Novel Methodology for Quantifying Left-Ventricular Dyssynchrony from Cardiac MRI Using Principal Component Analysis
Ramachandran, Raghav, Biomedical Engineering - School of Engineering and Applied Science, University of Virginia
Epstein, Frederick, Department of Biomedical Engineering, University of Virginia
Introduction: Cardiac resynchronization therapy (CRT), which restores normal electrical conduction and mechanical activity to the heart, is very effective for selected patients with systolic heart failure, particularly those with regional variations in left-ventricular (LV) motion, i.e. LV dyssynchrony. Unfortunately, CRT is associated with a 30-40% non-response rate due to sub-optimal patient selection based on electrophysiological criteria rather than measures of regional LV dyssynchrony. Current clinical measures of regional LV dyssynchrony, based on echocardiography and MRI of LV motion, are subjective and time-dependent result in significant CRT non-response.
Methods: To overcome the limitations of current measures of LV dyssynchrony, new time-independent, user-independent, data-driven metrics of LV dyssynchrony, based on LV strains, have been developed and tested in a canine model of heart failure.
A MRI method known as 2D cine displacement encoding with stimulated echoes (DENSE) provides accurate 2D basal and mid-ventricular LV strains over the entire cardiac cycle.
The new metrics are known as First Order Regional Conformity Estimate (FORCE) and First Order Regional Metric of Estimated Disparity (FORMED). FORCE and FORMED quantify regional dyssynchrony in LV motion by decomposing LV strains into principal components over space and time respectively.
Results: FORCE, FORMED and CURE were each found to be significantly different between canines with LV dyssynchrony and LV synchrony in the case of circumferential strains but not in the case of radial strains. In a preliminary analysis of 23 CRT patients a LV end-systolic volume of >=15% indicated positive response and cutoffs of CURE<0.7, FORCE< 0.795 and FORMED>0.9 indicated LV dyssynchrony. For these cutoffs, CURE, FORCE and FORMED had true negative rates (specificity) of 0.714, 0.714, 0.429 for a true positive rate (sensitivity) of 1. Areas under Receiver Operating Characteristic (ROC) curves, which define classification of CRT responders/non-responders, for CURE, FORCE and FORMED were 0.8144, 0.8416 and 0.7525 respectively.
Conclusions: FORCE and FORMED are effective metrics of LV dyssynchrony in canine models of heart failure. Also, FORCE is time and user-independent and can be used clinically to optimize CRT patient selection which improves CRT patient response.
MS (Master of Science)
All rights reserved (no additional license for public reuse)