Ischaemia-reperfusion dynamics in acute myocardial infarction experimental swine model: new insights from quantitative CMR

Abstract

Aims

To reveal the pattern and dynamics of myocardial oedema induced by myocardial infarction (MI) during ischaemia and subsequent reperfusion, that remain largely unknown, as are the factors that contribute to reperfusion injury. To propose a time-resolved dynamic myocardial tissue characterization by quantitative CMR, with T1&T2 mapping and Pixel-wise standardized analysis to circumvent inter-animal differences and subjective ROI positioning.

Methods and results

We measured T1&T2 relaxation times at baseline, during a 40-min transient coronary occlusion, and after reperfusion in an open-chest swine MI model (n = 20; 2 shams) using MRI. Myocardial function, early and late gadolinium enhancement were also assessed. Pixel-wise standardized analysis was used to compare the image contents at each pixel across individuals and time points. A significant increase in cardiac T1&T2 times in the ischaemic regions occurred during ischaemia compared with baseline (mean ΔT1 = 118.8 ms i.e. + 11.1%, ΔT2 = 5.6 ms i.e. + 11.3%; P < 0.05). A global significant and marked increase in T1&T2 times further appeared immediately after reperfusion (mean ΔT1 = 256.8 ms i.e. + 23.3% mean ΔT2 = 11.9 ms i.e. + 23.6%, P < 0.001). This increase was associated with myocardial wall thickness changes, with regional and global dysfunction in the ischaemic myocardium. Three different reperfusion patterns were differentiated by the pixel-wise T1 signal analysis: effective reperfusion with microvascular obstruction (MVO), effective reperfusion without MVO and absence of effective reperfusion. We found no correlations between baseline, per-ischaemia, and post-reperfusion native T1&T2 times when effective reperfusion occurred.

Conclusion

Objective quantification of tissue response by pixel-wise analysis demonstrated rapid and significant changes in myocardial water content status post-reperfusion, with three different early-reperfusion patterns observed, suggesting distinct reperfusion mechanisms. The water content after reperfusion does not reflect its state before and it does not provide insight into the final tissue status observed within 3 h after recanalization.

Graphical Abstract

Graphical Abstract.

Introduction

Myocardial infarction (MI), defined as an irreversible myocardium injury resulting from prolonged ischaemia, is a complex phenomenon.¹ The last decades have seen substantial advances in the treatment of ischaemic heart diseases due to increasing knowledge of the molecular mechanisms involved during ischaemia-reperfusion (IR) injury. The rapid myocardial tissue changes have been documented using of cardiac magnetic resonance (CMR). The abrupt reperfusion often results in additional tissue injury, with frequent (∼40%) microvascular obstruction (MVO) lesions (no reflow/haemorrhage) whose determinants are still unclearly identified.^2-4 This reperfusion injury leads to additional inflammation signed by significant myocardial oedema and tissue infiltration^5–7. The oedema in ischaemic myocardium appears early during ischaemia, with a significant increase after reperfusion, remaining stable during the first week after myocardial/infarction. Recent experimental work using T2 mapping has refined our knowledge and revealed that oedema follows a bimodal pattern over the first week instead, with an initial wave appearing upon reperfusion while dissipating at 24 h.⁶ Since the pioneering experimental work by Abdel-Aty et al.^8,9 which examined the changes in ischaemic myocardial tissue using T2-weighted CMR at a very early stage, no further studies have focused on precise tissue quantification and characterization of this time frame. There is currently no data available on the respective evolution of T1, T2, and extra-cellular-volume (ECV) in the first hours of ischaemia reperfusion.

The main objective of our study was to monitor myocardial tissue changes at the hyperacute stage of myocardial infarction and reperfusion with a non-invasive multi-parametric and quantitative MR protocol. This was performed with user-independent advanced post-processing and statistical assessment up to the pixel level.

Methods

Experimental animal procedures and model

An open-chest swine experimental model of ischaemia-reperfusion (IR) was used. Animal experiments were approved by the State Committee on Animal Affairs of our Institution in agreement with the 2010/63/EU Directive from the European Parliament. We included a total of 26 swine with body weights ranging between 30 and 37 kg (24 animals underwent the IR procedure, 2 served as shams).

Figure 1 reports the study timeline with the MR protocol and experimental procedures.

Figure 1.

Chronogram summarizing the multi-steps CMR protocol and acute ischaemia model procedures. The coloured spheres mark the positioning of the microspheres’ injections: Gold (Gd: @5–10 min before occlusion), Samarium (Sa: @35–40 min of ischaemia), Lutetium (Lu: @5 min after reperfusion) and Lanthanum (La: @150 min of reperfusion).

Animal preparation. The animals were initially premedicated with intramuscular administration of ketamine (15 mg/kg), xylazine (2 mg/kg), and acepromazine (0.3 mg/kg). Following placement of continuous monitoring for heart rate, electrocardiogram (ECG), respiration, oxygen saturation at end tidal CO₂ (EtCO₂), propofol (5 mL at 1%), and fentadon (10 µg/kg) were injected intravenously. Each pig was initially intubated and mechanically ventilated. Central arterial and venous catheters were placed for continuous low perfusion (5–10 mL/h) of a mixture of ketamine, xylazine, and fentadon, given to avoid any painful perception of the animal. The body temperature, heart rate and arterial pressure, oxygen saturation, and EtCO₂ were monitored continuously during the protocol. The chest was opened by surgical median sternotomy, allowing direct access to the heart. At the end of the protocol, the animals were sacrificed without awakening by injection of T61 euthanizing agent (0.3 mL/kg).

Left anterior descending (LAD) coronary artery occlusion. Acute reperfused MI was induced via coronary artery ligation of the LAD¹⁰ at its most proximal possible position to induce a large area at risk (AAR). Two vessel loops were positioned below the dissected LAD artery. They stayed in place during the whole procedure, including the pre-occlusion period, and were tied up only during the occlusion period of 40 min for each animal. The complete but reversible vessel occlusion was verified visually by immediate bluing of the artery bed and immediate akinesia. Observed ECG changes attesting to infarction were observed and registered. Digital pictures of the open-chest swine’s hearts at each point were recorded for quality control (see supplementary material S1) and stable-isotope labelled microspheres (BioPAL, Worchester, MA) were collected for quality control of reperfusion state (see supplementary material S2 & S3).

At reperfusion, the LAD coronary artery patency was considered effective if the artery was visualized as successfully recanalized. Blood samples were taken at the end of the MR session for each animal to measure the haematocrit.

Two sham animals also underwent the complete imaging protocol, without the ischaemia intervention.

Histopathology protocol

An intravenous bolus of 100 mL of Evans Blue dye was administered at the end of the protocol and after re-occlusion of the LAD artery, before the sacrifice, to delineate the AAR for further macroscopic visual measurement. The hearts were then excised and sliced into five to eight sections, with the same thickness, along the short axis, one of the slices (usually the central one) corresponding to the imaged MR short axis. All the slices were weighted, and high-resolution pictures of each slice (on both sides) were acquired with a digital camera. The AAR assessment was further visually performed by manually delineating the unstained myocardial area,¹¹ further compared with the total LV per slice.

Triphenyltetrazolium chloride (TTC) assays were conducted to assess tissue viability and infarct size, with precise timing (>3 h of reperfusion), optimized staining (incubating tissue slices in TTC solution [2% TTC (Sigma-Aldrich) in 0.9 NaCl, at 37°C for 20 min]. Stained myocardial slices were photographed from both sides. Further planimetric area quantification of tissue slices (both sides) was conducted using free-hand tracing within the Image-J software. The infarct area was compared with the total LV.

CMR protocol

CMR was performed in a 1.5T MR system (MAGNETOM Avanto, Siemens Healthineers, Erlangen, Germany), using the standard combination of surface coil (12 elements) for cardiac exploration. Continuous dynamic quantitative MRI exams were performed at the highest possible measurement rate at key time points: prior occlusion (pre-ischaemia), per ischaemia (∼40 min), immediately after reperfusion and at 2 h of reperfusion in order to characterize tissue changes in the myocardium at all stages. At the final imaging time point, 2 h after reperfusion, an intravenous gadolinium injection was performed (Dotarem, Guerbet, France, 0.2 mmol.kg⁻¹, 0.5 M) to evaluate the non-perfused territory, and early and late gadolinium enhancement (EGE and LGE) of the myocardium (Figure 1).

ECG-gated balanced steady-state free precession (bSSFP) cine of long-axis views were first acquired under breath-hold (manual arrest of the ventilator during 15 s). Full stack of short-axis (SA) cine images (covering the whole left ventricle) and two long-axis were acquired before ischaemia and after reperfusion, to provide full function analysis. Additional more basal and apical SA slices, as well as LA slices were also acquired before ischaemia and after reperfusion to fully depict the animal myocardium with T1 and T2 mapping. Then, a short function cine protocol (including one SA and two long axis scans) as well as T1 and T2 mapping (acquired in short-axis mid lesion position) were chosen and repeated continuously at all stages of the acute infarct model. Supplementary data online, Table S1 lists the main MR parameters (Supplementary Appendix).

First-pass ECG-gated perfusion sequence was used to monitor the bolus injection of gadolinium and was followed by EGE and LGE acquired using an inversion recovery 3D-turbo-gradient-echo-technique. EGE imaging was performed immediately after FP perfusion sequence (3.5 ± 0.7 min after Gd bolus injection) and LGE imaging 15.3 ± 0.6 min after Gd injection. EGE and LGE were used to classify the reperfusion patterns based on the existence/extent of a non-reperfused area (EGE), and on the extent of gadolinium enhancement (LGE). Quality control of myocardial perfusion status used stable-isotope labelled microspheres (see supplementary material section S2&3).

Post Gadolinium T1 mapping was also obtained, and the ECV map was calculated using the inline ECV post-processing tool on the MR console. Haematocrit values was obtained by extemporaneous microcentrifugation (Haematokrit 200, Hettich, Germany) with blood drawn immediately after post-gadolinium T1 maps, at the end of the MR procedures, and before the animal was moved for LAD re-occlusion.

Post-processing and image data analysis

Function analysis

Complete function analysis was performed offline for all the animals with Circle Cardiovascular Imaging software (CVI42 version 5.13, Calgary, Canada).

Left ventricle end diastolic volume, end systolic volume, ejection fraction and myocardial mass were assessed at all time points of the protocol.

Strain analysis was conducted using the tissue-tracking module of the CVI42 software applied to the cine images in long and short axes views at different time points of the protocol. Automatic segmentation of 11 segments, including at least one segment in the ischaemic zone and at least 1 in the remote region, allowed a deeper analysis of the tissue characterization. For a regional contraction evaluation, the evaluation was performed on cine images obtained in pre-ischaemia (values considered as baseline), ischaemia, and reperfusion period. The last cine sequence was completed just before the LGE imaging to confirm no enhancement in the segments considered remote.

Analysis of the lesion size at the end of the IR protocol

All lesion size (LS) analyses were performed on the time-resolved mid-ventricular SA T1 and T2 images, LGE images, and the corresponding histological Evans Blue dye slices and expressed in % of the total myocardium. The reference AAR was estimated from Evans Blue dye macroscopic in the ImageJ software (NIH, USA).

For lesion size evaluated on the MR images, the analysis was conducted in CMRSegTools processing freeware (v 1.5.1, CREATIS, Lyon, France: https://www.creatis.insa-lyon.fr/CMRSegTools/), following a semi-automatic procedure, and expressed in grams (g) and as a percentage of total LV mass. The acute IR lesion was measured as the enhanced area on the short-axis late-gadolinium slices acquired at 10 min post-gadolinium injection {SchulzMenger:2020ku}. The % of no-reflow (NR) was calculated from the same images (defined as hypointense areas located within hyperintense areas) and also expressed as a percent of the total MI mass (%). The endocardial surface length (ESL), defined as the length of endocardial infarct scar surface that included >50% of the whole thickness of myocardium,¹² was measured as a surrogate of the AAR on the corresponding CMR images.

Lesion volumes on T1 and T2 maps over time were determined by the same blinded investigator experienced in experimental CMR. After optimal contrast adjustment, the lesions’ edge was traced manually on each map acquired at the most representative portion of the lesion. The areas of hyperintensity were then multiplied by the slice thickness to calculate a lesion volume that enables comparison with histology and facilitates the comparison between techniques and animals.

Compared with histology, LGE, and EGE images, the lesion areas were expressed in % of the total myocardium.

Pixels visibly belonging to no-reflow or haemorrhage (if any) were excluded before deriving the mean T1 and T2 values within these regions of interest (ROI). Control values for each of the T1 and T2 parameters were determined from a ROI drawn in unaffected tissue (remote) in opposed myocardial segment with visually non-altered tissue. Moreover, an ROI was also placed in the cavity to monitor blood values. For each quantitative variable, ratios of values in altered tissue to contralateral remote values were also calculated. The difference and percent difference between territories (lesion and remote) at each time point and the percent enhancement between time points in the same territory (lesion or remote territories) were recorded and analysed.

Pixel-wise processing and temporal analysis of the lesion patterns

To compare the image contents at each pixel across individuals and time points and circumvent the fact that individuals differ in anatomy and pathophysiology, we spatially aligned the image contents of all given acquisitions/individuals to a standard reference.¹³ The reference consists of an arbitrarily designed geometry: two concentric disks with endocardial and epicardial radii of respectively 30 and 50 pixels, represented on an image of 80 × 80 pixels. Spatial correspondences were obtained by defining (both on each individual image and the reference) radial and circumferential coordinates, which ranged from 0 to 1 for endocardium-to-epicardium (radial) and anticlockwise direction starting from the LV-RV junction (circumferential). Spatial alignment was achieved by mapping each individual's coordinates onto the reference coordinates, through linear interpolation tailored for data defined on a scattered grid (here, the myocardial coordinates). No additional regularization, spatial smoothing, or outliers’ removal was involved. This process corresponds to the 2D version of the process first described in [13], for which code is publicly available (https://github.com/nicolasduchateau/CMR-imageDataAlignment).

After this, statistical comparisons were made on the spatially aligned data at each pixel of the reference anatomy to characterize the distribution of values within a given group of individuals. Each animal's lesion zone was taken from the LGE segmentation after transporting it to the reference anatomy. We reported the median value and first/third quartiles for each identified subgroup in each zone (excluding MVO) and separated the statistical analysis for each acquisition time point. Results were displayed as coloured images over the reference anatomy, representing synthetic images at each time point, as shown in Figure 2.

Figure 2.

(Pattern 1: R + MVO+: AMI with effective tissue reperfusion and MVO): non-contrast tissue characterization mapping techniques obtained in the open chest model at all main stages of the IR protocol; Post-Gd T1maps, Early Gd Enhancement (EGE), Late Gd Enhancement (LGE) and cine bSSFP (end diastolic phase) T1 weighted images after reperfusion; Reference macroscopic histology (Evans Blue dye). Bottom images display the whole T1 maps along the open chest MR protocol at baseline (blue frame), during ischaemia (green frame) and post reperfusion (orange frame).

Statistical analysis

The normal distribution of each data subset was checked using graphical methods and a Shapiro–Wilk test. Quantitative variables were reported as mean ± standard deviation (SD) or median and 95% confidence interval [CI 95%]. Statistical analysis was performed using one-factor ANOVA mixed-effects models with the Geisser–Greenhouse correction. Post-hoc analyses used Bonferroni’s correction when testing for differences between the experimental groups (P < 0.05 was considered as statistically significant). All statistical analyses were performed using Prism 10.4.1 (La Jolla, CA, USA).

For pixel-wise analysis, no correction for multiple comparisons was used, as we limited the pixel-wise analysis to descriptive statistics (pixel-wise average) across subgroups of animals, and did not directly use pixel-wise data for statistical comparisons between subgroups. A single acquisition was considered for each subject at each stage of the IR protocol.

Results

Among the 24 animals that underwent the ischaemia-reperfusion procedure, 20 successfully survived the whole protocol and were imaged at all steps, while 4 died before the end of the protocol (2 during ischaemia, 2 at the reperfusion). Digital pictures of the heart at each time point of interest (as shown in supplementary material, Supplementary data online, Figure  S1) allowed visual quality control of optimal occlusion, resulting akinesia, recanalization, Evans Blue opacification.

All animals that survived ischaemia were macroscopically reperfused at the end of the ischaemic period, when the coronary snare was released.

Area at risk, myocardial infarct size, global and regional left ventricular function.

Table 1 shows the mean myocardial area at risk and infarct size as well as the changes in volume, function and T1 times in the ischaemic and remote myocardium during the protocol.

Table 1.

Changes in LV volume, mass, global, regional function (strain), and MR-derived myocardial characteristics during the ischaemia-reperfusion protocol

			Baseline	Ischaemia	Reperfusion
LV mass (at ED) (g)			83.3 ± 10.5	83.3 ± 12.4	88.0 ± 15.1
LV EDV (ml)			53.3 ± 11.0	NA	71.4 ± 25.3.6
LV ESV (ml)			34.2 ± 7.4	NA	45.8 ± 18.6
Ejection fraction LAX (%)			52.1 ± 10.9	29.0 ± 13.6	35.1± 10.7
Heart ratea			87.6 ± 3.1	100.4 ± 2.9	99,2 ± 6.3
RR interval (ms)			687.5 ± 34.4	597.3 ± 27.4	608 ± 38.1
Infarct size (% LV)			—	—	26.2 ± 8.7
ESA ∼Area at risk (%LV)			—	—	27.1% ± 10.7
ESL mid slice (% mid slice)			—	—	42.8% ± 10.9
ECV Remote segments (%)			—	—	26.4 ± 4.7
ECV Ischaemic segments (%) b			—	—	77.6 ± 13.2
	Err	Ischaemic	14.4 ± 9.8	6.8 ± 5	13.1 ± 6
		Remote	18.5 ± 7.8	24.7 ± 10	15.0 ± 6.5
	Ecc	Ischaemic	−9.7 ± 6.1	−5.8 ± 3.7	−9.6 ± 4.6
		Remote	−11.8 ± 5.7	−15.6 ± 4.7	−8.5 ± 5.2
	Ell	Ischaemic	−9.2 ± 5.9	−5.3 ± 4.5	−7.5 ± 5.1
		Remote	−15.2 ± 8.5	−12.4 ± 6.5	−15.4 ± 6.2
	Sham	Err	22.8 ± 1.4	—	21.6 ± 1.7
		Ecc	−14.5 ± 1.5	—	−14.4 ± 0.8
		Ell	−13.6 ± 1.4	—	−12.8 ± 0.6

Descriptive statistics are provided for the 20 animals that survived the whole IR procedure and with imaging completed at all time-points. Data are mean ± SD; LV: left ventricle; LV mass = left ventricle mass (end diastole)

ESA: endocardial surface area; ESL: endocardial surface length; LVEF: left ventricular ejection fraction; LV: left ventricle; EDV: end-diastolic volume: ESV: end-systolic volume; Ecc (circumferential), Ell (longitudinal), Err (radial) are lagrangian strains.

^aHeart rate data were collected every 5 min during occlusion and reperfusion, every minute during surgery; cal 10 measures were performed at baseline during catheter positioning.

^bECV values are given for pattern I and II in Gd enhanced regions only (exclusion of MVO areas). NA = not available.

Compared with baseline, there was an increase in mass, end-diastolic and end-systolic volumes with a reduced ejection fraction at reperfusion in all animals. Regional function in the ischaemic myocardium as assessed with the normal strains (E_cc, E_rr, E_ll) declined significantly during the ischaemia period with a moderate improvement at reperfusion. There was no significant change in longitudinal regional function in the remote myocardium between baseline and reperfusion. Conversely, there was an increase in radial and circumferential function during ischaemia, with a decline of these regional functions upon reperfusion.

The mean acute hyperenhanced lesion size was 26.2 ± 8.7% LV with an estimate (Prussian blue dye) of the area at risk size of 27.1% ± 10.7% LV. There was no significant change in the extra-cellular volume fraction during the entire protocol in the remote segments. There was a significant increase of the ECV in the hyperacute lesion compared with the remote myocardium at reperfusion (77.6 ± 13.2% LV vs. 26.4 ± 4.7% LV, respectively; P < 0.001).

Global T1 and T2 variations in time during ischaemia-reperfusion

All T1 time variations in the ischaemic and remote myocardium are reported in Table 2. First, we found a significant effect of ischaemia on T1 and T2 with substantial variations in time (P < 0.0001). There were significant differences in T1 and T2 between the remote and ischaemic regions (P < 0.0001) (Figure 2–4).

Table 2.

Changes in T1 and T2 values over time in milliseconds at key time points of the ischaemia-reperfusion protocol for each group and/or pooling all animals together

Time points			Baseline	Early Isch.	Late Isch.	Early reperf.	Late reperf.	Post Gd
Ischaemic segments
T1	All Pigs (N = 20)	Mean (ms)	1057.3	1122.1	1155.2	1363.3	1339.0	713.8
		SD (ms)	52.7	48.4	47.8	143.9	160.5	422.6
	No Reperf (N = 9)	Mean (ms)	1062.0	1128.3	1164.3	1337.3	1346.6	1145.0
		SD (ms)	57.0	53.5	54.1	95.4	79.6	104.8
	Reperf MVO + (N = 5)	Mean (ms)	1039.9	1106.8	1143.4	1467.4	1441.4	277.9
		SD (ms)	59.5	64.4	70.5	241.7	317.7	61.0
	Reperf MVO- (N = 6)	Mean (ms)	1039.6	1112.9	1140.9	1262.7	1187.2	319.7
		SD (ms)	27.7	31.1	21.9	160.7	116.1	57.6
T2	All Pigs (N = 20)	Mean (ms)	48.4	50.9	52.2	61.9	60.6
		SD (ms)	2.9	4.1	4.2	8.2	7.7
	No Reperf (N = 9)	Mean (ms)	48.9	51.1	51.3	56.9	55.7
		SD (ms)	3.7	5.7	4.0	4.9	4.6
	Reperf MVO + (N = 5)	Mean (ms)	49.7	49.9	51.2	68.3	67.7
		SD (ms)	2.1	1.7	2.1	8.4	8.7
	Reperf MVO- (N = 6)	Mean (ms)	46.6	51.6	55.2	64.8	63.0
		SD (ms)	1.2	3.1	5.6	10.2	8.0
Remote segments (N  =  20)
T1	Mean (ms)		1041.9	1038.4	1044.4	1065.8	1063.8	510.0
	SD (ms)		42.6	25.5	24.3	28.2	43.5	105.9
T2	Mean (ms)		48.6	49.1	49.1	49.1	48.6
	SD (ms)		3.2	2.3	2.3	2.7	2.6

Data are mean ± standard deviation (SD) for animals (n = 20) that underwent the whole IR procedure with imaging at all steps.

Figure 4.

(Pattern 3: No Reperf (R-) : AMI without effective tissue reperfusion as documented after gadolinium bolus injection, by the lack of Gd enhancement of the lesion, on both cine post-Gd, early post-Gd imaging at 2 min (EGE), and late post-Gd imaging (10–15 min after injection (LGE). Non-contrast tissue characterization mapping techniques obtained in the open chest model at all main stages of the infarct protocol at baseline, during ischaemia and post reperfusion; post-Gd T1maps, dark blood STIR TSE, LGE and cine bSSFP (end diastolic phase) T1-weighted images after reperfusion T1 maps post reperfusion and reference macroscopic histology (Evans Blue dye).

A noticeable T1 increase was observed during ischaemia in the lesion (mean difference[95CI%])= + 63.5 [23.5;103.6]ms (P = 0.0019), followed by a significant increase in the ischaemic myocardium immediately after reperfusion compared with both baseline (+256.8[189.9;323.5]ms (P < 0.0001) and end-ischaemia values (=+166.6[84.7;248.6]ms (P = 0.0001).

A non-significant T1 increase was observed during ischaemia also in the remote myocardium (mean difference[95CI%]) = + 22.3 [2.9;47.6]ms (P = 0.0942), followed by a noticeable increase in the non-ischaemic myocardium immediately after reperfusion, compared with baseline (+40[6.9;73.1]ms (P≤0.0139) and at late reperfusion (+38.6[9.9;67.36] ms (P≤0.0058).

We observed no correlation between end-ischaemia native T1 or T2 values, and 3 h after reperfusion in the reperfused patterns (MVO + and MVO-), except in the absence of effective reperfusion (no reperfusion pattern, see supplementary materials Supplementary data online, Figure  S4).

Acute ischaemia-reperfusion myocardial tissue responses

Upon analysis of all imaging studies and the analysis of the post-gadolinium enhanced data (EGE and LGE), we could differentiate three different reperfusion patterns in our animal population after 3 h of reperfusion, show-cased in Figures 2–4 as well as in Figure 5–6, synthetizing all results after alignment on the standardized template.

Figure 5.

Coloured images synthetizing the results of the pixel-wise statistical analysis, after alignment and spatial alignment to a common reference, of all image contents of T1 dynamic acquisition (left) and LGE (right) from all individual. The first six columns display the representative average image pattern (mean of pixel values over the aligned images) at each stage of the IR protocol and for each of the identified patterns. The last two column uses a similar representation for the image labels obtained from LGE: average pattern and standard deviation.

Figure 6.

Coloured images synthetizing the results of the pixel-wise statistical analysis, after alignment and spatial alignment to a common reference, of all image contents of T2 dynamic acquisition (left) and LGE (right) from all individual. The first five columns display the representative average image pattern (mean of pixel values over the aligned images) at each stage of the IR protocol and for each of the identified patterns. The last column useq a similar representation for the image labels obtained from LGE: average pattern and standard deviation.

The first pattern was effective tissue reperfusion, with microvascular obstruction (MVO) at the core of the myocardial infarction (R + MVO + pattern). MVO was detected on the EGE and LGE images. It was present in 6/20 animals (30%), and detected on the T2* images in 5/20 of the same animals (25%). The T1 and T2 maps co-registered to pathology and late gadolinium images are illustrated in Figure 2. In this R + MVO + pattern, a moderate T1 increase was observed during ischaemia in the area at risk [ΔT1=+6% (65.6 ± 13.3 ms)], followed by a significant (T1,T2) increase after reperfusion [ΔT1=+27.7 ± 10.3% (301 ± 106 ms), ΔT2=+29.8 ± 10.7% ((16 ± 6.3 ms)].

The second pattern was effective reperfusion without MVO (R + MVO- pattern). This pattern R + MVO- is illustrated in Figure 3. MVO was absent on EGE and LGE with a significantly smaller infarcted area compared with the first pattern (34.7 ± 12.8%LV vs. 42.0 ± 7.7%LV, respectively; P < 0.05). Like the first pattern, T1 and T2 increased non-significantly during ischaemia [ΔT1=+6% (67.6 ± 13.6 ms)], with a steep increase at reperfusion [ΔT1=+24.7 ± 10.8% (265 ± 113 ms), ΔT2=+30 ± 17.1% ((15 ± 8.3 ms)].

Figure 3.

(Pattern 2: R + MVO -: AMI effective tissue reperfusion without MVO): non-contrast tissue characterization mapping techniques obtained in the open chest model at all main stages of the infarct protocol at baseline, during ischaemia and post-reperfusion; post-Gd T1maps, early Gd enhancement (EGE), late Gd enhancement (LGE) and cine bSSFP (end diastolic phase), T1-weighted images after reperfusion and reference macroscopic histology (Evans Blue dye).

The third pattern was absence of reperfusion, despite coronary artery patency (R−). There was no significant EGE or LGE on the post-gadolinium sequences. This pattern R- is illustrated in Figure 4. Like the first two patterns, T1 and T2 increased slightly during ischaemia [ΔT1=+4% (42.7 ± 17.6 ms)], but with a significantly lower increase in T1 or T2 signal at reperfusion ΔT1=+21.1 ± 6.7% (229 ± 71.5 ms), ΔT2=+15.6 ± 8.3% (8 ± 4.2 ms).

These variations in T1 or T2 evolution during the different stages of our acute ischaemia reperfusion protocol according to the three different patterns that we describe here, are presented in Figure 7.

Figure 7.

Temporal evolution of T1 (top) and T2 (bottom) relaxation times along the acute infarct and reperfusion open chest model in swine. The results are presented per subgroup and correspond to the mean values in the whole lesion (pink) or remote (cyan) area defined on the LGE images (estimated after spatial alignment to a common reference). The thicker curves correspond to the average across each subgroup.

Compared with baseline values, during ischaemia the increase of T1 in the ischaemic myocardium was not significantly different between early and late ischaemia depending on the reperfusion pattern (all P-values non-significant). A slightly higher increase in T1 value in this first pattern (R + MVO+)T1 (+154[63.7;431.4]ms (P = 1.0) and second pattern(R + MVO-): +140[54.1;195]ms (P = 1.0)) were observed compared with the third pattern (R−) but again not reaching significance. T1 changes between early and late reperfusion also showed differences between groups with a slight heterogeneity of response between groups: in the first pattern (R + MVO+) T1 (+112[57;362]ms (P = 1.0) and second pattern (R + MVO-): +47[54;283]ms (P = 1.0)) were observed compared with the third pattern (R−) but again not reaching significance between groups.

Pixel-wise map analysis of the T1 myocardial maps

Figure 5 summarizes the results obtained using the pixel-wise quantitative analysis of T1 maps for all animals according to the 3 different pattern subgroups. This analysis quantitatively confirms the visual patterns displayed in Figures 2 to 4. All groups showed moderate T1 increase in the ischaemic segments during ischaemia [average T1 for pattern I: +123 ms (min: +98, max: +151); pattern II: +139 ms (min: +86, max: +227) ; pattern III: +104 ms (min: +11, max: +164)]. In all patterns, there was an intense T1 increase after reperfusion in the ischaemic segments [average T1 for pattern I: +248 ms (min: +72, max: +368); pattern II: +101 ms (min: −95, max: +248); pattern III: +176 ms (min: +30, max: +264)].

The same signal evolution was noted on the pixel-wise quantitative analysis of T2 maps with moderate increase during ischaemia followed by a steeper increase after reperfusion [average T2 for pattern I: +18 ms (min: +2, max: +35); pattern II: +10 ms (min: +6, max: +17); pattern III: +1 ms (min: −5, max: +8)]. There was a marked difference in the R- pattern (pattern III) compared with the two others, with a significantly slower increase in T2 signal after reperfusion (Figure 7).

Ischaemic myocardial area size assessment by CMR techniques and by pattern subgroup

The ischaemic area size was measured on each animal for each CMR sequence and compared with the area at risk size and the infarct size determined by pathology (Figure 8, supplementary appendix). Table 3 reports the mean lesion size in % of the myocardium, and also for each reperfusion pattern. Our data show that all techniques overestimated the area at risk size compared with the size of the reference AAR by pathology (mean difference[95CI%]) with, from max to min overestimation: ESL: +10[0.03;20.66]% (P = 0.049), T1 Native: +7.66[−2.65;17.98]% (P = 0.36), EGE: +6.44[−3.88;16.76]% (P = 0.626), ECV: +5.39[−5.69;16.46]% (P = 0.887), T2: +4.42[−6.02;14.87]% (P = 0.952), T1 Post Gd: +3.51[−6.8;13.83]% (P = 0.99), LGE +2.78[−7.78;13.37]% (P = 0.999).

Figure 8.

Comparison of the lesions size (LS) on SA middle slice, estimated on each CMR tissue characterization techniques obtained at the acute stage of the ischaemia-reperfusion protocol (3 h post-reperfusion), in a swine model. Reference values were obtained by macroscopic histology for myocardium at risk (Evans Blue dye, empty red circle, dotted horizontal line). T1, T2 maps, and cine bSSFP images used for the lesion sizing were these acquired last before Gd injection, i.e. at the end of the 3 h reperfusion period.

Table 3.

Lesion area in % of the myocardium in the central slice as measured by each MR technique and post-mortem histology by sub-groups

LS (%)	Late reperfusion CMR measurements							Histology measurements
			Post-Gd injection
Group	T1 native	T2	EGE	CINE	T1 post Gd	LGE	ECV	Evans Blue	TTC
All pigs	41.6 (10.5)	38.2 (10.4)	41.8 (11.0)	40.1 (12.0)	39.2 (11.7)	38.9 (12.5)	39.1 (11.3)	33.7 (8.8)	11.8 (15.1)
No Reperf	40.2 (7.5)	35.8 (6.7)	38.6 (7.2)	35.8 (6.5)	36.9 (9.2)	34.5 (8.9)	34.5 (8.7)	31.7 (8.9)	13.4 (10.7)
	50.2 (12.0)	47.2 (9.6)	46.1 (7.5)	51.8 (10.5)	42.8 (7.7)	42.0 (7.7)	49.2 (9.2)	41.7 (5.8)	8.1 (8.8)
Reperf MVO-	36.2 (11.7)	33.7 (13.8)	38.8 (11.9)	33.9 (11.5)	34.2 (10.2)	34.7 (12.8)	34.6 (10.9)	32.6 (9.2)	3.5 (4.8)

All the measurements are percentage of the total surface of the central slice and are provided as mean (standard deviation). Results are for animals that underwent the whole IR procedure and with MR imaging at all steps (all pigs: n = 20; no reperf: n = 9, reperf MVO+: n = 5, reperf MVO: n = 6).

Discussion

Using a multi-parametric quantitative CMR imaging protocol and pixel-wise analysis, we objectively tracked the dynamics of the acute changes during ischaemia and following reperfusion in experimental model. These dynamic changes revealed three distinct original patterns differentiated by LGE imaging after effective coronary artery reopening.

A crucial finding is that myocardial oedema as measured by T1 and T2 MR imaging biomarkers in the ischaemic area before reperfusion cannot be considered as a main determinant of the inflammatory status post reperfusion (see Figures 5 and 6). Indeed, T1 and T2 indexes follow a systematic increase after reperfusion rather independent from the amount of oedema in the myocardial AAR as a signature of the initial inflammatory status. Considering the well-established correlation between T1, T2, and total water content in IR models,^6,14,15 our study demonstrates that the inflammatory response (∼amount of oedema as measured by T1 and T2) after reperfusion was not related to the inflammatory status in the tissue pre-reperfusion.

Moreover, our study reveals that T1 and T2 exhibit slightly different and specific changes over time, with T1 showing a slightly higher sensitivity to changes occurring in the tissue during ischaemia, whereas the T2 increase is more restricted to the reperfusion period (see Figure 6 and S4). The amplitude of T2 changes are in the range of these reported by previous studies in swine models of I/R.⁶ We found no studies with T1 monitoring in I/R large animal models at the very early phase of AMI. The slight desynchrony of the dynamic T1 and T2 changes could indicate a potential complementarity between the indexes, and a tendency towards a specific sensitivity of each marker with the tissue water content in each involved compartment. The relaxation time variance especially in the pattern 1 (MVO positive) and pattern 3 (no reperfusion group), highlight the influence on T1 and T2 values of tissue components and complexity, including the proportion of free/bound water as well as its location (intracellular/extracellular) in the tissue.

Our study cannot provide more insights regarding water distribution at this stage. The specific sensitivity of each marker with the tissue water content in each involved compartment needs to be co-analysed with more specific markers and in larger cohorts, to understand what each imaging biomarker is more precisely revealing from the water distribution and exchanges within the insulted tissue, as well as their relation to a potential inflammatory status or grade. Current experimental methods to determine water distribution between intra- and extracellular tissue compartments are based on radionuclide labelling of macromolecules with suitable diffusion properties 8–10 that are difficult to translate to experimental and clinical practice.

A second important finding is the confirmation that the lesion size estimated on T1 and T2 maps, early or late post-Gd enhanced images (see Figure 8) are close to the true/reference measures of AAR at such early stage of myocardial infract (all P-values were non-significant), signing a mixture of oedema and necrosis in the tissue. It is well established that infarct size is still overestimated at 3 days after IR due to water content increase.^6,7 Our results also indicate a trend towards more oedema at the end of reperfusion (when MVO lesions are present), hence tending to indicate a higher inflammatory response in the lesion area in such case, encompassing the AAR in many cases. Our results are also coherent with previously acquired knowledge, reporting the non-exact equivalence between the AAR reference measure (Evans blue) and lesion size measured on T1 or T2 acquired post reperfusion,^5,7 and with strongest matching between the reference AAR and LGE or T1 Post-Gd. Our results also confirm the existence of inflammation in the remote area after reperfusion, though moderated.¹⁶

A third crucial finding of our study is that the occurrence of no-reflow and/or haemorrhage appeared also independent from the inflammatory status of the myocardium before ischaemia. Indeed, the second and third pattern do not exhibit a significantly higher T1 or T2 before reperfusion, i.e. an increased water content.

In pattern 1, the Gd chelates do not have access to the core MVO lesion and ECV measurements measured are only possible and valid outside the core lesion. A slightly higher increase T1 and ECV value in this MVO + subgroup compared with the MVO- group suggest that the water-content, i.e. inflammatory response is more severe in case of haemorrhage and MVO presence, which are well-known markers of severity of the IR lesion. Garcia-Dorado and co-authors have described an ECV increase, revealed by Gd chelates distribution, that strongly correlates with the T2 increase they measured, leading to the conclusion that early cell death during reperfusion occurs mainly through necrosis involving sarcolemmal rupture and intracellular space becoming accessible to Gd chelates.¹⁵ The same group also supported a role for oedema as a mechanism of cell death during IR injury.

Finally, slight demographic dispersion between animals can be observed between animals in term of lesion size, i.e. extension or intensity of the T1 or T2 changes, reflecting a variability of response to vessel reopening. The dispersion is well illustrated by the quantitative measures and visible in Figure 7 at most time point except baseline in the third pattern (no reperfusion), and in post Gd in the remote area. Of note, our survival rates are better than those reported in previous studies based on such challenging models⁸ and our general representation enables us to identify mean behaviours by group beyond the inter-animal variability.

Limitations and technical considerations: We studied a model of acute total occlusion in healthy animals with single-vessel occlusion and fixed ischaemia time, without pre-existing atherosclerotic disease. Also, the influence of the time to reperfusion, which remains one of the main differences in clinical care, was not explicitly explored in this study given the same size, while it could lead to substantial injury-related difference in the level of tissue inflammation (even though other factors such as the quality of life, food and physical activity could also contribute). The fixed ischaemia time (40 min) and the open-chest conditions intrinsically limitate generalizability. The bimodal-oedema timeline beyond 2–3 h which is also of high interest were also not studied.

T1 values derived from MOLLI acquisition are sensitive to HR dependence, with fluctuating RR intervals that can bias T1. Note that we used a T1 mapping sequence, that enable precise control of recovery period in seconds, independently from the number of heartbeats. T1 values are intrinsically heart-rate-adjusted at the acquisition as recommended by Kellman et al.¹⁷ Regarding the temperature, we were aware of the bias and it was also carefully regulated and controlled all along the measurements, it was not surprisingly very constant between animals at all steps (mean T° other all animals = 38.5°+−0.3°). The most critical part is during the open chest surgery where a minor global change in T° (<1°) is registered, regained at the end of surgery and prior the occlusion. We are using the Bair Hugger 750 patient warmer, a device routinely used in veterinary medicine to maintain the body temperature of animals during surgical procedures or when they are under anaesthesia. The Bair Hugger 750 is equipped with temperature control and safety features, providing a reliable solution for maintaining the body temperature of animals during surgery or recovery periods. The RR values being in the order of the RR observed in humans, we believe the reported T1 values along the protocol will still enable comparison with human studies.

Clinical perspectives

• Initial inflammation at time of therapeutic intervention (i.e. at end ischaemia) was suspected to be one of the major determinants of the prognostic after therapy, mainly determining the occurrence of reperfusion injury and increasing final infarct size. Our study suggests that there are important confounders that occur at the time of reperfusion, independent of the initial myocardial ischaemic insult.

• The proposed time-resolved monitoring would be of crucial interest in follow-up studies and to evaluate and validate new therapeutic approaches aiming at effectively treat myocardial reperfusion injury and offer an effective protection of myocardium at risk in the territory downstream of occlusion, by mean of so-called ‘cardioprotection’ strategies.

• The computer-assisted objective representation and visual synthetic rendering of the results used here avoid the need for subjective user inputs in placing regions of interest and improve the reliability of myocardial tissue assessment in time by CMR.

Conclusions

The present study demonstrates that measuring T1, T2 as well as CMR function is feasible in an animal model with a temporal resolution for a full quantitative MRI of 5 min, and offers efficient monitoring of the pathophysiological evolution of acute MI. Pixel-wise analysis strengthen the analysis and enable to clearly illustrate, synthetize while objectively quantify the tissue response before and after reperfusion. We observed an increase of cardiac MR T1, T2 imaging biomarkers in the infarcted region, indicating moderate tissue oedema during ischaemia, strongly accentuated after reperfusion. These ischaemia and post-reperfusion changes were associated with respective macroscopic wall thickness decrease/increase (cell swelling), coherent regional and global dysfunction. Each MR biomarker exhibits different sensitivity to water content and changes in the involved tissue compartments of interest, while also reflecting inter-individual variability. Small correlations between post-reperfusion and per-ischaemia native T1 indicate that the water content status post-reperfusion is poorly related with that prevailing before reperfusion, and is not the main determinant of the final status of the tissue in the observed AAR 3 h after recanalization.

Supplementary data

Supplementary data are available at European Heart Journal - Imaging Methods and Practice online.

Author contributions

Magalie Viallon (Conceptualization [equal]; Data curation [lead]; Formal analysis [lead]; Investigation [lead]; Methodology [lead]; Project administration [supporting]; Supervision [lead]; Validation [lead]; Writing—original draft [lead]; Writing—review & editing [equal]), Lorena Petrusca PhD (Data curation [equal]; Formal analysis [equal]; Investigation [equal]; Methodology [equal]; Writing—review & editing [equal])), Nicolas Duchateau PhD (Data curation [equal]; Formal analysis [equal]; Software [lead]; Visualization [lead]; Writing—review & editing [equal])), Lionel Augeul MD PhD (Conceptualization [supporting]; Investigation [supporting]; Writing—review & editing [supporting])), Michel Ovize MD PhD (Conceptualization [supporting]; Funding acquisition [lead]; Project administration [supporting]; Writing—review & editing [supporting])), Nathan Mewton MD PhD (Conceptualization [equal]; Investigation [supporting]; Methodology [equal]; Validation [equal]; Writing—review & editing [equal])), and Pr Pierre Croisille MD PhD. (Funding acquisition [lead]; Project administration [lead]; Supervision [equal]; Writing—review & editing [equal]))

Funding

This work was supported by the Agence Nationale de la Recherche (ANR-16-RHUS-0009: RHU MARVELOUS), by the Claude Bernard Université Lyon 1 (UCBL) program ‘Investissements d'Avenir‘. The authors also acknowledge the partial support from the LABEX PRIMES of Université de Lyon (ANR−11-LABX−0063), the Fédération Francaise de Cardiologie (MI-MIX project, Allocation René Foudon), and the Institut Universitaire de France.

Data availability

To limit the use of animals for research purposes, the imaging data from this study can be made available upon request to the corresponding author, provided the request includes the reasons, conditions, and objectives for the request.