Quasi-Conformal Technique for Integrating and Validating Myocardial Tissue Characterization in MRI with Ex-Vivo Human Histological Data

Abstract

Ventricular tachycardia caused by a circuit of re-entry is one of the most critical arrhythmias. It is usually related with heterogeneous scar regions where slow velocity of conduction tissue is mixed with non-conductive tissue, creating pathways (CC) responsible for the tachycardia. Pre-operative DE-MRI can provide information on myocardial tissue viability and then improve therapy planning. However, the current DE-MRI resolution is not sufficient for identifying small CCs and therefore they have to be identified during the intervention, which requires considerable operator experience. In this work, we studied the relationship of histological data (with 10 $\mu$m resolution), with in-vivo DE-MRI pixel intensities (PI) of one human heart. Integrating multi-modal data provided by different nature (in- vs. ex-vivo; 3D volume vs. 2D slices) is not straightforward and requires a robust integration pipeline. The main purpose of this work, is to develop a new technique for integrating histological information into the corresponding DE-MRI one. The proposed quasi-conformal mapping technique (QCM) integration were compared with state-of-the-art registration techniques (affine and non-rigid) on a benchmark of 418 synthetically generated datasets showing a more robust results. We used the QCM to quantitatively compare DE-MRI PI with the percentage of fibrosis extracted from histology. We show a positive correlation between the DE-MRI PI and the percentage of fibrosis extracted from histology (r = 0.97; p < 0.0001). Furthermore, we found a significant amount of viable tissue (up to 50$\%$) in areas commonly defined as core zone in DE-MRI (PI level > 60$\%$ of the maximum intensity value).