Introduction

Microvascular obstruction (MVO) and intramyocardial haemorrhage (IMH) are both frequent phenomena after reperfused ST-elevation myocardial infarction (STEMI) [1, 2] and partly represent reperfusion injury after primary percutaneous coronary intervention (PPCI) [3]. MVO or the so-called no-reflow phenomenon is caused by (1) embolization of particulate debris, (2) release of vasoconstrictor, thrombogenic and inflammatory substances, and (3) structural collapse of the capillary bed. MVO is associated with adverse ventricular remodelling and worse prognosis [4]. Equally, IMH is considered as severe damage after revascularization by PPCI or fibrinolysis which portends adverse prognosis [5]. Structural and functional degradation of the microcirculation causes extravasation of erythrocytes in the reperfused myocardium [6, 7]. The pathophysiology of IMH is not fully understood yet and it remains unresolved if IMH is a consequence of MVO or vice versa and CMR and other imaging modalities might help to better understand that phenomenon [814]. However, previous studies demonstrated that IMH presence is an independent predictor of adverse left ventricular (LV) remodelling and associated with larger infarct size [15]. O’Regan et al. demonstrated a positive correlation between IMH and MVO and also IMH and infarct size in one of the first human “in vivo” cardiac magnetic resonance (CMR) studies using T2 *-weighted images [16]. T2 *-weighted CMR is a sensitive and potentially more specific technique for detecting IMH [8, 9]. The T2 * sequence was originally used to illustrate cerebral haemorrhage [17, 18]. T2-weighted CMR was utilized for quantifying oedema [3, 5], detecting the area at risk (AAR) and calculating the myocardial salvage [5,1921]. In some cases a so-called hypointense core (HIC) could additionally be visualised within the increased signal intensity representing oedema. These regions of hypointense signal have been demonstrated to correspond to histological evidence of myocardial haemorrhage [8, 9]. However, studies have also demonstrated that MVO may also cause hypointense infarct cores [22]. Consequently, it does not seem possible to differentiate if HIC illustrates MVO, IMH or both by using exclusively T2-weighted sequences. To examine the exact extent and prognostic impact of both findings as well as influencing factors, it is crucial to precisely distinguish between IMH and MVO.

The aim of this study was therefore to evaluate the performance of T2 * and HIC for detection of IMH and to assess the relationship between IMH and clinical as well as CMR markers of myocardial damage.

Material and methods

Patient population

A total of 151 consecutive patients (114 men, 37 women) undergoing PPCI for STEMI at a tertiary care centre were included between August 2009 and March 2010. All patients presented with typical chest pain within 12 h (Table 1) and had ST-segment elevation of ≥0.1 mV in ≥2 extremity leads or ≥0.2 mV in ≥2 precordial leads.

Table 1 Patient characteristics

All patients gave written informed consent and the study was approved by the local ethics committee. This analysis is a substudy from the Abciximab Intracoronary versus intravenously Drug Application in STEMI (AIDA STEMI) trial (ClinicalTrial.gov Identifier NCT00712101) [2325]. Exclusion criteria for this CMR substudy were claustrophobia, pregnancy, clinical instability, metallic implants, implanted pacemakers or defibrillators.

The patients were categorized into two groups defined by the presence or absence of IMH by using T2 * imaging. Furthermore, cardiovascular risk factors, reperfusion times, medication, culprit vessel, thrombolysis in myocardial infarction (TIMI) flow grade before and after PPCI as well as LV function and volumes were assessed to define possible influencing factors of IMH (Table 1). For culprit lesion identification, the three main vessels—left anterior descending artery, right coronary artery or left circumflex artery—were also assessed for proximal, mid or distal lesion location. The blood flow in the culprit vessel was graded according to the TIMI flow grade before and after PPCI: grade 0, no reperfusion; grade 1, low reperfusion; grade 2, partial reperfusion; and grade 3, regular perfusion [26].

CMR imaging protocol

CMR was performed on a 1.5-Tesla scanner (Intera, Philips Medical Systems, Best, the Netherlands) using a five-channel phased-array surface coil within 8 days after PPCI.

  • First, for functional analysis standard steady state free precession (SSFP) sequences in breath-hold were obtained in four- and two-chamber view and a stack of short-axis slices to cover the whole LV as described previously [2729]. The scan parameters were as follows: repetition time (TR) = 3.6 ms, echo time (TE) = 1.8 ms, flip angle = 60°, slice thickness = 8 mm, maximum field of view (FOV) = 400 mm, matrix 256 × 256, in-plane resolution <1.56 × 1.56 × 8 mm.

  • Second, a stack of short-axis slices covering the whole LV using a black blood T2-weighted short tau inversion recovery (T2-STIR) sequence (TR = 2 RR intervals, TE = 80 ms, flip angle = 90°, slice thickness = 10 mm, maximum field of view (FOV) = 370 mm, matrix 512 × 512, in-plane resolution <0.72 × 0.72 × 10 mm) were acquired for detecting HIC within myocardial oedema (Figs. 1a, 2a) and the area at risk (AAR) by a free-breathing navigator technique.

    Fig. 1
    figure 1

    IMH with MVO and HIC: Short-axis CMR images obtained 2 days after ischemic reperfusion injury (pain-to-balloon time 523 min) in a 53-year-old man. In the anteroseptal myocardium a large transmural hyperintense region and HIC directly centrally located therein (white arrows) are present on the T2-STIR image (a); also, a major hyperintense, contrast-enhancing region (necrosis) with a clear MVO (white arrows) within can be visualised on the LGE image (b) and a huge haemorrhage indicated by a hypointense signal (white arrows) can be seen on the T2 *-weighted image at a TE of 15 ms (c) and within the T2 *-weighted parameter image indicating a T2 *-decay ≤20 ms (d), the threshold for haemorrhage in this study [16]

    Fig. 2
    figure 2

    IMH without MVO and HIC: Short axis CMR images obtained 3 days after ischemic reperfusion injury (pain-to-balloon time 126 min) in a 74-year-old woman with an occlusion of the circumflex artery (LCX). a Demonstrates an oedema in the inferior wall without a HIC, b the LGE in the inferior wall without any MVO and c an obvious haemorrhage (white arrows) on the T2 *-weighted image at a TE of 15 ms and within the T2 *-weighted parameter image indicating a T2 *-decay ≤20 ms, d the threshold for haemorrhage in this study [16]

  • Third, the same slice orientation as for the T2-STIR image was chosen for IMH detection in the middle of the myocardial oedema, centrally located within the “area at risk” of myocardial infarction (Figs. 1a, 2a). The T2 * sequence with the following parameters was used: TR = 17 ms, TE = 1, 3, 6, 8, 10, 12, 15 ms, flip angle = 20°, slice thickness = 10 mm, maximum field of view (FOV) = 240 mm, matrix 256 × 256, in-plane resolution <0.94 × 0.94 × 10 mm. Images at seven different TE (Figs. 1c, 2c) to calculate relaxation curves, a parameter image (Figs. 1d, 2d) and the T2 * decay [16] were obtained from the parameter images. The occurrence or absence of HIC with or without IMH was documented.

  • Fourth and finally, late gadolinium enhancement (LGE) images (Figs. 1b, 2b), sensitive for necrosis and MVO, were acquired 10–15 min after intravenous administration of 0.15 mmol/kg Gd-DTPA by using a 3D-inversion recovery gradient echo sequence (IR-GRE) with the following parameters: TR = 2.9 ms, TE = 1.5 ms, flip angle = 15°, slice thickness = 10 mm (5 mm increment), maximum field of view (FOV) = 370 mm, matrix 280 × 280, in-plane resolution <1.32 × 1.32 × 10 mm.

CMR image analysis

All images were analysed using a dedicated workstation (ViewForum R4.2V1C2 SP1, Software Release 2005-11-04, Philips Medical Systems, Best, the Netherlands). LV volumes and function for every patient were assessed by manually outlining the endocardial contours of the LV from the SSFP short axes sequences [5]. Short-axis T2-STIR sequences were used to detect and to assess oedema and HIC (Figs. 1a, 2a). In the LGE images the size of myocardial infarction was measured by using a threshold of two standard deviations of the remote myocardium [5, 27]. In addition, the presence or absence of MVO was documented (Figs. 1b, 2b). The salvaged myocardium after PPCI was expressed by the myocardial salvage index (MSI) calculated from the difference between oedema and necrosis divided by oedema [21]. IMH was considered as the region with decreased signal intensity in the T2 *-weighted images based on the paramagnetic effects of deoxyhaemoglobin and the reduction of the T2 time [30]. The used sequence provided T2 *-weighted images with the described different TE additionally to a parameter image, which was used to calculate the T2 * decay. According to previous publications a decay at the T2 * parameter image ≤20 ms was considered as positive haemorrhage [16] (Figs. 1d, 2d).

Statistical analysis

The study cohort was characterized by mean ± standard deviation (SD) for continuous variables and by absolute and relative frequencies for categorical variables. Continuous variables were compared within the group of patients with and without IMH by Student’s t test. Proportions were compared by Pearson’s χ2 test (Tables 1, 3) and Fisher’s exact test where appropriate.

Furthermore, a stepwise logistic regression analysis was applied to identify possible predictors of the occurrence of IMH. All statistically significant clinical and CMR variables of the univariate analyses (Tables 1, 3) were included in the final multivariable model. All variables selected by this algorithm were included in a final model in order to get odds ratio estimates with the corresponding 95 % confidence intervals (Table 4). All tests performed were two-sided with a significance level of 5 %. All statistical analyses were done by means of SPSS software, version 16.0 for Windows.

Results

CMR was performed at a mean of 2.8 days (range 0–8 days) after PPCI. The mean patient age of the 76 patients with IMH was 60 ± 11 years. There were no significant differences between patients with or without IMH concerning common patient characteristics except for hypercholesterolaemia (Table 1).

Functional and volumetric assessment

LV ejection fraction was significantly lower in patients with IMH compared to those without (43.5 ± 9.7 % vs. 51.1 ± 8.6 %, P < 0.001) (Fig. 3). Patients with IMH also had a higher LV-EDV (163 ± 40 ml vs. 145 ± 30 ml, P < 0.01) and LV-ESV (94 ± 36 ml vs. 72 ± 23 ml, P < 0.001).

Fig. 3
figure 3

Relationship between LV ejection fraction and haemorrhage. Patients with IMH had a significantly lower LV-EF than those without IMH (43.5 ± 9.7 % vs. 51.1 ± 8.6 %, P < 0.001)

Comparison of T2 * mapping and T2-weighted imaging for IMH detection

By using T2 * mapping IMH was detected in 50 % of all patients (76/151). T2-weighted STIR images demonstrated HIC in 54 % (82/151) and LGE images demonstrated MVO in 66 % (100/151) of all patients (Fig. 4).

Fig. 4
figure 4

Venn diagram of the frequency of IMH, MVO and HIC out of the basic population. It demonstrates the occurrence of IMH without MVO and/or HIC. CMR cardiovascular magnetic resonance, IMH intramyocardial haemorrhage, MVO microvascular obstruction, HIC hypointense core in T2-STIR

IMH occurred in the majority of patients (95 %; 72/76) concomitantly with MVO (Fig. 1) and in only 5 % (4/76) without (Fig. 2). When IMH and MVO were present at the same time IMH occurred in centre areas of the MVO. This phenomenon was observed especially in large infarcts with large MVO areas. Additionally, two of these four patients showed neither HIC nor a concomitant MVO. HIC occurred in 6 % (5/82) of patients without MVO.

Eighty-four percent (64/76) of patients with IMH showed HIC in T2-STIR images (Table 2), whereas in 16 % (12/76) of patients IMH was not accompanied by HIC.

Table 2 Differences in the detection of HIC and MVO in IMH+ versus IMH patients

Influencing factors for IMH occurrence

Differences could be demonstrated related to potential influencing factors for the occurrence of IMH. The following factors were analysed:

Location of occlusion

In 42 % (64/151) of patients the LAD, in 39 % (59/151) the RCA and in 18.5 % (28/151) the LCX was occluded. The culprit vessel in the majority of patients with IMH was the LAD (50 % vs. 35 %, P = 0.06) and without IMH was the RCA (29 % vs. 49 %, P = 0.01) (Table 1).

TIMI flow before PPCI

In IMH+ patients a TIMI flow ≤ 1 before PPCI was significantly more frequent than in IMH patients (72 % vs. 54 %, P < 0.001) (Fig. 5).

Fig. 5
figure 5

Relationship between TIMI flow before PPCI ≤1 and haemorrhage: In IMH+ patients a TIMI flow ≤1 before PPCI was significantly more frequent than in IMH patients (71 % vs. 51 %, P < 0.001), which could be considered as a potential influencing factor for the development of IMH

Medication

A statistically significant difference in premedication of the IMH+ vs. IMH group could only be demonstrated with regard to the intake of beta-blockers (34 % vs. 16 %; P = 0.01) (Table 1). All other premedication, like angiotensin-1 antagonists, ACE inhibitors or the intake of acetylsalicylic acid (ASA), clopidogrel or statins were not significantly different in both groups.

Infarct size/myocardial salvage index (MSI)

IMH+ patients had significantly larger infarcts (32.6 ± 20.3 ml vs. 12.9 ± 9.5 ml, P < 0.001) (Fig. 6) and a lower MSI (36 % ± 21 % vs. 59 % ± 25 %, P < 0.001) (Fig. 7).

Fig. 6
figure 6

Relationship between infarct size and haemorrhage: the IMH group had significantly larger mean infarct sizes (32.6 ± 20.3 ml vs. 12.9 ± 9.5 ml, P < 0.001) at the LGE sequences

Fig. 7
figure 7

Relationship between myocardial salvage index and haemorrhage. MSI in the IMH group was significantly lower than in the non-IMH group (0.36 ± 0.21 vs. 0.59 ± 0.25, P < 0.001)

Predictors of IMH

A stepwise logistic regression model was applied to identify possible predictors of the occurrence of IMH (Tables 1, 3). After the use of a univariate analysis several significant parameters were left for the multivariate analysis, like TIMI flow grade before PPCI ≤ 1 or the MSI. Nevertheless, the final multivariate analysis revealed that patients with a HIC, or even more likely with the presence of MVO, demonstrated a high OR for the occurrence of IMH (Table 4). These results again demonstrate the close relation between HIC and MVO and the frequent appearance of both phenomena together with IMH. Surprisingly, an intake of beta-blockers demonstrated the highest OR for the occurrence of IMH. However, there was neither a significant difference between IMH+ and IMH patients according to the presence of arterial hypertension, nor a tendency toward higher systolic or diastolic blood pressure in the IMH+ group. However, in the IMH+ group more patients were on a combined antihypertensive drug therapy with beta-blockers as compared to the IMH group.

Table 3 CMR results
Table 4 Characteristics associated with IMH/predictors of IMH

Discussion

In our large study (n = 151) we could demonstrate the importance of using dedicated T2 *-weighted CMR sequences to detect IMH after acute reperfused STEMI to definitely distinguish IMH from MVO. In some cases IMH could be detected even without a visible MVO (Fig. 2), which is in contrast to several other studies [5, 6, 11, 31, 32]. Our data indicate that T2-STIR sequences only, which are generally used to visualise AAR and MSI, are not able to reliably detect IMH as a HIC within the oedema. Furthermore, potential influencing factors for the occurrence of IMH could be evaluated.

Our most important findings can be summarized as follows:

  1. 1.

    IMH is a common finding in patients with acute STEMI after PPCI and was present in approximately 50 % of our patients.

  2. 2.

    Our data demonstrated that the majority of T2 * defects were also defects on T2 imaging (HIC), but some T2 defects had had no defect on T2 * imaging. Although not directly proven by our works, experimental studies suggest [3336] that T2 * imaging is more accurate for IMH than T2 imaging and therefore should be the preferred technique for detection of IMH.

  3. 3.

    However, according to the multivariate analysis HIC on T2 imaging and MVO on LGE images are highly correlated with occurrence of defect on T2 * images indicating IMH.

CMR after reperfused STEMI and the role of the used sequences

CMR is considered as the gold standard for the assessment of ventricular volumes and mass [28, 37] as well as in patients with acute reperfused myocardial infarction as the preferred method for the noninvasive evaluation of viability, infarct size and myocardial salvage with a high prognostic value [5, 38, 39]. The prognostic value of MVO and IMH after acute reperfused myocardial infarction has previously been described [5, 11, 16]. However, there is disagreement in the literature about the best CMR sequences to detect IMH [5, 7, 10, 11, 15, 30, 3336, 38, 40, 41] and how to definitely differentiate it from MVO.

Pedersen et al. demonstrated in an animal study that T1-weighted sequences are more accurate than T2- or T2 *-weighted sequences for the detection of IMH [40]. In contrast to this study Bradley and Kali et al. [17, 34] showed the highest sensitivity for IMH detection by using T2 *-weighted CMR sequences, which is in line with previous data in neuroradiology [17, 18, 42] and animal models using T2 * mapping in the heart [10]. Lotan et  al. declared difficulties in the detection of haemorrhage by using T2-weighted images only. That study group stated that the reciprocal influencing effects of the hyperintense oedema and HIC may mask IMH [31]. Furthermore, a postmortem study by Jackowski et al. [35] demonstrated that MVO without haemorrhage and HIC on T2-weighted images exist.

Nevertheless, there is a strong relationship between the occurrence of IMH and MVO [41]. Also this study demonstrated MVO in 95 % of patients with IMH.

But it has also been described before that some patients may exclusively present IMH or MVO [38]. In the current study 5 % of patients with IMH did not show any MVO. Therefore, we suppose that T2 *-weighted sequences could detect small amounts of haemorrhage with a very high sensitivity, despite a slightly lower spatial resolution as compared to the T2-STIR, but also only slightly higher than the LGE sequences.

However, in previous studies and also in our study it could not definitely be clarified if IMH can be present without MVO and further studies are still needed to examine this finding. Prior analysis had demonstrated that T2 *-weighted sequences in particular can reliably detect even small amounts of cerebral haemorrhage [18, 42, 43]. But the spatial resolution of the presently used LGE sequences may not be high enough to depict every MVO, while the corresponding T2 *-weighted sequences do for the detection of IMH. The spatial resolution of the T2 * sequence in our study was slightly higher than the spatial resolution of the LGE sequence.

Nevertheless, artefacts are common on T2 *-weighted images and could mimic IMH. To reduce this potential source of misinterpretation a hypointense signal on the T2 * images was only defined as IMH if the decay on the parameter image was at most 20 ms as previously described [16]. In another study an even more sensitive threshold with a T2 * decay of less than 30 ms was proposed [32], which may additionally increase the number of IMH+ patients without any visible MVO or HIC.

The occurrence of IMH without MVO on the LGE images may also be caused by passive diffusion of contrast agent into the small infarct core, obscuring the presence of MVO on the LGE images [44].

Furthermore, it is not yet understood what a HIC in T2-weighted STIR sequences represents when IMH and MVO could not be demonstrated (2 % (3/151) of all patients in this study). In one of only a few histologically confirmed postmortem analyses [35] HIC always represented MVO with or without IMH.

Relationship between IMH and potential influencing factors

Previous publications and this study confirmed that IMH is not always detectable in patients after acute reperfused STEMI. In about half of our patients IMH was detectable, which is consistent with findings from Beek et al. with 49 % of patients with IMH [3] or Robbers et al. with 60 % [44], but much more frequent as described in other studies [5, 45]. Therefore, potentially influencing factors for the occurrence of IMH should be evaluated. In our study most patients with IMH showed an occluded LAD (50 %) and patients without IMH presented more often an occluded RCA, which had already been described by Amabile et al. [45]. The majority of our patients (42 %) had their culprit lesion in the LAD territory (Table 1) and the IMH+ mostly in the proximal segment (50 % vs. 38 %), whereas in the LCX it was mainly located at the middle segment, with an equal distribution within both groups. The second most involved territory was the RCA region (39 %), in which the IMH+ demonstrated an almost equal distribution of proximally and distally located lesions. But in the IMH group the culprit lesion was mainly located proximally in more than 50 % of patients. Infarct size roughly correlates with the location of the culprit lesion and a proximal lesion in the LAD territory causes usually larger infarcts than a proximal lesion in the RCA. Furthermore, IMH correlated significantly with infarct size.

Patients with a low TIMI flow before PPCI (≤1) seem to have a higher risk for the development of IMH (Fig. 5). Seventy-two per cent of our patients with a TIMI flow grade ≤1 demonstrated an IMH, which was significantly different to those patients without IMH. Similar results could be demonstrated in prior studies, e.g. Eitel et al. reported on 87 % patients with a TIMI flow grade ≤ 1 before PPCI [5] and a HIC, which was used as an indicator for IMH in this study. Nevertheless, in the final multivariate analysis (Table 4) only HIC and especially MVO were significant predictors of IMH. Furthermore, the beta-blocker intake at admission was a strong predictor of IMH in patients. In the IMH+ group more patients were on a combined antihypertensive therapy indicating a potentially more severe arterial hypertension, but this is rather speculative and has to be proven using another study design. However, several large trials on the use of beta-blockers in myocardial infarction demonstrated that the use of beta-blockers reduced the risks of reinfarction and ventricular fibrillation, but increased the risk of cardiogenic shock [46, 47]. These findings changed the clinical practice of beta-blocker use in patients presenting with acute coronary syndromes, emphasizing the potential harm of beta-blockers in patients at increased risk of cardiogenic shock. Nevertheless, this unexpected finding may also be an epiphenomenon.

Relationship between IMH and ventricular function

Other studies have already demonstrated the relationship between the presence of IMH and significantly larger infarct sizes, decreased left ventricular function, less myocardial salvage and an adverse prognosis [3, 5, 11]. IMH after STEMI is seen as a reperfusion injury and closely related to larger infarct sizes [5]. In this study we confirmed a significantly lower LV-EF, larger LV-EDV and LV-ESV and larger infarct sizes in patients with IMH. These findings were also seen in studies by Eitel et  al. and Beek et al. [3, 5, 44]. Beek et al. did not find any prognostic influence of MVO [3]. O’Regan et al. supposed that the occurrence of IMH is related to a reduced myocardial salvage [16]. In our patients a significant impact of IMH on ventricular function and volumes could be demonstrated, therefore it might be considered to be a potentially important prognostic factor.

Limitations

This is a retrospective study on limited patients with acute reperfused ST-elevation myocardial infarction. A prospective study on a bigger study cohort would be desirable to confirm our results and even more important to confirm the prognostic significance of IMH in this group of patients.

As a result of time constraints in this study, like in others [16], we used only one single slice T2 * mapping sequence in the most representative slice in the middle of the myocardial oedema. An additional retrospective analysis of our data revealed that that was always in agreement with the centre of the LGE area. More slices could not be acquired and therefore possible haemorrhage in other areas might be missed. Therefore, it can be considered that the frequency of IMH is probably even higher than described in the present study. Further studies should therefore use a whole heart T2 * mapping approach. We also could not compare the total size of IMH to the size of HIC or MVO because of the one single slice T2 * mapping.

The period of time after infarction at which CMR was performed could also have been an important influencing factor for the IMH detection rate, because of the different degradation stages of haemoglobin, which have an effect on the signal intensity [17, 48, 49]. In this study the CMR analysis was performed within 8 days after infarction (mean 2.8 days). It is known that the oxyhaemoglobin of fresh bleeding does not have any paramagnetic effect and therefore can not be detected by T2 *-weighted MRI. As a result of oxidative denaturation, deoxyhaemoglobin with its detectable paramagnetic effects occurs usually after 12–72 h, which is known from cerebral MRI studies [48]. Therefore, it can be hypothesised that in these patients additional haemorrhage may have been detected if MR imaging had been performed later. In addition to that, Kali et  al. described the good commitment of T2 *-weighted CMR for the assessment of changes in iron deposition in postinfarcted myocardium [33, 34]. The paramagnetic effects of iron are still visible on T2 *-weighted images after months.

Conclusions

IMH is a frequent finding in reperfused STEMI associated with reduced ventricular function and myocardial damage, which can be assessed by CMR. Our data indicates—in concordance with other recently published data [3336]—that T2 * mapping seems to be more suitable than T2-weighted images alone for the detection of IMH in reperfused acute myocardial infarction. Despite the fact that it is rather closely correlated with a HIC in T2-weighted images and MVO in LGE images, it can also occur alone according to our results.

In a larger trial with additional follow-up examinations and especially the use of a full left ventricular coverage of T2 * mapping, it has to be proven if this is really a separate phenomenon to MVO with a clinical and prognostic impact or just another possibility to detect MVO with a higher sensitivity as compared to T2-weighted STIR and LGE sequences. However, this may allow for a better risk stratification in patients after acute reperfused myocardial infarction.