Diagnostic performance of cardiovascular magnetic resonance native T1 and T2 mapping in pediatric patients with acute myocarditis

After obtaining approval by the Institutional Review Board at Ann & Robert H. Lurie Children’s Hospital of Chicago, a retrospective chart review was performed. Data from patients <21 years with a diagnosis of acute myocarditis and a clinical CMR study with native T1 and T2 performed from 2014 to 2017 were included. In the absence of a gold standard diagnostic test and similar to both a multi-center retrospective pediatric study [17] and a comprehensive adult meta-analysis [15], clinical diagnosis of acute myocarditis was used for study inclusion. Clinical diagnosis was determined by experienced clinicians and based on a presentation that included acute chest pain, viral respiratory symptoms, elevated troponin, recent infectious symptoms, and/or depressed systolic function on echocardiogram. All patients underwent CMR within 4 weeks of diagnosis of myocarditis.

CMR exams that included myocardial mapping data from age matched healthy pediatric subjects served as the control group. Typical control subjects were healthy patients who underwent CMR to assess for a coronary anomaly (n = 9) or cardiomyopathy/arrhythmogenic right ventricular cardiomyopathy (n = 30). All controls had normal cardiac structure and normal biventricular size and global systolic function determined by echocardiogram and CMR, no evidence of late gadolinium enhancement (LGE) determined by CMR, and no underlying systemic illnesses known to alter mapping values. Patients with a previous diagnosis of myocarditis, chronic cardiac disease, congenital heart disease, or an obvious alternative diagnosis (e.g. myocardial infarction) were excluded.

All CMR studies were performed on a 1.5 T CMR scanner (Siemens Healthineers, Erlangen Germany). The CMR protocol was adapted from the previously published myocarditis consensus paper [3]. The protocol consisted of balanced steady state free precession (bSSFP) cine images in ventricular short- and long-axis views to assess for cardiac size and systolic function. Volumetric and functional data analysis was performed on dedicated software (Q Mass, Medis Medical Imaging Systems, Leiden, The Netherlands). T2-weighted (T2W) short-tau triple inversion-recovery (STIR) short axis to assess for edema, T1-weighted (T1W) turbo-spin-echo long axis imaging sequences pre and post contrast to assess for early gadolinium enhancement (EGE), and LGE were performed. Using the established LLC, myocardial edema was defined as a signal intensity ratio of myocardium relative to skeletal muscle ≥2.0 on T2W imaging [3]. We performed early gadolinium enhancement (EGE) imaging and LGE imaging 3 and 10–15 min, respectively. After the administration of contrast, 0.2 mmol/kg gadopentetate dimeglumine (Magnevist, Bayer HealthCare, Whippany, New Jersey, USA) or 0.15 mmol/kg of gadobutrol (Gadavist, Bayer HealthCare). EGE was defined as a signal intensity ratio of myocardium relative to skeletal muscle ≥4.0 on post-contrast images or an absolute increase in myocardial enhancement of ≥45% [3]. The presence of LGE was confirmed by at least two experienced observers at the time of the original study interpretation. General anesthesia was utilized as clinically necessary per our clinical protocol.

Myocardial native T1 maps were obtained using a free-breathing motion-corrected, electrocardiogram (ECG)-triggered, modified Look-Locker inversion recovery (MOLLI) sequence with images acquired at end-diastole before and 12 min after contrast injection in the short axis plane at the base, mid cavity, and apical LV myocardium, similar to previous protocols [18]. While acquisition times varied depending on the patient’s heart rate, 11 s was typically needed for each breath hold. Sequence parameters were optimized by adjusting the number of recovery heartbeats based on the patient’s heart rate to allow for adequate T1 recovery between inversion pulses (Table 1). In order to compensate for loss of spatial resolution due to cardiac motion at high heart rates, we optimized spatial resolution by reducing in-plane pixel size. Detailed T1 MOLLI parameters are provided in Table 1. T1 values were obtained using post-processing software (Medis Medical Imaging). Endocardial and epicardial contours were traced on each pre and post contrast image to exclude blood, epicardial fat, or artifact. The myocardium was divided into 16 segments as standardized by the American Heart Association (AHA) [19]. T1 values pre- and post-contrast for each myocardial segment were averaged to give a mean global T1 value. ECV was calculated using the following equation: ECV = λ × (1 − hematocrit), where the partition coefficient λ = ΔR1 (myocardium)/ΔR1 (blood) and ΔR1 is the change in signal intensity between pre- and post-contrast images [20].

T2 mapping was performed using a dark blood turbo spin echo sequenced with a T2 preparation pulse and a bSSFP readout in the short-axis plane at the base, mid-chamber, and apex at end-diastole during free breathing with motion correction [21]. Detailed T2 imaging parameters are provided in Table 1. T2 mapping echo times were 0 ms, 24 ms, and 55 ms and did not vary by heart rate. Images were motion corrected. T2 maps were acquired prior to contrast administration. T2 values were obtained using post-processing software (Medis Medical Systems) and endocardial and epicardial contours were manually traced on each image to exclude blood, epicardial fat, or artifact. T2 values for each myocardial segment were recorded and each value was averaged to provide a mean global T2 value. Maximum segment T2 values were also recorded, similar to the methodology from adult studies [11, 22].

All descriptive patient values are reported as percent distributions and continuous variables as mean +/− standard deviation or median values with interquartile ranges. Two sample independent t-tests were used to compare normally distributed data and Wilcoxon rank sum tests used for non-normal data. Receiver operating curves (ROC) were generated for T1, T2, and ECV and the Youden index was used to determine ideal upper limit cut-off values for distinguishing patients with myocarditis from healthy patients, with their own respective sensitivity and specificity. The predicted probabilities of logistic regression were used to create ROC curves of the combined mapping parameters. A p-value < 0.05 was considered statistically significant. Data were analyzed in SAS 9.4 (SAS Institute Inc., Cary, North Carolina, USA).

A total of 23 patients with a clinical diagnosis of myocarditis and 39 healthy controls met the inclusion criteria. The median age at the time of CMR for the myocarditis patients was 15.2 years (range, 0.6–18.8) and for the controls was 16.3 years (range, 10.5–18.3, p = 0.054). CMR was performed on average 4 days (range, 1–26) after admission and all CMR studies were performed within 11 days of diagnosis aside from one (26 days). Patient demographic and clinical data are summarized in Table 2. On presentation, patients demonstrated arrhythmias in 17% of cases, required inotropes 26% of the time, and all but one patient (96%) had an elevated troponin. Of note, there was no difference in left ventricular ejection fraction by CMR between patients with myocarditis (54.5% ± 7.3) and controls (57.0% ± 4.8, p-value 0.17).

Mean global values for all mapping parameters were significantly elevated in patients with clinically suspected acute myocarditis compared to controls, with native T1 1098 ± 77 vs 990 ± 34 ms, T2 52.8 ± 4.6 ms vs 46.7 ± 2.6 ms, and ECV 29.8 ± 5.1% vs 23.3 ± 2.6% (all p-values < 0.001; Table 3). For each parameter, we established ideal cut off values from corresponding ROC curves to distinguish patients with myocarditis from controls, and to assess sensitivity and specificity. For global native T1 the ideal cutoff value was 1015 ms (AUC 0.936, sensitivity 91%, specificity 86%), for global T2 48.5 ms (AUC 0.908, sensitivity 91%, specificity 74%); and for ECV 25.9% (AUC 0.918, sensitivity 86%, specificity 89%) (Table 4, Fig. 1). Regional T2 mapping values were also significantly higher for patients with myocarditis as compared to controls. The maximum segmental T2 value for myocarditis patients was 61.2 ± 7.0 vs 52.4 ± 4.2 ms for controls (p-value < 0.0001). Finally, we created ROC curves using a combination of mapping parameters. Combining the global native T1 and global T2 parameters yielded an AUC 0.953 (sensitivity 83%, specificity 96%) while combining the global native T1 and maximum segmental T2 (MaxT2) parameters yielded an AUC of 0.947 (sensitivity 87%, specificity 93%) (Table 4, Fig. 1).

We found only 57% (13/23) of patients with a clinical diagnosis of acute myocarditis met LLC. Broken down by each criterion, 86% had LGE, 57% had abnormal T2W imaging (edema), and 13% had abnormal EGE (hyperemia) (Table 2). Using the ROC curve generated cut-off values, ten patients with a clinical diagnosis were missed by the LLC. Of these ten, nine had at least one abnormal mapping value, seven demonstrated abnormal values across all three parameters, and two patients demonstrated abnormal maximum T2 values without any LGE (Fig. 2). Of particular interest, we identified two patients who demonstrated normal findings across all three of the LLC but who demonstrated abnormal mapping data across all three global mapping parameters. One of these patients exhibited markedly abnormal tissue mapping values, with global T1 1270 ms, global T2 60 ms, segmental max T2 of 66 ms and ECV 39.6%, and no perceptible LLC criteria (Fig. 3).

Our single center study using one CMR scanner is limited by a small sample size. The use of clinical criteria for the diagnosis is not an ideal substitute for a gold standard test but is a limitation of other similar diagnostic cohort and case control studies of myocarditis. Our study demonstrated overall lower CMR and global mapping values for healthy controls than previously published adult studies [8, 9, 15, 18, 21]. Some of this may due to the technical differences of obtaining CMR in children, who have more heart rate variability and often require anesthesia. Additionally, while consensus guidelines note these differences are expected, variable sequence parameters, post-processing techniques and vendor software, and possibly naturally occurring lower mapping values in children compared to adults all limit the generalizability of our data to other centers [5, 9, 10, 16].