Head-to-head comparison of multiple cardiovascular magnetic resonance techniques for the detection and quantification of intramyocardial haemorrhage in patients with ST-elevation myocardial infarction

Cardiovascular magnetic resonance (CMR) represents a valuable non-invasive modality for studying the ischemia/reperfusion (I/R) myocardial injury in patients with ST-segment elevation myocardial infarction (STEMI) [1]. Intramyocardial haemorrhage (IMH) is a marker of severe I/R damage being associated with microvessel wall destruction and interstitial erythrocyte extravasation [2]. In experimental and clinical studies, IMH is related to unfavourable clinical outcomes [3‐14]. Thus, non-invasive detection and quantification of IMH by CMR may play a key role in the risk stratification of STEMI patients as well as in the development of an imaging-based biomarker for testing treatments aiming to minimise I/R damage and improve patients’ prognosis [15, 16]. To date, T2* mapping is claimed to be the ‘reference standard’ for post-infarction IMH detection and quantification [17, 18]. Multi-echo T2* imaging represents the comparator against which other imaging techniques are evaluated. However, the inclusion of IMH assessment in clinical studies has been hindered due to the fact that multi-echo T2* imaging is prone to off-resonance artefacts resulting in a relevant proportion of patients with uninterpretable images [19]. Moreover, multi-echo T2*w imaging does not allow for the concomitant detection and quantification of infarct-related oedema, which provides relevant complementary information in STEMI [20]. Black-blood T2-weighted (T2w) short-TI-inversion recovery (STIR) and bright-blood T2prep steady-state-free precession (SSFP) as well as T1 mapping (T1 map) and T2 mapping (T2 map) have been used in previous studies for the visualisation and quantification of ischemia-related oedema and IMH [3‐14]. With respect to IMH identification and quantification, these studies were limited by the lack of a properly defined reference standard, spectrum bias due to the absence of a healthy control group and the absence of a direct comparison among the diverse techniques. They also seldomly reported the precision of the techniques for IMH quantification, essential information for sample size calculation when planning randomised controlled trials.

Based on these premises, we studied a cohort of STEMI patients and healthy control using T2w-STIR and T2prep-SSFP as well as T2/T1 maps and multi-echo T2*w imaging. We set T2*w imaging as the reference standard, and we hypothesised that, when compared with T2*w, CMR alternative techniques may (1) be associated with improved image quality, diagnostic confidence, and non-inferior diagnostic accuracy for IMH detection; (2) provide reliable surrogate estimates of IMH (for mapping techniques); and (3) offer better intra- and inter-observer reproducibility.

The major study findings can be summarised as follows. First, T2 map, T1 map, and T2w-STIR have a good-to-excellent per-subject and per-segment diagnostic accuracy for IMH diagnosis. The image quality, diagnostic confidence, and intra- and inter-observer reproducibility were higher for mapping techniques than T2w-STIR or T2prep-SSFP imaging. Second, hypocore on T2 map and T1 map correlated strongly but slightly overestimated the IMH extent with an overall good-to-excellent agreement. Third, hypocore extent quantified on T2 map and T1 maps showed better intra- and inter-observer reproducibility than IMH size measured on T2*w. Overall, these results strongly support the use of mapping for diagnosing and quantifying the hypocore as a surrogate measure of IMH as detected by multi-echo T2* imaging.

In our study, we used T2*w imaging as the clinical gold standard for IMH, which has been validated against histology and mass spectrometry for post-reperfusion haemorrhagic infarcts [10]. Although we paid particular attention in setting our clinical protocol by limiting the breath-hold duration and maximising B₀ field homogeneity, as many as 8% of patients screened for study inclusion were finally excluded due to off-resonance artefacts on T2* imaging [19]. In the remainders, T2*w imaging performed well with respect to the overall image quality and diagnostic confidence in combination with a good intra- and inter-observer reproducibility for both qualitative (presence or absence of IMH) and quantitative data (IMH quantification). T2*w imaging, however, does not allow concomitant detection and quantification of infarct-related oedema, which could provide important additional information in STEMI in the early post-infarction phase [25, 26]. This limitation comes with the following drawbacks: Firstly, if T2*w imaging is used for IMH detection/quantification, an oedema-sensitive (e.g. T2 or T1 maps) technique has to be included in CMR protocol leading to prolonged scanning time and reduced patients’ comfort. Secondly, the proportion of patients excluded because of poor T2*w gives rise to a substantial increase in required study sample size [15‐19]. Our results indicated T2 map, T1 map, and T2w-STIR had the best accuracy for assessing the presence or absence of IMH on per-subject and per-segment basis. However, T2w-STIR, alike T2prepSSFP, had lower image quality and diagnostic confidence compared with mapping techniques [27], and in 2 subjects (4%), T2w-STIR and T2prepSSFP images were excluded from the analysis because of very poor imaging quality. Within the remainders, the operator was uncertain or very uncertain in attributing or excluding IMH diagnosis based on T2w-STIR and T2prepSSFP images in 2 (3%) and 5 (8%) cases, respectively. In contrast, T2 map and T1 map had good-to-excellent image quality in combination with high or very high diagnostic confidence for IMH diagnosis in all cases. Our study results endorsed and expand previous knowledge about the use of mapping techniques for diagnosing and assessing IMH in STEMI patients. In particular, our findings are in line with those reported by Bulluck et al [28] in a smaller cohort of subjects and confirming the good sensitivity and specificity of T1 and T2 maps for IMH detection. That study was limited by the lack of a healthy control group (spectrum bias) and of direct comparison with T2wSTIR and T2prep-SSFP, which have been largely used in previous experimental and clinical studies for IMH detection [3‐5, 7]. Our study superseded the limitation of the previous literature by reporting a properly chosen reference standard (i.e. T2*w), a head-to-head comparison of the most often used CMR techniques for infarct-related oedema and IMH imaging, and by including an age- and gender-matched healthy control group to reduce spectrum bias. We also detailed the precision of the diverse techniques for IMH quantification. Therefore, our study results provide a comprehensive background for researchers aiming to use CMR-based IMH identification and quantification for improving risk stratification in STEMI patients or for planning treatments to mitigate the deleterious I/R phenomenon.

It has to be acknowledged that T2/T1maps do not provide a direct measure of iron content within the infarct core. Based on our study results, and in particular the good diagnostic accuracy of T2/T1 maps in detecting IMH, it is reasonable to reserve T2* imaging to patients showing hypocore on T2 or T1 maps and withhold it in those without. In addition, given the settings of the current study which utilised a 1.5 magnetic field, dedicated studies on 3-T scanners are needed to test the diagnostic accuracy and precision of T2/T1 maps in the visualisation and quantification of IMH.

Finally, we found that hypocore by T2 map or T1 map slightly overestimated the size of IMH as quantified by T2*w imaging. Likewise, hypocore on parametric imaging and IMH represent two diverse aspects of I/R injury. In pre-clinical models of reperfused STEMI, histology data invariably showed a central necrotic core devoid of inflammation and blood flow due to the extensive irreversible microvascular damage alongside spotty areas of IMH [29‐31]. It is likely that hypocore on parametric imaging represents the central necrotic core of reperfused infarcts, thus explaining why the hypocore areas slightly but consistently overestimate T2*w-based IMH. Furthermore, we found only a moderate positive relationship between the hypocore T2 values and T2* relaxation times and no correlation was found between T1 value and T2* relaxation time. This finding underpins that hypocore and IMH are two distinctive albeit closely interrelated phenomena of I/R injury. Concurrently larger hypocores on mappings are more likely to detect IMH as highlighted by the receiver operating curve analysis showing that a hypocore extent of 1% on T2 map had a sensitivity of 95% for diagnosing IMH. In IMH-positive cases, the T2 shortening is principally caused by magnetic susceptibility effects due to the compartmentalisation of paramagnetic deoxy- or methemoglobin inside the red cells, which usually occurs between 1 and 3 days after post-infarction IMH. In contrast, the determinants of T1 shortening in cases with IMH are likely more complex. Deoxyhaemoglobin, the predominant intra-erythrocyte form of degraded haemoglobin, is rather inaccessible to water molecules due its three-dimensional conformation resulting in negligible effect in T1 shortening. Methemoglobin, on the other hand, is prevalent in the haemorrhagic infarct between 4 and 14 days and exerts strong paramagnetic T1 shortening [8‐12].

The study holds several limitations. Firstly, the head-to-head comparison of the diverse sequences was performed on one single slice (target slice). As a result, it was not possible to investigate whether the alternative techniques were able to visualise hypointense or hypo-T2/T1 regions within the infarcted myocardium not otherwise detected by T2* imaging. Histological validation was not possible in the current study and therefore, we were unable to investigate the contribution of diverse pathophysiological components of I/R injury on the relaxivity properties of the myocardium. Several studies adopted a T2* relaxation time < 20 ms as the gold standard for post-infarction IMH detection and quantification [5, 6, 9, 28, 31]. This cut-off is based on pathological values derived from explanted hearts of patients who died of hemochromatosis, and it has not been validated histologically in large animal models of I/R [32]. To the best of our knowledge, only Kali et al validated multi-echo T2* imaging with histology-based gold standard for post-infarction IMH [10]. Because IMH and microvascular obstruction (MVO) are closely related [29‐31], it was not possible to discriminate the relative contribution of IMH or MVO to the relaxation time changes of the hypocore. Finally, T2 relaxation time of the hypocore region cannot be used to assess the severity of IMH.