The effect of microvascular obstruction and intramyocardial hemorrhage on contractile recovery in reperfused myocardial infarction: insights from cardiovascular magnetic resonance

All patients had CMR on a 1.5 T system (Intera CV, Philips Healthcare, Best, The Netherlands) within 3 days of their index presentation and at 7–10, 30 and 90 days post-AMI. The same CMR protocol was used for each of the visits. A stack of images covering the whole LV, with the same slice geometry, position and slice thickness were used for all sequences. Cine imaging used a steady state free precession (SSFP) pulse sequence (echo time (TE) 1.4 ms; repetition time (TR) 2.8 ms; flip angle 55°, spatial resolution 2 × 2 × 10 mm, ≥18 phases per cardiac cycle), covering the whole heart in parallel short axis slices. To minimize differing volume effects between image types, 10 mm slice thickness was used for all sequences. Tagged CMR used a complementary spatial modulation of magnetization (CSPAMM) pulse sequence (spatial resolution 1.67 × 1.67 × 10 mm, no gap, tag separation 8 mm, ≥18 phases, typical TR/TE 30/6 ms, flip angle 20°). T2w CMR used a dark-blood T2w short tau inversion-recovery fast spin echo sequence (TE 100 ms, TR 2x R-R interval, flip angle 90°, spatial resolution 1.43 × 1.43 × 10 mm). T2* images were obtained with a dual echo T2* gradient echo sequence (TE 4.6/9.2 ms, TR 12.3 ms, flip angle 30°, spatial resolution 1.43 × 1.43 × 10 mm). A dose of 0.2 mmol/kg of gadolinium-DTPA (dimegluminegadopentetate; Magnevist, Bayer, Berlin, Germany) was then administered using a power injector (Spectris, Solaris, PA). A short-axis LGE stack was acquired after 10 minutes (inversion recovery-prepared T1 weighted gradient echo, inversion time according to a Look-Locker scout, TR/TE 4.9/1.9 ms, flip angle 15°, spatial resolution 1.35 × 1.35 × 10 mm). For follow-up, care was taken to ensure similar slice positioning, by aligning the proximal border of the most basal slice of the short axis stack to the mitral valve annulus in end-diastole and comparing slice position to the index scan.

Images were analyzed offline using commercial software (MASS 7.2; Medis, Leiden, The Netherlands and Tagtrack 1.8; Biomedical engineering, ETH Zurich, Switzerland). Infarct location was determined by CMR, according to standard guidelines [18]. In addition to the alignment of slices during image acquisition, we verified accurate alignment of serial scans by comparing features such as the presence and shape of papillary muscles. Left ventricular volumes and wall thicknesses were analyzed from SSFP cine-imaging. Infarcts and MO were measured from LGE images. Infarct was defined as an area of LGE ≥2 standard deviations (SD) above remote myocardium, and infarct volume estimation included any hypointense core. This cut-off was chosen for consistency with analysis of T2w images. MO was defined visually as the hypointense core within the infarcted zone and planimetered manually. Volumes of infarct and MO were calculated from planimetered areas across the whole LV stack by the modified Simpson’s method. The presence and extent of myocardial hemorrhage was assessed by combined analysis of T2w and T2* sequences [8]. On T2w images, areas with mean signal intensity more than 2 SD below the periphery of the area at risk (AAR) were considered to be hemorrhage [11, 13]. On the T2* images, the presence of a dark core within the infarcted area by visual inspection of the images was used as confirmation of myocardial hemorrhage. Only when T2w and T2* images showed concordant findings was an area considered to represent hemorrhage.

For CSPAMM analysis, endocardial and epicardial borders were drawn by a semi-automated process, and a midline calculated automatically. To minimize partial volume effects, strain was measured in the single short axis slice which demonstrated maximal infarction on LGE imaging. One short axis slice per patient at each time point was analyzed for strain. In all cases, this slice was in the same position along the long axis for each of the time points in a given patient, and every slice had LGE on the adjacent slices. Circumferential Lagrangian strain was measured at endocardial, mid-myocardial and epicardial layers through the infarct and remote zones (Figure 1). Remote strain measurement was taken in a 30 degree arc of myocardium diametrically opposite to the infarct zone. No patient had LGE in remote myocardium. To minimize the effects of passive post-systolic shortening [19], strain was measured at end systole, taken as the phase at the time of end-systole on the corresponding SSFP cine slice.

Transmural extent of infarction was graded into quartiles at each time point from anonymized LGE images at the same position as the selected CSPAMM slice by consensus of two observers (AK and SP), blinded to the results of other sequences.

Statistical analysis was performed using IBM SPSS® Statistics 19.0. Continuous variables were expressed as means ± SD. Correlation between T2w and strain data were derived using Spearman’s rank test; differences in strain and size measurements over time were evaluated using repeated measures analysis of variance (ANOVA); post-hoc testing was performed with the Bonferroni correction. Differences in transmurality quartiles over time were evaluated with the Friedman test. Normality for strain data was established using the Kolmogorov-Smirnov test. Differences in infarct pathophysiology at a single time point were evaluated using one-way ANOVA; post-hoc testing was performed with Tukey’s test. Univariable analyses were performed to identify predictors of reduced strain at 90 days. Variables with a probability value <0.1 in the univariable analysis were included in a multivariable analysis, which was based on a logistic regression model with a repeated measures variable (to adjust for non-independence of strain data). All statistical tests were 2-tailed; P values <0.05 were considered significant. Error bars for mean values denote standard error.

50 patients met the inclusion criteria. Two patients were excluded due to claustrophobia, 2 refused follow-up and 1 died before completing follow-up. In 6 other patients the tagging software failed to accurately track the tagged images due to artifact. Therefore 39 patients completed baseline and follow up scans and were included in the statistical analysis. Of these, 10 patients had tagging and T2w imaging at either day 2 or day 7, with 29 having these images at both these time points. All 39 had imaging at day 30 and day 90.

Patient data were divided into three groups for analysis; patients without MO or IMH, patients with MO but no IMH, and patients with both MO and IMH. No patient had IMH without MO. Patient characteristics were similar between the three groups (Table 1). Infarct characteristics are shown in Table 2.

Patients without MO or IMH had similar infarct size to patients with MO, but those with IMH had significantly larger infarcts than patients without IMH both at baseline (mean infarct volume 41 ± 20 ml vs. 21 ± 16 ml, p = 0.02) and 90 days (mean infarct volume 21 ± 16 ml vs. 14 ± 10 ml, p < 0.01). Infarct size decreased significantly over time in all three groups (no MO group, F = 7.5, p < 0.01; MO with no IMH group, F = 9.8, p < 0.01; MO and IMH group, F = 7.2, p = 0.01) (Table 2 and Figure 2). There was also a significant decrease in infarct transmural extent over time in the patients with MO and IMH (p < 0.01), but not in patients with MO and no IMH (p = 0.1) or without MO or IMH (p = 0.6) (Table 3). There was no significant difference in infarct transmural extent between the groups at any time point (p > 0.05 at each time point).

Within the infarct zone, examining strain across all layers showed overall recovery with time (F = 44, P < 0.001). For individual layers, endocardial, mid-myocardial and epicardial strain recovered over the 4 time points (p ≤ 0.01 for each) for patients without MO or IMH. For patients with MO, regardless of the presence of IMH, there was no significant recovery of endocardial (F = 3.1, p = 0.05; F = 1.4, p = 0.3 in the absence and presence of IMH, respectively) or mid-myocardial strain (F = 2.1, p = 0.1; F = 2.9, p = 0.05 respectively), but epicardial strain recovered significantly (F = 7.7, p < 0.01; F = 3.3, p = 0.03 respectively; Figure 3).

Analysis of individual time points showed differences between infarcts with MO and IMH evolving over time. At day 2, there was no significant difference in infarct strain in endocardial, mid-myocardium or epicardial zones according to the presence of MO or IMH (F = 0.27, p = 0.8; F = 0.27, p = 0.8; F = 1.9, p = 0.2 for endocardial, mid-myocardial and epicardial strain, respectively). By day 7, there was significant difference in endocardial and mid-myocardial strain between the groups, but not epicardial strain (F = 6.9, p = 0.03; F = 4.3, p = 0.02; F = 1.9, p = 0.2 respectively). At day 30 and day 90, there were significant differences in endocardial, mid-myocardial and epicardial strain according to the presence of MO or IMH (day 30: F = 3.7, p = 0.03; F = 4.5, p = 0.02; F = 5.5, p < 0.01; day 90: F = 7.9, p < 0.01; F = 11, p < 0.01. F = 10, p < 0.01 respectively).

Remote myocardial strain was similar over time, and similar at each time point (p ≥ 0.2 for all) regardless of infarct characteristics (Figure 3).

At each time point, infarct zone endocardial strain was not associated with infarct transmural extent (F = 1.1, p = 0.3; F = 0.3, p = 0.8, F = 1.1, p = 0.3; F = 2.3, p = 0.1 for days 2, 7, 30 and 90 respectively). Endocardial strain was chosen as it was consistently within the infarct zone.

Univariable linear regression analysis (examining the variables in Table 4) showed that presence of MO, presence of hemorrhage and total infarct volume, but not infarct transmurality were significantly associated with decreased strain in the infarct zone at 90 days. Of these, the presence of MO and/or IMH, but not infarct volume, were significantly associated with strain on multivariable logistic regression analysis (Table 4).

This study has limitations. The number of patients is relatively small, though sufficient to generate significant results and in keeping with other CMR studies in this demographic, where serial imaging including early post AMI is challenging. Five patients did not have strain and T2w imaging at baseline. It is possible that any MO/IMH may have resolved by day 7, although of 22 patients with MO at day 2, all but one had MO at day 7. We used a dual-echo T2* method in this study. Multi-echo techniques permit more reliable quantitative estimates of T2* but were not available to us at the time of this study. For this reason, we only performed a qualitative assessment of IMH size from T2* images. A variety of cut-offs to define LGE enhanced myocardium have been proposed in the literature; we chose 2SD to ensure consistency with T2 image analysis. Using concurrent LGE images for tagging at each time point may be affected by geometric changes in the LV or infarct zone for follow up scans, and a potential change in the position of myocardium that is sampled. Resolving three circumferential layers in remodeled infarcted myocardium may have been limited by the resolution of CMR, though the harmonic phase analysis software is able to produce a displacement map between tagging intersections to <1 mm [31], and the infarct zone did not thin in the vast majority of patients with timely reperfusion. End systolic strain may not account for infarct zone dyssynchrony, though this was not observed visually in any patient.