In this work, we have found that T1 is elevated in regions with abnormal wall motion and that these regions co-localize with regions with elevated T2. There is strong correlation between myocardial T1 and T2 SI ratios. Further, T1-mapping detected myocardial involvement beyond that identified by wall motion abnormality. We have demonstrated for the first time that non-contrast T1-mapping detects acute myocardial edema with high diagnostic accuracy compared to conventional dark-blood and newer bright-blood T2w-CMR at 1.5 T.
CMR for the detection of acute myocardial edema
Given that edema is a generic tissue response to acute insults, edema imaging appears to be a highly suitable tool for assessing acute myocardial injury. The “gold standard” for defining myocardial edema on imaging, however, is controversial [
22]. Experimental animal models using fluorescent microspheres to delineate myocardium area-at-risk in acute MI correlated to T2w-CMR imaging [
4]; however, topographic correlation between T2w images and histopathological validation of pure myocardial edema without infarction in humans is lacking, largely due to the difficulty in accurately measuring edema in histopathological samples. We thus decided to study patients presenting with cardiac conditions known to involve acute edema but without co-existing infarction or other pathology detected by LGE. In Takotsubo cardiomyopathy and acute myocardial stunning, inflammation and edema are main features described on human endomyocardial biopsies and animal models [
23‐
25]. Although patients with regional edema did not fulfill criteria for the typical patterns of Takotsubo cardiomyopathy, and possible mechanisms may include coronary spasm or plaque events, the cases precipitated by an emotionally stressful event may represent an atypical form of stress-induced cardiomyopathy not fitting the traditional description of apical ballooning or other morphologic variants commonly described [
17]. There was no viral history to suggest a myocarditic process. The common feature for all patients in this study was that all had acute wall motion abnormalities associated with edema without infarction on CMR. While not proven histologically in our cohort, it is highly likely that segments with acute wall motion abnormality would demonstrate features of acute edema. Our assumption was strongly supported by the fact that segments with abnormal wall motion showed numerically higher average T2 SI ratio and T1 values than controls.
ShMOLLI T1-mapping: a novel method for imaging myocardial edema
We recently developed the ShMOLLI technique for fast T1-mapping [
16], which overcomes many of the limitations of both conventional MOLLI T1-mapping and dark-blood T2w-CMR. Compared to MOLLI, ShMOLLI is ~50 % shorter in breath-hold times; it is also heart rate- and T1-independent over a wide range of T1 values, and thus is more accurate in measuring long T1s in edema and at fast heart rates [
16]. Compared to dark-blood T2w-CMR , ShMOLLI is robust to tachycardia (shown in simulation and phantoms for rates up to 100 bpm [
16] and tested in clinical subjects with sinus tachycardia and atrial fibrillation at ventricular rates of up to 150 bpm; unpublished own data). It does not suffer from signal dropout due to through-plane cardiac motion and breath-holds are considerably shorter.
Bright-blood T2 is excellent for detecting large areas of regional edema when there is obvious normal myocardium to be used as a reference, as shown recently [
3] and in this work. The main challenge of extending bright-blood T2 imaging for the detection of diffuse edema was the variability of the image quality of reference skeletal muscle, which was often affected by off-resonance artifacts, bright signals from adjacent fat, bright blood vessels within the muscle or signal dropouts. For this reason, STIR outperformed bright-blood ACUT2E in this setting as STIR images have a more homogenous signal across the field of view, and thus the quality of the skeletal muscle was generally better, more consistent and more reliable. This study suggests that bright-blood T2 may not always be suitable as a replacement method for conventional dark-blood T2w-CMR.
T1-mapping using ShMOLLI has now been shown to be highly sensitive to acute changes in myocardial water characteristics which, up to this point, have been evaluated mainly using T2w-CMR. From a mechanistic point of view, T1-mapping and T2-weighted MRI likely detect acute myocardial injury via distinct, but complementary and overlapping mechanisms. Interestingly, we detected significant T1 elevation even in patient segments with normal wall motion (Table
1), a finding accompanied also by significantly higher T2 SI in segments with normal wall motion in Takotsubo patients (by dark-blood STIR but not bright-blood ACUT2E) as well as in patients with regional edema (by bright-blood ACUT2E but not dark-blood STIR). This suggests that T1-mapping may be the more sensitive technique in detecting acute myocardial changes in both global and regional pathologies, and/or detects myocardial changes beyond that identified by wall motion abnormality and the T2-weighted methods tested. Previous experimental work demonstrated that T1 relaxation times are prolonged in regionally ischemic tissue from dog hearts, and while largely reflective of increased free water content of the ischemic tissue, T1 changes exceeded the changes in water content as the duration of ischemia was increased [
13]. The prolongation of T1 values was found to be unrelated to changes in metabolic parameters such as tissue content of oxygen, carbon dioxide, hydrogen ions, lactate, adenosine triphosphate or phosphocreatine [
13]. Rather, it was postulated that T1 changes may be related to changes in both total water content as well as the relative amounts of water in intracellular and extracellular compartments. Furthermore, changes in sodium and potassium contents and their distribution may impact on the motional freedom of water protons, contributing to prolongation of T1 values in ischemic tissue [
13]. Indeed, both of these processes are part of edema evolution in ischemic injury to the myocardium [
1]. Thus, while increased T1 values in acutely injured myocardium may largely reflect water characteristics, there may be additional changes that are more readily detected by T1-mapping and to which T2w methods are less sensitive. The precise nature of what these changes are that prolong T1 values will need to be further investigated, but ShMOLLI T1-mapping is a distinct yet complementary technique to T2w-CMR, which adds to our current repertoire of available sequences for assessing acute myocardial injury.
T1-mapping holds potential to quantify both extent and severity of disease and may provide further insights into conditions where T2w-CMR has proven useful. These include quantifying myocardial salvage in acute MI, determining the age of myocardial injury, diagnosing acute myocarditis and cardiac transplant rejection. It may also be useful in describing subacute conditions where myocardial inflammation is implicated, such as cardiac sarcoidosis [
26], amyloidosis, and hypertrophic cardiomyopathy [
6]. T1-mapping has also been shown to detect changes in chronic cardiac conditions involving diffuse fibrosis [
27‐
29], which is characterized by increased collagen and aqueous content within the myocardium, leading to interstitial expansion.
Unlike the T2w methods examined, ShMOLLI T1-mapping does not require
a-priori knowledge of where affected and unaffected myocardium (if any) should be for image analysis; it reports absolute T1 values, obviating the need for a reference ROI, and thus has an excellent performance in different models of edema, whether regional or global. Accordingly, T1-mapping can be useful in pathologies where edema maybe diffuse, patchy or not immediately obvious to the eye. Immediate visualization of the extent of injury may be facilitated by dedicated color schemes based on quantitative thresholds, an example of which is shown in Figure
2.
Quality and reliability of T1-maps
As our T1-mapping method is novel, we aimed to be as accurate as possible in setting the most optimal cut-off value for detecting edema and thus were stringent on selecting only the best segments (without any or with only minimal artifacts) to include in the analysis. This was one of the reasons for the apparently higher rejection rate for T1-mapping compared to other modalities. T1-maps were assessed for quality and reliability in three ways: visual inspection of the T1-map itself, visual inspection of all seven of the raw T1 images for potential artifacts, and the use of R
2 maps to ensure that the T1 fit is robust for all segments on the T1-map. The feature of R
2 maps is described in detail with examples in the Additional file
1, and serve as a more objective way to detect segments with poor T1 fit that may not be obvious to the eye; T2-weighted methods tested in this study, on the other hand, do not have this added feature for quality assurance and so their rejection rate may be underestimated. We consider the fact that ShMOLLI T1-mapping has multiple ways for detecting regions with potentially compromised T1 values a strength of the method which makes it more robust, especially to observer bias. ACUT2E is also a relatively novel method that has been validated in acute MIs [
3], but not in detecting global edema or non-ischemic pathologies; sometimes, it is unclear whether darker areas within the myocardium are due to SSFP artifacts (as in Figure
2B, middle panel) or whether it is representative of normal myocardium (as in Figure
2B (right panel) versus signal drop-out/artifact in the inferolateral wall (less likely in this case). These reasons contribute to the apparently higher rejection rate of T1-mapping, but do not reflect that the method is artifact-prone; rather, we were strict on selecting only the best segments for the purpose of establishing T1 thresholds with the greatest accuracy.
As presented in the Results section, to check for bias against any of the methods tested, when we re-included all rejected segments affected by artifacts, it was reassuring to note that all of the original relative relationships on the ROC analysis were preserved, with T1-mapping maintaining its superior diagnostic performance with a significantly larger AUC of 0.87 compared to the T2-weighted methods (Figure
5). Relative relationships amongst methods were also preserved for ROC analyses according to patient groups (Additional file
1: Figure 6A and B).
Limitations
ShMOLLI T1-mapping sometimes demonstrated typical SSFP susceptibilities to off-resonance artifacts and partial volumes effects, especially in very apical slices. These may be circumvented with frequency shifting and potentially by wide-band SSFP. As artifacts may cause falsely high or low T1 values, it is important to have methods for assessing the quality and robustness of T1-maps, such as those used in this study (see Additional file
1). ShMOLLI underestimates the true T1 value by 4 % but in a consistent manner [
16]. Thus, when comparing T1 values measured by different T1-mapping techniques, corrected T1 values and thresholds should be used.
As discussed, it is challenging to establish an in-vivo and histopathological reference gold standard for acute myocardial edema; thus, this study used acute regional wall motion abnormality without evidence of infarction as the best clinical surrogate for edema in a carefully selected study population. Pre-contrast increase in myocardial T1 value is non-specific and may be seen in a number of conditions, including acute myocardial injury, chronic infarct scar [
14] and cardiac amyloidosis [
30], and thus must be interpreted within the clinical context. Moreover, changes of T1 and T2 within the myocardium are dependent on different mechanisms, but the two are closely related. The precise mechanisms leading to increased T1 relaxation times in acutely injured myocardium remain to be established.
Thirdly, T2-mapping is another technique which may be used to detect edema [
31] but was not studied in this work. We did, however, compare T1-mapping to the state-of-the-art bright-blood ACUT2E sequence [
10], which has recently been shown to be superior to conventional dark-blood STIR imaging in acute MI [
3]. Future work will have to directly compare T1 and T2 mapping techniques.