Comprehensive Fully Automated Left-Ventricular Ejection and Contractility Analysis

This study presents a novel fully-automated mechanism for MRI-based profiling of left-ventricular (LV) chamber quantities and mechanical properties (3D strains, twist and torsion) for the cardiac cycle. Comprehensive LV contractile analysis was conducted by extending a validated semi-automated methodology for spatiotemporal myocardial boundary detection, whose underlying algorithm consists of a multi-threshold image quantization process and 3D phase-unwrapping of complex MRI images acquired with the Displacement Encoding with Stimulated Echoes (DENSE) sequence. Demonstrating this tool's feasibility included cross-validating computations of LV torsion using two different techniques (a boundary deformation-based and a strain-based one) and by showing how torsional mechanics correlate with ejection fraction (EF) in the phase following isovolumetric contraction. The quantization and displacement results acquired in all 2D partitions at subsequent timeframes lead to 3D LV reconstructions and determination of contractile parameters in N=15 healthy subjects. Statistical tests consisted of Bland Altman analysis of results from the two torsional computations and establishing correlations between EF and torsion.Bland Altman biases and limits of agreement between the two methods for regional torsions were 0.20 ± 2.4° and 0.19 ± 1.97° for the mid-ventricular and basal sub-regions (both relative to apex), respectively. Strong Pearson's correlations found between torsion and EF were R² = 0.80 (p <; 0.001) and R² = 0.82 (p <; 0.001) for the mid-ventricular and basal segments, respectively. The results from the statistical analysis demonstrate that the methodology developed for comprehensive cardiac contractility analysis can be used reliably in healthy subjects, with a potential for extending application in diagnosing cardiac dysfunction.