Assessment of myocardial injury after reperfused infarction by T1ρ cardiovascular magnetic resonance

Ten pigs underwent 90 min of coronary artery occlusion and were randomized into two groups: group I, a 1-week post-MI terminal study (n = 5) and II, a 4-week terminal study (n = 5). Furthermore, five animals (group II; n = 5) underwent a baseline CMR immediately prior to coronary artery occlusion. The experimental model was chosen to emulate human post-percutaneous coronary intervention ischemia injury. All the animals of the 4 week group underwent both 1 and 4 week CMR. Figure 1 depicts an overview of the experimental protocol.

A prophylactic antiarrhythmic regimen of 150 mg amiodarone, 1 mg/kg lidocaine and 1 g magnesium sulfate was administered intravenously. A left thoracotomy was performed and the pericardium was opened. One or more coronary snares were positioned at the branches of the circumflex artery to induce an infarction of approximately 20% LV size. The dimensions of the infarction were determined by visible color changes at the epicardium. The exact position of the ligation sutures were decided upon after gross inspection of arterial anatomy, unique to the animal. Ischemia was confirmed by visible color changes in the ischemic region and ST segment elevations on ECG. After 90 min of ischemia, the coronary snare was removed, an inter-costal nerve block was performed with bupivacaine at the surgical site, and the chest was closed in muscle layers.

CMR studies were performed on a 1.5 T whole-body system (Avanto; Siemens Healthcare, Erlangen, Germany) with 40 mT/m gradient and 12 channel RF receiver arrays. Intraventricular pressure was interfaced to physiological monitoring software and filtered to facilitate dual respiratory and cardiac gating (LabView, National Instruments, Inc., Austin Texas). All 2D images were acquired in the short-axis during breath-holding and 3D with dual cardiac and respiratory gating. Breath-holding was performed by temporarily disabling the animal ventilator.

2D T1ρ single-shot balanced steady-state free precession (bSSFP) sequences were performed using a spin echo, spin lock (SL) T1ρ pulse cluster (90_x – SL_y - 180_y – SL_-y - 90_-x) [8, 14, 15]. T1ρ images were acquired with different TSL times using the following parameters: TSL = 2, 10, 18, 26, 34, 42, 50 ms, B₁ = 500 Hz, spatial resolution = 1.4 × 1.4 mm², slice thickness = 6 mm, flip angle = 70°, TE = 1.45 ms, TR = 2.9 ms, N_Seg = 55, bandwidth = 900 Hz/pixel, linear k-space phase encoding ordering, parallel imaging with acceleration factor = 2, 34 reference k-space lines obtained in a separate heartbeat, and four heartbeats for T1 relaxation between shots. The T1ρ amplitude was set at the highest available within scanner specific absorption rate limits (B₁ = 500 Hz).

2D T2 maps were obtained using a single-shot T2 prepared (90_x – 180_y – 90_-x) bSSFP sequence and 8 images were acquired with different contrast times TE = 2, 10, 18, 26, 34, 42, 50 ms using the same readout as with the T1ρ images.

2D T1 maps were obtained with a modified Look-Locker sequence, utilizing a single-shot acquisition with eight inversion times (two inversion pulses: five images obtained after inversion 1, 10 T1 recovery beats, and three images obtained after inversion two) [16]. Other parameters were: spatial resolution = 1.4 × 1.4 mm², slice thickness = 6 mm, flip angle = 35°, TE = 1.2, TR = 2.4 ms, N_Seg = 57, bandwidth = 1080 Hz/pixel, linear k-space encoding, parallel imaging acceleration factor = 2, 34 reference k-space lines obtained in a separate heartbeat.

Retrospective, short axis, multi-slice cine CMR was performed with a temporal resolution = 40 ms, flip angle = 70°, bandwidth = 940 Hz/pixel, spatial resolution = 1.1 × 1.1 mm², slice thickness = 6 mm.

The animals received a 0.1 mmol/kg intravenous injection of gadolinium contrast for LGE imaging (MultiHance; Bracco Diagnostics, Inc; Princeton, NJ). Imaging was performed 10 minutes after injection of contrast agent using an inversion time (TI) scout sequence to determine the inversion time to null myocardial tissue signal. LGE CMR was obtained using a 3D multishot phase-sensitive inversion recovery (PSIR) bSSFP sequence at spatial resolution = 1.2 × 1.2 mm², flip angle = 50°, TE = 1.6 ms, TR = 3.2 ms. slice thickness = 2 mm, and parallel imaging acceleration factor = 2 [17].

After ex vivo CMR, the heart was flash-frozen using liquid nitrogen in a 4.5 L cryogenic container (Fisher Scientific, Waltham, MA). The heart was sectioned into uniformly thick short axis slices using a commercial-grade prosciutto slicer (330 M; Berkel) obtaining slices of uniform thickness of approximately 2 mm to be matched to CMR studies. Slices were submerged in a PBS solution with 0.1 M triphenyl tetrazolium chloride (TTC) and incubated at 50 ° C for 15 min [18]. Slices were removed from TTC solution and mounted on glass slides with aqueous mounting media (Aquatex; EMD Millipore). TTC stained viable tissue deep red, distinguishing scar tissue and viable myocardium. Slides were imaged at 800 dpi resolution using an optical scanner (Perfection V700; Epson). Tissue sections from remote and infarct regions were selected for further histological analysis and fixed in 10% neutral buffered formalin. Tissue samples were stained with Masson’s trichrome. Collagen was stained blue whereas keratin, muscle fibers, and cytoplasm were stained red and nuclei dark red.

Cine CMR image series were used to calculate indexed LV mass (Mass), wall thickness (WT), end-diastolic volume (EDV), end-systolic volume (ESV), ejection fraction (EF) and cardiac output (CO). Epi- and endocardial contours were drawn manually at ED and ES (excluding papillary muscles) using standard techniques (Qmass 7.5, Medis, Leiden, The Netherlands) [19, 20].

For all images, scar size was assessed using full width at half maximum (FWHM) thresholding in a mid-LV slice with visible enhancement on LGE (QMass) [20]. FWHM thresholds were determined by drawing two regions-of-interest (ROIs): one in an area of non-enhancement and normal wall motion and a second around the hyperintense myocardium and used to define the maximum signal for FWHM threshold. ROIs were not adjusted to include hypointense regions of hemorrhage. Residual blood pool or pericardial regions of high brightness were manually removed from the ROI. Scar size was reported as the ratio (%) of the infarct ROI volumes and remote myocardium volume.

Scar transmurality was computed as the ratio of hyperintense (infarct) to non-enhanced myocardium (%) in a mid-ventricular short axis slice. The myocardium was divided into six circumferential wedges and transmurality was reported for anterolateral and posterolateral segments in which infarction was observed (two of six segments). Scar transmurality <5% in a segment was excluded as noise.

Motion correction was used to align T1ρ and T2 images using optical flow estimation of the image deformations [8]. T1 and T2 mapping was performed using a 3-parameter modelwhere A and B are additional free parameters, and \( \mathrm{T}\mathrm{S}\mathrm{L} \) the contrast evolution time (spin lock time) (MatLab, Natick, MA). Infarct T1ρ, T2 and native T1 (T1 in the absence of gadolinium contrast agent) were calculated from the thesholded region-of-interest (ROI) and remote or baseline myocardium from a normal region (QMass).

Contrast-to-noise ratio (CNR) was calculated for T1ρ, T2 and native T1:where ∆TX=TX _inf –TX _rem and TX _inf (TX _rem ) was the mean relaxation time in the infarcted (remote) tissue and σ(TX _rem ) was the standard deviation of the relaxation times observed in remote myocardium [21].

Descriptive statistics were reported as mean ± standard deviation (SD). Comparison of statistical means was performed using 2-way analysis of variance. Correlations were assessed with Pearson’s r. Bland-Altman testing was performed to test for mean bias and variation between imaging methods [22]. Pairwise infarct size changes of CMR measurements were assessed with 1-way analysis of variance. Histological infarct size changes were assessed with Student’s t-test.

All pigs (s = 10) survived the 90 min ischemia and reperfusion study to their terminal CMR at 1 or 4 weeks. At 4 weeks post-MI, there was a 13.4 ± 5.4% reduction in EF compared to baseline and associated with a 22.7 ± 8.0 mL increase in ESV and a 24.2 ± 14.1 mL increase in EDV. End diastolic wall thickness decreased by 4.0% after 4 weeks (p < 0.05) and ΔWT and ΔLVEDV between baseline and 4 weeks were correlated (r = 0.90, p < 0.05). Additional details are reported in Table 1.

Hyperintense regions were observed at the lateral wall on all T1ρ, T2, and T1 maps and in LGE images at 1 and 4 weeks, consistent with ischemic injury to the circumflex coronary circulation (Fig. 2; full LV coverage shown in Additional file 1: Figure S1).

As illustrated in Fig. 3, infarct T1ρ and T2 relaxation times were significantly increased at 1 and 4 weeks in infarcted myocardium in comparison to healthy myocardium prior to infarction with T1ρ = 50.9 ± 5.5 (baseline), 125.2 ± 12.1 (1 week) and 124.3 ± 15.6 ms (4 week) and T2 = 40.5 ± 7.3 (baseline), 92.1 ± 1.4 (1 week) and 99.6 ± 12.6 ms (4 week). T1 data was not acquired at baseline. T1ρ, T2, or T1 times were not significantly different from 1 to 4 weeks in the infarcted regions. In addition, there was no difference between remote myocardium at 1 or 4 weeks and healthy myocardium at baseline. ∆T1 (infarct-remote) = 421.3 ± 108.8 (1 week) and 480.0 ± 33.2 ms (4 week), ∆T1ρ = 68.1 ± 11.6 and 74.3 ± 14.2 ms, and ∆T2 = 51.0 ± 10.1 and 59.2 ± 11.4 ms. CNR_T1ρ = 12.0 ± 6.2 (1 week) and 12.3 ± 3.2 (4 week), CNR_T2 = 8.0 ± 3.6 and 10.3 ± 5.8 and CNR_T1 = 7.0 ± 3.5 and 6.9 ± 2.4.

Infarct size was not significantly different for T1ρ, T1 and T2 compared to LGE (ss = 0.14) and, among animals who underwent both 1 and 4 week CMR, significantly decreased from 1 to 4 weeks (s < 0.01). Individual infarct size changes were ΔT1ρ = −3.8%, ΔT1 = −3.5% and ΔLGE = −2.8% from 1 to 4 weeks (Table 2). No change in infarct size was observed for T2.

There was excellent correlation of infarct size between T1ρ and LGE (r = 0.98, p < 0.001), good correlation for T1 and LGE (r = 0.79, p < 0.01) and poor correlation for T2 and LGE (r = 0.26, p = 0.44) (Fig. 4). Infarct size mean bias between T1ρ and LGE = 0.3, T2 and LGE = −0.4 and T1 and LGE = 2.3.

Infarct size at histology was 8.9 ± 2.3% at 1 week and 10.1 ± 4.1% at 4 weeks. There was no significant difference in infarct size at 1 or 4 weeks or as compared to any CMR measurement (p < 0.41; Fig. 5).

Infarct transmurality was not significantly different for T1ρ, T1 and T2 compared to LGE (s = 0.7) and was unchanged from 1 to 4 weeks (p = 0.15).

There was good correlation of infarct transmurality between T1ρ and LGE (r = 0.94; p < 0.001), T1 and LGE (r = 0.86; p < 0.001) and moderate correspondence between T2 and LGE (r = 0.86; p = 0.02) (Fig. 6).

T1ρ is increased in acute and chronic myocardial infarction [6, 7, 11, 13], but little is known about the evolution of T1ρ post-infarction and its relationship with scar size. T1ρ progressively increased in non-reperfused MI in mice at 1, 3 and 7 days post-infarction, but remained constant from 7 – 20 days [13]. We did not examine the earliest times post-infarction (<24 h), however, we found significant T1ρ, T2 and native T1 changes at 1 week post-MI and confirmed histologically that these changes corresponded to early scar formation and extensive collagen deposition. This suggests that endogenous contrast changes at 1 week may be in part associated with formation of collagen, increased water mobility in collagen, and not necessarily increased myocardial water content in edema.

Furthermore, while T2 and native T1 are believed to be biomarkers for myocardial edema in the area-at-risk after acute MI [2, 23], recent studies have more closely examined this paradigm, suggesting that (at least) T2 has complex temporal behavior early post-infarction. Fernandez-Jimenez, et al. and Carrick, et al. reported two waves of edema with histological quantification of myocardial water content and associated it with bimodal T2 evolution at <24 h and at 1 week in pigs [24, 25]. Kim, et al. also reported in canines that T2 was not sensitive to the area-at-risk at 3–5 days week post-infaction [26]. While we did not directly measure area-at-risk using an approach such as injected microspheres, there was no evidence that any endogenous contrast was associated with at-risk myocardium at 1 week (was not larger than LGE infarct size). Furthermore, 90 min of ischemia-reperfusion in this animal model would generate significant myocardial salvage (between 5–15% salvage) [27], which was not observed by inspection of differences between endogenous contrast and LGE infarct size.

T2 had reduced correlation with LGE as compared to T1ρ and T1, which may be partly explained by differences in the way post-infarction hemorrhage affects endogenous contrast. Degradation of hemoglobin byproducts in the hemorrhage in T2 and T2* CMR studies in patients and animals contribute to increased magnetic susceptibility-induced dephasing on CMR [21, 28]. Accurate infarct size measurement should manually correct for hemorrhage, which is not labeled properly by semi-automatic thresholding. In the case of T1ρ CMR, the spin locking pulse may mitigate the effects of magnetic susceptibility-induced dephasing of the transverse CMR signal. Thus T1ρ may more accurately reflect total infarct size on semi-automated thresholding. Manual correction may introduce bias and inadvertently label remote or salvaged myocardium, so we did not manually correct semi-automated thresholding.

Another reason for the poor T2 agreement with LGE is that different methods of preparation may give different T2, e.g. preparations with adiabatic refocusing pulses instead of a non-selective Hahn spin echo [29]. The frequency and amplitude modulation of the adiabatic refocusing pulse, refocusing pulse spacing, and the peak amplitude of the adiabatic pulse may improve T2 results by partly suppressing hemorrhage-induced signal dephasing in a similar fashion to T1ρ [30, 31]. In addition, T1ρ is higher than T2 in almost all situations in which it has been measured in human tissues, including the myocardium [9, 11, 13]. This is attributed to the T1ρ dispersion, which is the variation in T1ρ with the preparatory pulse amplitude B₁. As the amplitude B₁ approaches zero, T1ρ approaches T2 (with the additional assumption that there is a refocusing pulse between the SL pulses). In our comparison of T1ρ, T2 and T1, we used identical readout sequences to eliminate a major cause of measurement variation. In addition, we matched the time between adjacent T2 or T1ρ preparations (4 heartbeats) and discarded the first scan because of variations in T1 recovery between preparations.

Infarct size is expected to decrease from sub-acute to chronic post-infarction times [32]. T1ρ, T1 and LGE showed a decrease in infarct size from 1 –s 4 weeks, but a similar change was not observed histologically. This is likely explained by the small number of animals and the absence of histological data from the same animal at 1 and 4 weeks. The use of serial imaging of the same animal and measurement of infarct at the same mid-ventricular position in vivo, likely permitted this detection by in vivo mapping. Yorkshire swine have well-characterized coronary artery anatomy of a similar size and distribution to adult humans. While there is considerable variability in the size, orientation and vascular bed associated with the branches of the circumflex artery in pigs, the overall infarction size was prospectively determined by direct inspection of the circumflex and branch coronary artery distribution at the time of surgery. Therefore the pig infarction model produced reproducible infarct size and spatial distribution and localized to the inferior posterolateral wall.

We found that T1ρ and T1 infarct transmurality was correlated with LGE CMR at 1 and 4 week post-MI. However, differences in T1ρ infarct transmurality might be explained by the varying acquisition times in the cardiac cycle. Although infarct size is not influenced by cardiac cycle, scar transmurality (derived from LGE CMR) has shown to vary between end-diastolic and end-systolic assessment [33]. LGE MR images were derived during end-diastole, however T1ρ images were derived during end-systole, which is a plausible explanation for the differences in scar transmurality.