A novel multilayer anatomical constraint algorithm for channel detection from computed tomography to guide catheter ablation of ventricular tachycardia
A Zaman, X Lu, V Maslova, M Grehn, M Tsygankov, J Xie, A Schweikard, O Blanck, D Frank, E LianAbstract
Background
Accurate identification of arrhythmogenic substrates is essential for effective ablation of ventricular tachycardia (VT). Magnetic resonance (MR) is often not successful due to implanted devices. Proprietary software for computed tomography (CT)-based wall thinning analysis of the left ventricle (LV) is not accessible everywhere. Therefore, a custom-made algorithm for determination and visualization of the wall thinning was proposed. The algorithm was extended by a strain analysis in order to enhance accuracy by taking the mechanical function into account.
Purpose
To evaluate the algorithm prospectively in order to predict potential VT channels from CT scan and thereby guiding the VT ablation.
Methods
Contrast enhanced cardiac CT scans (end-diastolic and end-systolic) were obtained for patients scheduled for VT ablation. After segmentation the LV myocardium was bounded by two surfaces: endocardium and epicardium. A Frangi vesselness filter, typically used for vessel detection, was applied to identify regions of relative thinning (<7mm) with preserved boundaries, suggestive of potential conduction channels within border zones and scar tissue. Sudden changes in WT The wall thickness (WT) was calculated and channels were automatically determined and visualized as segmented structures by the algorithms. The WT model and channel model were merged into the mapping system. The algorithm was assessed [A1]retrospectively with 10 cases and prospectively in 2 cases during ablation procedure in patients with ischemic cardiomyopathy. Electro-anatomical mapping (EAM), e.g. voltage map, functional mapping, e.g. decrement-evoked potential (DEEP) map, and where possible, a propagation map was performed.
Results
In electro-anatomical mapping the wall thinning and resulting channels determined by the algorithm could be validated, in general. There was a great overlap between low voltage area and WT areas in all cases. Channel model showed in average 3 possible channel areas. VT activation map and functional mapping showed VT isthmus, which was always corresponded to the part of CT-based proposed channels.
Conclusion
Anatomically constrained, layer-specific strain estimation from CT imaging was performed. In comparing with corresponding functional mapping data suggest that this technique can assist in defining the critical sites in VT ablations. The next goal would be the determination of clinical relevant channels which are part of the reentrant cycle.