Unsupervised stratification of patients with myocardial infarction based on imaging and in-silico biomarkers

TL;DR

Post-MI ventricular tachycardia risk prediction is challenging; this work combines LGE-CMR-derived 3D cardiac anatomy with a fast cellular automaton (Arrhythmic3D) to simulate thousands of VT scenarios per patient. ARRISK, a normalized AR-index derived from simulation results, stratifies patients into ZERO/LOW/HIGH risk and correlates with clinical outcomes, often outperforming imaging-based risk markers. Biophysical validation with openCARP confirms the consistency of predicted exit sites and reentry circuits. The approach enables automated, rapid, patient-specific risk assessment suitable for clinical workflows and may guide targeted ablation strategies.

Abstract

This study presents a novel methodology for stratifying post-myocardial infarction patients at risk of ventricular arrhythmias using patient-specific 3D cardiac models derived from late gadolinium enhancement cardiovascular magnetic resonance (LGE-CMR) images. The method integrates imaging and computational simulation with a simplified cellular automaton model, Arrhythmic3D, enabling rapid and accurate VA risk assessment in clinical timeframes. Applied to 51 patients, the model generated thousands of personalized simulations to evaluate arrhythmia inducibility and predict VA risk. Key findings include the identification of slow conduction channels (SCCs) within scar tissue as critical to reentrant arrhythmias and the localization of high-risk zones for potential intervention. The Arrhythmic Risk Score (ARRISK), developed from simulation results, demonstrated strong concordance with clinical outcomes and outperformed traditional imaging-based risk stratification. The methodology is fully automated, requiring minimal user intervention, and offers a promising tool for improving precision medicine in cardiac care by enhancing patient-specific arrhythmia risk assessment and guiding treatment strategies.

Figures (7)

• Figure 1: LGE-CMR-based Digital Twin construction workflow and data preprocessing for simulations
• Figure 2: Segmented models for patients P7 and P5 featuring personalized anatomy, including scar, border Zone, core zone and the extraction of slow conduction channels that extend across the core zone.
• Figure 3: Results across multiple example cases, showing all simulations where ventricular tachycardia was induced, visualized on the ventricle of each case. The core zone is depicted in purple. Yellow stars indicate pacing sites, while magenta squares represent the scar exit sites where ventricular tachycardia was initiated. Blue circles highlight areas where exit points frequently cluster around specific hotspots. Despite the reentries being triggered from various pacing sites and parameter configurations, the exit sites consistently cluster in only a few regions.
• Figure 4: Detail of the phenotype of patient P5. The border zone is shown in red, the core zone in purple, and the slow conduction channels in green. The blue circle highlights the high-risk slow conduction channels that correspond with the results shown in Fig. \ref{['fig:ventClusterSims']}.
• Figure 5: Left: Visualization of the patient dataset using t-SNE. The color of each point indicates the clinical classification of the patient (red: HIGH risk, blue: ZERO risk). The scale of the axes is derived from the t-SNE projection and does not correspond to the original feature space. Right: ARRISK performance on the 51-case dataset. The outer border represents clinical risk, while the inner circle denotes the ARRISK score (red: HIGH risk, pink: LOW risk, blue: ZERO risk).