Cardiac dyssynchrony analysis using circumferential versus longitudinal strain: implications for assessing cardiac resynchronization

Abstract

Background: QRS duration is commonly used to select heart failure patients for cardiac resynchronization therapy (CRT). However, not all patients respond to CRT, and recent data suggest that direct assessments of mechanical dyssynchrony may better predict chronic response. Echo-Doppler methods are being used increasingly, but these principally rely on longitudinal motion (epsilonll). It is unknown whether this analysis yields qualitative and/or quantitative results similar to those based on motion in the predominant muscle-fiber orientation (circumferential; epsiloncc).

Methods and results: Both epsilonll and epsiloncc strains were calculated throughout the left ventricle from 3D MR-tagged images for the full cardiac cycle in dogs with cardiac failure and a left bundle conduction delay. Dyssynchrony was assessed from both temporal and regional strain variance analysis. CRT implemented by either biventricular (BiV) or left ventricular-only (LV) pacing enhanced systolic function similarly and correlated with improved dyssynchrony based on epsiloncc-based metrics. In contrast, longitudinal-based analyses revealed significant resynchronization with BiV but not LV for the overall cycle and correlated poorly with global functional benefit. Furthermore, unlike circumferential analysis, epsilonll-based indexes indicated resynchronization in diastole but much less in systole and had a lower dynamic range and higher intrasubject variance.

Conclusions: Dyssynchrony assessed by longitudinal motion is less sensitive to dyssynchrony, follows different time courses than those from circumferential motion, and may manifest CRT benefit during specific cardiac phases depending on pacing mode. These results highlight potential limitations to epsilonll-based analyses and support further efforts to develop noninvasive synchrony measures based on circumferential deformation.

Figure 1

A, Schematic for determination of TUS dyssynchrony index. Strain is plotted at a given time as a function of spatial location of the segment. Data are processed by Fourier series decomposition. The zero-order (S₀) and first-order (S₁) terms are shown plotted vs spatial position. A perfectly synchronous heart would appear as a straight line (solely S₀ term), whereas one that was perfectly dyssynchronous would appear as a sinusoid (S₁ term). The index relates the relative power of these 2 terms (Appendix). B, Schematic for calculating vector-strain index. For a given short-axis slice, strains at each segment are determined, and this value was multiplied by a unit vector pointing in the direction of the segment. The vectors are then added, and the sum V reflects the primary orientation and magnitude of contraction. If all segments contract at the same time, the magnitude of V is zero; with increasing dyssynchrony of contraction, the magnitude of V increases.

Figure 2

Temporal-spatial maps of ε_cc and ε_ll for RA (LBBB), LV, and BiV pacing modes. Each subplot within a map shows the time-dependent (x axis) circumferential or longitudinal (y axis) motion at a spatial point around the LV. A zone of increased ε_cc and ε_ll strain is demonstrated in the lateral wall of the RA (LBBB) paced heart (A, D). Both LV and BiV modify this lateral wall strain pattern, although the effects are less pronounced with ε_ll (E, F) than with ε_cc (B, C). See text for details.

Figure 3

Comparison of full-cycle ε_cc- and ε_ll-derived temporal uniformity in the LBBB dyssynchrony model. With both LV and BiV pacing, the ε_cc-derived uniformity index positively correlates with enhanced EF (A). However, unlike ε_cc, ε_ll uniformity is not enhanced with LV pacing despite improved hemodynamic function (B). The dynamic range of ε_ll uniformity is much less than that of ε_cc uniformity. Ejection fraction is adjusted according to its mean value for each respective animal. C, ε_cc and ε_ll temporal uniformity in the LBBB dyssynchrony model averaged over systole, diastole, and the full cardiac cycle. *P<0.003, †P=0.04 vs RA (LBBB) dyssynchronous baseline; ‡P=0.05 vs BiV. Data are mean±SEM.

Figure 4

Time-dependent plots of cardiac dyssynchrony and resolution of dyssynchrony based on analysis of ε_cc and ε_ll variance (A, B) and spatially weighted strain disparities (C, D). The strain variance rises (more dyssynchrony) and peaks at late systole and early diastole. For each pacing mode, LV and BiV are compared with RA (dyssynchronous) pacing at early, mid, and late time points in both systole and diastole (gray highlight). The time-dependent spatially weighed vector index (RVVPS) is also displayed for ε_cc and ε_ll (C, D). *P<0.009, †P<0.05 vs dyssynchronous (RA pacing, LBBB) baseline.

Figure 5

Schematic of difference in net dyssynchrony based on geographic distribution of regions of delayed activation. When regions are locally clustered, the net impact on dysfunction is greater, whereas if they are dispersed throughout the wall, the impact is less. Importantly, both situations could result in similar numbers of delayed segments overall (counting dyssynchrony indexes) and similar variance of the delays. In contrast, a vector index would be zero for dispersed discoordinate shortening but non-zero when discoordinate motion was geographically clustered.