligence transparency and derive risk factors warranting further investigation.
18 June 2024 in Nature Cardiovascular Research
CYCarolyna A. P. Yamamoto
+9
Atrial fibrillation (AF), the most common heart rhythm disorder, may cause stroke and heart failure. For patients with persistent AF with fibrosis proliferation, the standard AF treatment—pulmonary vein isolation—has poor outcomes, necessitating redo procedures, owing to insufficient understanding of what constitutes good targets in fibrotic substrates. Here we present a prospective clinical and personalized digital twin study that characterizes the arrhythmogenic properties of persistent AF substrates and uncovers locations possessing rotor-attracting capabilities. Among these, a portion needs to be ablated to render the substrate not inducible for rotors, but the rest (37%) lose rotor-attracting capabilities when another location is ablated. Leveraging digital twin mechanistic insights, we suggest ablation targets that eliminate arrhythmia propensity with minimum lesions while also minimizing the risk of iatrogenic tachycardia and AF recurrence. Our findings provide further evidence regarding the appropriate substrate ablation targets in persistent AF, opening the door for effective strategies to mitigate patients’ AF burden.
29 October 2018 in Journal of Cardiovascular Electrophysiology
20 September 2023 in JACC: Clinical Electrophysiology
9 October 2024 in IEEE Transactions on Biomedical Engineering
Myocardial fibrosis is a crucial marker of heart muscle injury in several heart disease like myocardial infarction, cardiomyopathies, and atrial fibrillation (AF). Fibrosis and associated scarring (dense fibrosis) are also vital for assessing heart muscle pre- and post-intervention, such as evaluating left atrial (LA) fibrosis/scarring in patients undergoing catheter ablation for AF. Although cardiac MRI is the gold standard for fibrosis assessment, current quantification methods are unreliable due to their reliance on variable thresholding and sensitivity to MRI uncertainties, lacking standardization and reproducibility. Importantly, current methods focus solely on quantifying fibrosis volume ignoring the unique MRI characteristics of fibrosis density and unique distribution, that could better inform on disease severity. To address these issues, we propose a novel threshold-free self-calibrating probabilistic method called “Fibrosis Signatures.” This method efficiently encodes ∼9 billion MRI intensity co-disparities per scan into standardized probability density functions, deriving a unique MRI fibrosis signature index (FSI). The FSI index quantitatively encodes fibrosis/scar extent, density, and distribution patterns simultaneously, providing a detailed assessment of burden/severity. Our self-calibrating design mitigates impacts of MRI uncertainties, ensuring robust evaluations pre- and post-intervention under varying MRI qualities. Extensively validated using a novel numerical phantom and 143 in vivo LA 3D MRIs of AF patients (pre- and post- ablation and serial post-ablation scans) and compared to 5 existing methods, our FSI index demonstrated strong correlations with traditional fibrosis measures and was able to quantify density and distribution pattern beyond entropy. FSI was up to 9 times more reliable and reproducible to MRI uncertainties (noise, segmentation, spatial resolution), highlighting its potential to improve cardiac MRI reliability and clinical utility.