Quantification of local changes in myocardial motion by diffeomorphic registration via currents: Application to paced hypertrophic obstructive cardiomyopathy in 2D echocardiographic sequences

Abstract

Time-to-peak measurements and single-parameter observations are cumbersome and often confusing for quantifying local changes in myocardial function. Recent spatiotemporal normalization techniques can provide a global picture of myocardial motion and strain patterns and overcome some of these limitations. Despite these advances, the quantification of pattern changes remains descriptive, which limits their relevance for longitudinal studies. Our paper provides a new perspective to the longitudinal analysis of myocardial motion. Non-rigid registration (diffeomorphic registration via currents) is used to match pairs of patterns, and pattern changes are inferred from the registration output. Scalability is added to the different components of the input patterns in order to tune up the contributions of the spatial, temporal and magnitude dimensions to data changes, which are of interest for our application. The technique is illustrated on 2D echocardiographic sequences from 15 patients with hypertrophic obstructive cardiomyopathy. These patients underwent biventricular pacing, which aims at provoking mechanical dyssynchrony to reduce left ventricular outflow tract (LVOT) obstruction. We demonstrate that our method can automatically quantify timing and magnitude changes in myocardial motion between baseline (non-paced) and 1 year follow-up (pacing on), resulting in a more robust analysis of complex patterns and subtle changes. Our method helps confirming that the reduction of LVOT pressure gradient actually comes from the induction of the type of dyssynchrony that was expected.

Machine learning‐based phenogrouping in heart failure to identify responders to cardiac resynchronization therapy

“…only a peak value (such as end-diastolic and end-systolic volume or peak systolic strain). Prior to analysis, these traces need to be referenced to a common temporal framework 18 (see online supplementary Methods S1). The 49 segmental strain traces available from the 2-chamber and 4-chamber views were converted into two basal, two mid-LV and two apical segments, by isolating and averaging groups of eight consecutive traces.…”

Machine-learning–based exploration to identify remodeling patterns associated with death or heart-transplant in pediatric-dilated cardiomyopathy

“…11 A detailed description of the temporal normalisation procedure can be found in the Supplementary Information. 12 .…”

Machine Learning Approaches for Myocardial Motion and Deformation Analysis

“…(73,74) explicitly addressed the problem through spatiotemporal tensor decomposition. Duchateau et al (75) tuned up the contributions of the spatial, temporal, and magnitude dimensions to analyze changes in deformation patterns through registration. Jia et al (31) and Guigui et al (32) transported temporal trajectories without explicitly extracting motion or deformation descriptors beforehand.…”