A CMR study of the effects of tissue edema and necrosis on left ventricular dyssynchrony in acute myocardial infarction: implications for cardiac resynchronization therapy

The study protocol was approved by the local Ethics Committee and all subjects gave written informed consent before study participation. Between November 2006 and August 2010, 22 patients (20 male/2 female, age = 56.9 ± 12.4 years) were examined at baseline 7.5 ± 4.1 days after first anterior MI and again at follow-up 4 months later (120.8 ± 21 days after MI). Ten patients participated in a previously published study [1]. Exclusion criteria for study participation were contraindications to CMR (typically incompatible metallic implants, claustrophobia). All recruited patients completed the study and no patients were lost to follow-up. Detailed characteristics and medications of all patients are given in Table 1, respectively.

All measurements were performed on a 1.5 T MRI scanner (Achieva, Philips Healthcare, Best, The Netherlands). All acquisitions were performed during end-inspiration breath-hold periods. After acquiring survey images and reference data for sensitivity encoding (SENSE) [21] LV short-axis and long-axis images were acquired in order to plan subsequent scans. To assess global LV function, short-axis images covering the entire LV were then acquired in all patients using a steady state fee precession (SSFP) sequence (cardiac phases = 25, temporal resolution = 45 ms, slice thickness = 8 mm, 13–16 contiguous slices).

For the 3D tagging data acquisition an accelerated acquisition scheme was applied [1, 16]. Imaging parameters were as follows: FOV = 108x108x108 mm³, matrix size = 28x14x16, receiver bandwidth = 220 Hz/pixel, number of profiles per EPI-segment = 7, number of excitations per heart phase = 4, echo time = 3.3 ms.

Edema imaging was performed with a standard T2-weighted (T2w) double-inversion black-blood spin-echo sequence (FOV = 350 × 350 mm², pixel size = 1.4 × 1.4 mm², repetition time = 2 R to R intervals, echo time = 100 ms, slice thickness = 8 mm) in 3 short-axis slices (basal, midventricular, and apical).

LGE short-axis images were acquired in all patients 15 minutes after administration of a bolus of 0.25 mmol/kg body weight of Gadobutrolum (Gadovist, Bayer Schering Pharma, Germany). An inversion recovery segmented gradient echo sequence was employed with an inversion time set to null normal myocardium (FOV = 350 × 350 mm², pixel size = 1.5 × 1.5 mm², repetition time = 7.4 ms, echo time = 4.4 ms, slice thickness = 8 mm, no slice gap). All MR data, and specifically the T2-weighted and LGE data, were acquired using a cardiac surface coil.

The 3D tagging data were post-processed with HARP [22] using an extended software tool originally designed for the analysis of two-dimensional tagging data (TagTrack, GyroTools Ltd., Zurich, Switzerland). Eight to eleven short-axis midwall contours, consisting of multiple landmark points arranged in steps of about 5˚, were defined on different slices between base and apex. Midwall circumferential shortening (csh in %) was defined as the length change of a contour relative to the original length at end-diastole (first heart phase). For a detailed analysis of csh in 48–66 segments of each LV, each midwall contour was divided into sectors of 60˚ starting at the anterior epicardial junction of the right and the left ventricle on the equatorial level and numbered consecutively from S1 to S6 in clockwise direction as viewed from the apex. The time to maximum csh (T_max) was automatically determined for each segment. Segments in which T_max was detected in the first three or in the last acquired time frame where inspected by an observer and corrected manually if necessary. As a measure for mechanical LV dyssynchrony the coefficient of variation in csh (CV _csh in %) at end-systole, as well as the circumferential uniformity ratio estimate (CURE)-index were calculated [23, 24]. CURE ranged from 0 (pure dyssynchrony) to 1 (synchronous). End-systole was defined in all patients as the moment of aortic valve closure assessed from SSFP images.

Data sets were post-processed using an in-house software (GTVolume, GyroTools Ltd., Zurich, Switzerland). For calculations of LV volumes and ejection fraction, endo- and epicardial borders were delineated manually on all acquired SSFP slices. On T₂-weighted images, edema was measured as high-signal (>2 SD above remote tissue) along the LV mid-myocardial circumference on 3 short-axis images distributed equally along the LV long axis and was expressed as percentage of LV circumference. This evaluation along the LV mid-myocardial centreline was chosen to avoid the subendocardial region where tissue signals can be difficult to distinguish from signal of stagnant blood of the LV cavity. In analogy, on LGE images, necrosis was quantified manually as percentage of LV mid-myocardial circumference on 3 short-axis images. For this analysis 3 short-axis LGE images were selected from the stack of LGE images according anatomical landmarks that matched those of the T₂-weighted images. The salvaged myocardium was calculated as area-at-risk (%LV circumference) minus necrosis (%LV circumference) and is expressed as %LV circumference (Figure 1). The total amount of necrosis and fibrosis was also quantified on LGE images covering the entire LV by summing of LGE tissue of all LV slices (expressed as %LV mass) in the acute and follow-up studies at 4 months.

Statistical analysis was performed by using the SPSS software package release 15.0.1 (Chicago, IL, USA) and MedCalc (Mariakerke, Belgium). Mean values + SD were calculated for all groups. The paired Student’s t-test or Wilcoxon test and independent samples t-test or Mann–Whitney test were used to test for differences within and between groups. All tests were two-tailed. The Kolmogorov–Smirnov test was used to test for normality. P-values less than 0.05 were considered statistically significant. The study was powered for the CURE index determined by 3D tagging CMR. A previous study [1] yielded a difference in CURE of 0.08 between controls and patients with old MIs assessed by CMR. Assuming that in the acute phase edema would decrease CURE by another 50% of this difference (= expected difference of 0.04 in CURE in the acute vs chronic MI), and considering a SD of 0.03 (intra-observer difference in (1) for CURE), a sample size of n = 18 (α = 0.05; power 80%) would be needed. A larger patient group was recruited to account for possible dropouts and a potentially smaller difference of the CURE index at 4 month follow-up, which was not known a priori.

MR tagging techniques [7] well as other modalities [25, 26] could successfully demonstrate and quantify dyssynchrony in patients with chronic MIs. In this chronic stage, scar tissue is the substrate causing dyssynchrony, most likely by the mechanism of post-systolic shortening, by which the energy stored in scar tissue during contraction is released during early diastole [7]. In the present study, dyssynchrony was also observed in the setting of acute MI. The amount of hyperintense tissue on T2-w images correlated with the degree of dyssynchrony indicating that the underlying substrate for dyssynchrony in the acute phase is tissue edema, composed of both, viable edemateous tissue as well as the central necrotic tissue. The mass of necrotic tissue alone did not correlate significantly with the degree of dyssynchrony in the acute phase of MI. Thus, measurements that consider both, viable and non-viable edematous tissue best explain intraventricular LV dyssynchrony in the acute stage. This finding also suggests that dyssynchrony in the acute phase of MI may represent at least in part a transitory phenomenon.

As was hypothesized, this study could demonstrate a regression of LV dyssynchrony during healing of acute MI (Table 2). Healing of acute MI was also associated in this patient population with an increase in LV end-diastolic and end-systolic volumes, as well as with an increase in LVEDV/LV mass, a measure of adverse remodelling. At 4 months scar tissue did shrink slightly as demonstrated earlier [27]. The reduction of dyssynchrony during the healing period was correlated with the amount of edematous tissue measured in the acute phase of MI (Figure 5a). As tissue edema can build up in necrotic as well as in viable myocardium, further analysis sought to investigate, which edema component is best predictive of regression of dyssynchrony. Interestingly, necrosis in the acute phase of MI inversely correlated with regression of dyssynchrony, thus indicating, that dyssynchrony tends to persist in ventricles with smaller necrotic mass, while dyssynchrony regresses in ventricles with larger necrosis. This might be explained by the fact, that in ventricles with very large scars, no sufficient contracting myocardium is left to cause intraventricular unloading and thus, post-systolic shortening. Extent of necrosis in the acute phase was positively correlated with LV dilation during infarct healing and a borderline correlation was observed for adverse LV remodelling during healing. Conversely, the edema mass in viable tissue during the acute phase of MI was not predictive for changes in LV dyssynchrony during healing, thus indicating, that salvaged myocardium (=edema in viable myocardium) is associated with dyssynchrony in the acute MI (Figure 2c), while it is not predictive for regression of dyssynchrony during healing (Figure 5e).

From a practical point of view the transient nature of the intraventricular dyssynchrony during the acute phase of MI may suggest, that the indication for a CRT should be made with caution during the acute phase and a re-evaluation of dyssynchrony in the chronic phase is recommended when remodelling after MI has occurred. Also, in studies aimed to evaluate responder rates to CRT in patients after AMI, the stage of LV remodeling and the presence or absence of edema in the LV myocardium should be taken into account to enable a comprehensive assessment of the benefit of CRT in this patient population.

The extent of tissue edema was measured in 3 short-axis levels only, while tagging and LGE data were acquired over the entire LV. These 3 levels for edema assessment were chosen as the examination time could not be expanded deliberately since cooperation in acute MI patients is often decreasing during long examinations. A fully 3D edema quantification given as percentage of LV mass (and not as percentage of LV circumference) might have yielded tighter correlations with dyssynchrony. As mentioned in the methods section, for edema quantification, the extent of hyperintense tissue along the circumferential midline of each slice was calculated to avoid problems in discriminating edema tissue with high signal from high signal of stagnant blood in the LV cavity. As the edema distribution was typically transmural, this approach might not have affected substantially the correlations found between edema and dyssynchrony. Having these limitations in mind, we included in this study anterior infarcts only in order to study rather large infarcts. Thus, the findings of this study should not be extrapolated to smaller lateral or inferior infarcts.

Application of the accelerated 3D tagging acquisition technique allows assessing detailed 3D motion patterns of the entire LV in only three breath-holds. While the spatial resolution of the final motion parameters is relatively low (voxel size = 3.9 x 7.7 x 7.7 mm³) with respect to current imaging standards, regional motion changes typically occur on a larger scale in the vast majority of cardiac diseases including coronary artery disease and left bundle branch block. Hence, the achieved spatial resolution appears sufficient for most applications.

As a consequence of the rather small sample size, we cannot exclude that some correlations could be missed (or might be underestimated), however, the positive correlations found are not affected by this fact and show the importance of edema in the viable tissue to cause dyssynchrony. Furthermore, we tried to compensate for a small sample size by selecting a well defined patient population and by applying a strictly standardized imaging protocol.