Novel topological method to define scar heterogeneity relates to arrhythmia vulnerability

Abstract Funding Acknowledgements Type of funding sources: Public Institution(s). Main funding source(s): KU Leuven BOF, C14/18/079 Background Healing of myocardial infarction (MI) can result in heterogeneous scar structures, resulting in border zone conduction channels with altered electrophysiological properties, that lead to arrhythmia vulnerability. Identification of scar heterogeneity and potential conduction channels based on contrast enhanced MRI has shown to be viable. However, current methods rely on manual identification and systematic approaches are lacking. Objective To provide a systematic methodology to annotate scar heterogeneity on contrast enhanced MRI and to correlate it to arrhythmia vulnerability. Methods Antero-septal MI was induced in 10 domestic pigs by 120-minute occlusion of the left anterior descending artery followed by reperfusion. Late gadolinium enhanced MR imaging was performed one month after infarction. The endocardium and epicardium of the left ventricle were segmented and the voxels inside the left ventricle (LV) were extracted. All LV voxels with signal intensity above the middle of the myocardial signal intensity range were defined to be infarct voxels and otherwise intact voxels. The infarct voxel structure and intact voxel structure were analyzed using a novel computational cubical homology approach that automatically identifies "holes" in the respective structures. The so called topological "holes" were annotated using a minimal length cycle encircling the hole (see Figure 1). This cycle annotation corresponds with scar heterogeneity as it pinpoints regions of intertwined fibrosis and viable myocardium which can potentially act as functional arrhythmia substrate. Scar heterogeneity, annotated as cycles, was compared against two measures of arrhythmia vulnerability. On one hand, the number of premature ventricular complexes (PVCs) were determined during an incremental isoproterenol infusion protocol (ranging from no infusion to 0.04 µg/kg with increments of 0.01µg/kg each 5 minutes). And on the other hand, the ease of induction is defined as the complementary percentage of the steps of the programmed electrical stimulation (PES) protocol needed to induce a ventricular tachycardia. Results Scar heterogeneity cycles were automatically annotated in all animals. In 7 of the 9 animals a ventricular tachycardia was successfully induced during the PES protocol. The ease of induction correlated with the circumference of the myocardial infarct (r = 0.69, p = 0.040). Furthermore, the ease of induction tended to correlate with the circumference of the largest heterogeneity cycle (r = 0.61, p = 0.083). The total number of PVCs throughout the isoproterenol infusion protocol correlated with the circumference of the heterogeneity cycle (r = 0.72, p = 0.031). Conclusion A novel automated and systematic method to identify scar heterogeneity on contrast enhanced MR imaging is introduced. Furthermore, the presence of large heterogeneous scar structures, potentially adding an arrhythmia substrate, correlated with arrhythmia vulnerability.