The effect of scar imaging quality on in-silico pace-mapping of post-infarction ventricular tachycardia

Abstract

Abstract Funding Acknowledgements Type of funding sources: Foundation. Main funding source(s): Wellcome Trust and British Heart Foundation. Background In-silico pacing mapping is a technique which involves using an anatomically-personalised image-based computational model to generate a ‘virtual’ clinical pace map for enhanced pre-procedural ablation planning of post-infarction ventricular tachycardia (VT). Utilising, as it does, a direct clinical ECG (or implanted device) recording of the clinical VT in order to construct the map, it has a number of advantages over other computational methods which attempt to simulate patient-specific VTs to identify ablation targets. However, accurate model construction relies on adequate scar imaging data, which may not be available with current clinical cardiac MR (CMR) sequences. Purpose To investigate the impact of scar imaging quality on the ability of in-silico pace-mapping to determine the VT origin. Methods VTs were simulated in three post-infarction porcine heart models reconstructed from high resolution (1mm isotropic) contrast-enhanced CMR data. Visual analysis of the VT circuits was performed to identify the VT site of origin (exit site) in each case (as the gold standard). The ECG of each simulated VT was computed and used as input for the construction of the in-silico virtual pace-mapping approach. Prior to pace map construction, the scar anatomy was modified to represent three different levels of imaging resolution: high resolution scar mapping (default); low resolution scar (representing standing clinical CMR sequences), obtained by performing dilation and erosion in the scar to close all inner pathways; and, finally, no scar (cardiac anatomy only, representing CT-based reconstructions). In-silico pace-mapping was performed by pacing the heart in N = 1000 sites from randomly chosen locations, on both surfaces and intramurally, within close proximity (<5mm) from the infarct. ECG signals computed following each paced activation wavefront were compared to the VT ECG through correlation analysis to construct the virtual pace map. The distance d between the VT exit site and the pacing location with the strongest correlation was also computed. Results As can be seen in Figure 1, correlation between VT and paced ECGs was found to be strongest at pacing locations near the VT exit site in all models. Correlation coefficients were similar in value and distribution in models with high- (min: -0.49±0.27 and max: 0.92±0.05) and low-resolution scars (min: -0.46±0.29 and max: 0.93±0.05). While the correlation map in the absence of scar had smaller values when compared to the other two models (min: -0.50±0.24 and max: 0.71±0.09), the VT exit was still located near the pacing location with the strongest correlation: d = 9.47±2.83 (high-res), d = 6.63±6.35mm (low-res) and d = 18.60±7.10mm (no scar). Conclusion In-silico pace-mapping successfully detected the VT exit sites regardless the resolution of the scar. This finding suggest that the level of infarct anatomical detail has a minor effect on simulated paced activation to localize VT ablation.