Our experimental protocol was approved by the Institutional Animal Care and Use Committee and was performed in compliance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.
Canine preparation
Male hounds (18–20 kg, n = 6) were initially anesthetized with an intravenously administered mixture of ketamine (5.0 mg/kg) and diazepam (0.5 mg/kg). Following intubation, animals were mechanically ventilated (Model 2000, Hallowell EMC, Pittsfield MA, USA) and anesthesia was maintained by continuous administration of isoflurane (2.5–3% volume/volume), with intravenous fentanyl (50–100 μg every 30 min) used for analgesia. Heart rate and blood oxygen saturation were monitored and an electrocardiogram was recorded. The right femoral artery was surgically prepared and cannulated using a 6-F arterial sheath (Pinnacle, Terumo Medical Corporation, Elkton, MD, USA). Heparin was administered as needed to maintain a blood activated clotting time above 300 s. A 6-F coronary guide catheter (RunWay Kimny Mini, Boston Scientific, Natick, MA, USA) was introduced to cannulate the ostium of the left main coronary artery, and initial coronary angiography (Philips BV Pulsera, Philips Healthcare, Best, The Netherlands) was performed. A 2- to 4-mm angioplasty balloon (Maverick, Boston Scientific, Natick, MA, USA) was introduced into the left anterior descending or the left circumflex coronary artery over a guide wire, inflated, and left in position for 180 min to produce MI. After removing the balloon catheter, a second coronary angiography confirmed reperfusion. The femoral artery was decannulated, surgically ligated, and the wound was closed.
MRI protocol
The dogs underwent cardiac MRI on a 1.5-T scanner (Signa Horizon CV/i, GE Healthcare, Milwaukee, WI, USA) before the induced MI and at various time points after MI. Four dogs were imaged and sacrificed 4 days after reperfusion to compare in vivo MRI findings in the acute phase of MI to same-day histology. Two dogs were monitored for 8 weeks to follow the evolution of MI with MRIs on days 3, 6, 14, and 56. For image acquisition, animals were anesthetetized and mechanically ventilated, as described above. Imaging was performed during breath-hold at end-inspiration. Native T2 maps and LGE images were generated for the purpose of in vivo tissue characterization at each time point. A flowchart of the protocol is shown in Fig. 1.
T2 mapping and T2-weighted imaging
Double inversion-recovery (black-blood) fast-spin-echo images were generated during breath-holds with varying echo times (TE). Acquisition was timed to the same end-diastolic phase of the cardiac cycle with all TEs. Six short-axis images were generated covering the entire LV with the following parameters: field of view 300 mm, image matrix 256 × 256, slice thickness 10 mm, flip angle 90°, echo train length 24, and TE 12, 20, 30, 45, 60, 75, 90, and 120 ms. The image obtained with a 60-ms TE was used as the T2-weighted image [19].
LGE imaging
At day 4 in the acute phase of MI, LGE imaging was performed 48 h after the administration of 0.05 mmol/kg of Gd-N-(2-butyryloxyethyl)-N’-(2-ethyloxyethyl)-N,N’-bis[N’’,N’’-bis(carboxymethyl)acetamido]-1,2-ethanediamine (Gd-ABE-DTTA), a persistent contrast agent [20, 21]. To perform LGE imaging at the other time points, 0.2 mmol/kg gadopentetate dimeglumine (Gd-DTPA, Magnevist®, Bayer HealthCare Pharmaceuticals Inc., Wayne, NJ, USA) was administered and LGE acquisition was performed 12 min after the administration of the contrast agent. LGE images were acquired using a 180°-prepared, segmented, fast gradient-echo pulse sequence in every other cardiac cycle. The inversion time was set to the optimal value to null the signal in the healthy myocardium. Identical slice orientations and positions were used for LGE and T2 maps to enable accurate co-registration of images generated by the two methods.
Image analysis
MR images were converted to text images and analysed in ImageJ (Wayne Rasband, National Institutes of Health). Calculations were performed in either Microsoft Excel or ImageJ. Colour coding was processed in ImageJ.
Percent edema map
Voxel-wise R2 was calculated based on the T2 maps from the TE dependence of the signal intensity (SI) by means of a two-parameter, least-squares curve-fitting routine using the following formulas:
$$ \mathrm{SI}={\mathrm{SI}}_0\times {\mathrm{e}}^{\left(-\mathrm{TE}\times \frac{1}{\mathrm{T}2}\right)} $$
$$ \mathrm{R}2=\frac{1}{\mathrm{T}2} $$
where SI is the signal intensity and SI0 is the SI at the theoretical TE = 0 ms time point, also representing maximum SI.
Since post-MI myocardial R2 is in a linear relationship with the dry-to-wet weight ratio (DWR) [22], changes in tissue R2 (∆R2) can be interpreted as changes in tissue water content. ∆R2 attributable to tissue changes was calculated as follows:
$$ \Delta \mathrm{R}{2}_v=\mathrm{R}{2}_0-\mathrm{R}{2}_v $$
where R20 is the R2 measured in healthy myocardium and R2
v
is the observed R2 in any given myocardial voxel v. Thus, in healthy myocardium ∆R2
v
= 0, in ME regions ∆R2
v
> 0, and in regions where the water content is decreased (mature scar), ∆R2
v
< 0. To determine a universal percent edema (PE) scale in terms of ∆R2v values, the value PE = 0% was arbitrarily assigned to the healthy myocardium, where ∆R2v = 0, corresponding to R20 = 18.7±1.2 s–1 based on the control R2 maps generated prior to intervention. The corresponding DWR in this normal myocardium was 0.23±0.01. PE = 100% was assigned to the ∆R2 of pure water, which, by definition, corresponds to a DWR of zero. Based on the R2 of pure water (R2H2O = 0.27 s–1), ∆R2 of pure water can be calculated as follows:
$$ \Delta \mathrm{R}{2}_{H2O}=\mathrm{R}{2}_0-\mathrm{R}{2}_{H2O}=18.7{\mathrm{s}}^{-1}-0.27{\mathrm{s}}^{-1}=18.43{\mathrm{s}}^{-1} $$
This value is the change in R2 when proceeding from healthy myocardium (DWR = 0.23) to pure water (DWR = 0). This ∆R2H2O corresponds to the entire theoretical range of change in water content covering the corresponding range of PE from PE = 0% to PE = 100%, and thus is used to convert any observed R2 to the corresponding PE value. To calculate the PE values for all myocardial voxels, ∆R2
v
was calculated for each voxel v. In this manner, the R2 map was transformed into a ∆R2 map. ∆R2
v
was considered zero in all voxels where R2
v
was within the range of R20±2 standard deviations. Subsequently, voxel-wise PE
v
values were calculated for the entire slice (percent edema per slice, PES) to generate a PEM as follows:
$$ {\mathrm{PE}}_v=\left(\Delta \mathrm{R}{2}_v\div \Delta \mathrm{R}{2}_{H2O}\right)\times 100 $$
LGE evaluation
Endo- and epicardial contours were manually traced, and the remote myocardium was selected using a region of interest. Further steps were automated to avoid observer bias. MI was defined as pixels that displayed an SI value above the mean SI of the remote myocardium plus sixfold the standard deviation (6 SD) [23]. MI pixels were counted, and the ratio of infarcted to total myocardial area of a given slice (percent infarct per slice, PIS) was subsequently calculated.
Tissue characterization map
Voxel-wise TCMs were generated by a computer routine based on criteria (Fig. 2) that combine the information from PEM and LGE images. Tissue characterization was based on the presence or absence of ME, or the presence of “negative ME”, while also taking into account voxel enhancement in the LGE image. Importantly, while “PE = 0” in healthy myocardium indicates the absence of ME, ME is present in hemorrhagic MI. Nevertheless, ME is not seen in the PEM of such hemorrhagic MI due to the cancelling effect of ∆R2 induced by methemoglobin. Since MH occurs in the center of the MI and is enhanced in LGE, it can be differentiated from healthy myocardium and quantified (percent hemorrhage per slice, PHS) using our method.
Histopathology
Triphenyl tetrazolium chloride (TTC) staining was performed in vivo prior to inducing cardiac arrest [24]. A solution of 12.5 ml/kg of 2% TTC saline was intravenously administered and maintained in the circulation for at least 20 min [24, 25]. The animal was then euthanized with a high dose of pentobarbital followed by 100 ml of 2M potassium chloride solution. The heart was excised, rinsed with saline, and sliced in 3-mm increments using a commercial meat slicer. Both sides of each TTC slice were photographed using an Olympus C 4000 Zoom Digital camera. TTC photograph analysis was adapted from work by Ruifrok et al., in which the three colour channels (red, green, and blue) were separated in ImageJ [26]. After splitting the three channels (Fig. 3), the red channel was displayed as a greyscale image, where viable tissue is shown as dark grey and irreversibly injured regions are bright. MI borders were traced using these images. To highlight MH selectively, the blue channel image was displayed in the red channel and merged with the green channel, resulting in a composite image where hemorrhagic regions appear as light brown within the green-yellow non-hemorrhagic region. This method was further validated with hematoxylin-eosin and Prussian blue microscopic histology.
The PIS and PHS were calculated and compared with MRI data by adding MI and MH areas in three TTC slices that corresponded to one MRI slice. Values were expressed as a percentage of total LV myocardial area in that slice.
Statistical analysis
Statistical analysis was performed using SigmaStat (version 2.03; SPSS, Inc.). A normality test was performed to determine whether the sample had a Gaussian distribution. Data with a normal distribution are reported as mean±SD. Data with a non-normal distribution are reported as median [25th, 75th percentiles]. Student’s t test was used to compare data with normal distribution and equal variance, while the non-parametric Wilcoxon test was employed for non-normal distributions. Pearson’s correlation and Bland-Altman analyses were performed to compare the PIS and PHS from MRI to those obtained with TTC staining. Overestimations of results from TCM were calculated for each slice and each dog, using the TTC results as reference. Rejecting the null hypothesis at α = 0.05 with a p value < 0.05 indicated statistical significance.