Abstract
Purpose: Conducting channels (CCs) are responsible for the initiation and maintenance of infarcted-derived ventricular tachycardia (VT). It is necessary to accurately locate CCs to finish infarct-derived VT by radiofrequency ablation (RFA). It is not possible to detect CCs by evaluating only a few 2D DE-MRI (delay enhancement-MRI) slices. For detecting CCs it is necessary to build a 3D reconstruction of the anatomy of the left ventricle (LV) and of the infarct scar (core and border zone (BZ)) from a high-resolution DE-MRI stack. There is no consensus about which of the semi-automatic existing methods to delineate infarct scars from DE-MRI is more accurate or about the thresholds to apply for properly differentiating between core, BZ and healthy myocardium.
Methods: We have constructed a patient-specific 3D bi-ventricle model by manual delineation from a high-resolution DE-MRI stack including several 3D reconstructions of the ischemic region obtained from eight different semi-automatic algorithms.
Results: Figure 1 shows the 3D reconstructions of the scar obtained by different methods and different thresholds for the same method. The geometry and the volume of the scar depend strongly on the chosen method and thresholds.
Conclusions: When different 3D reconstructions of the scar are compared by visual inspection, it is clear that the number, size and location of detected CCs are strongly method-dependent. We have visually compared the segmented geometries to electroanatomical voltage maps (EAMs) of endocardium and epicardium of the same patient, but it is difficult to drag consistent conclusions. It is necessary to carry out a larger study involving DE-MRI images from several patients and using a more objective methodology capable of comparing 3D segmented scar volumes to 2D high-density EAMs.
