A Framework Combining Multi-sequence MRI for Fully Automated Quantitative Analysis of Cardiac Global And Regional Functions

In current clinical settings, there are several technological challenges to perform automated functional analysis from cardiac MRI. In this work, we present a framework to automatically segment the heart anatomy, define segments of the left ventricle, and extract myocardial motions for quantitative analysis of cardiac global and regional functions. This framework makes use of the cardiac MRI sequences that are widely available in clinical practice, and improves the performance of the automated processing by combining information from multiple MRI sequences. We employed 20 pathological datasets to evaluate the proposed framework where the automatic analysis was compared with the manual intervention assisted analysis. The results showed high correlation between the two methods for the global function analysis (volume: R-2>0.8, ejection fraction: R-2=0.88), and for the regional dyssynchrony analysis (wall motion: R-2=0.89; thickening: R-2=.81). We also found that the automated method could fully include apical and basal volume, resulting in consistent overestimation of the left ventricle volume (similar to 40mL, P<0.05) and small underestimation of ejection fraction (-0.024, P<0.001).