The study had institutional review board approval, followed the Declaration of Helsinki principles. Patients were retrospectively selected from a myocarditis registry at our institution, with retroactive written informed consent according to institution regulations. Consecutive patients were included, who underwent baseline CMR within 7 days of the initial clinical episode and follow-up CMR at 90 (± 15) days, with a study period between January 2016 and December 2019. All participants were referred to our center with a first episode of suspected acute myocarditis. Patients underwent a diagnostic work-up in line with the latest guidelines of the European Society of Cardiology, including a 12-lead electrocardiogram at presentation, and met the updated Lake Louise Criteria for a CMR-based diagnosis of acute myocarditis [7, 8]. Inclusion criteria required positive LGE on baseline CMR. Coronary artery disease was ruled out in all participants either through invasive coronary angiography, CT coronary angiography, or based on a low pretest probability (< 30 years of age). Medical records were scrutinized for underlying comorbidities that could potentially impact LV myocardium, leading to the exclusion of patients with a history of prior myocarditis, congenital heart disease, cardiac surgery, reduced LV ejection fraction, known cardiomyopathies, and systemic conditions with potential myocardial involvement, such as collagen disorders, eosinophilic disorders, sepsis, tuberculosis, or malignancies undergoing chemotherapy. The study population is outlined in Fig. 1.

The studies utilized magnetic resonance scanners, either a 1.5 Tesla (Achieva, Philips Medical Systems) or a 3.0 Tesla (Skyra, Siemens), equipped with dedicated phased array coils for cardiac examinations. All acquisitions were breath-held in end-expiration. Cine-balanced steady-state free precession images assessed cardiac function and geometry in standard orientations. Native T1 and T2 Mapping, along with T2-weighted images of the LV with and without fat suppression, were obtained in short axis. LGE images were acquired 10 minutes postadministration of 0.2 mmoL/kg gadobutrol (Gadovist, Bayer Schering Pharma), following standard orientations. The optimal inversion time for LGE was determined individually using an inversion time scout sequence, ranging between 190 and 270 ms. For 1.5 Tesla CMR studies, a three-dimensional (3D) LGE sequence was employed (field of view 350 × 350 mm2, matrix dimensions of 256 × 256, repetition time/echo time of 3.6/1.8 ms, a flip angle of 15°, in-plane resolution of 1.5 × 1.5 mm2, and a slice thickness of 8 mm). At 3.0 Tesla, a two-dimensional (2D) Phase-Sensitive Inversion-Recovery sequence was used (field of view 330 × 330 mm2, matrix dimensions of 192 × 256, repetition time/echo time of 663/2 ms, a flip angle of 20°, in-plane resolution of 1.3 × 1.3 mm2, and a slice thickness of 8 mm). The CMR diagnosis of myocarditis was established based on regional or global myocardial edema, defined by increased native T2-signal or T2-mapping values, along with evidence of myocardial injury indicated by elevated native T1-mapping values, increased extracellular volume, or positive LGE.

LGE involvement in the LV was assessed using commercial postprocessing software (Intellispace Portal, Version 10, Philips Healthcare). The quantitative evaluation involved manually delineating LV endo- and epicardial boundaries on short-axis LGE images. Subsequently, various clinically available LGE quantification techniques were applied. These included semiautomated methods, where regions of interest (ROIs) were initially placed in both non-enhanced areas and the seemingly normal remote myocardium. In the next step, the hyper-enhanced myocardial regions were segmented in several ways: automatically using the full width at half maximum (FWHM) method or relative to gray-scale thresholds set at predefined values, 5 and 6 standard deviations above the mean signal intensity for the remote myocardial tissue (referred to as the 5 SD and 6 SD techniques), or by manually adjusting the threshold by the reader to achieve the best match (referred to as visual assessment with a user-defined threshold, VAT). Lastly, a full manual (FM) quantification method was employed, where the reader manually delineated the extent of LGE on each LV slice. The extent of LGE was calculated as a percentage relative to the entire LV myocardial volume. Change in LGE extent during follow-up was defined as a ≥ 1% difference in LGE volume between baseline and follow-up relative to the entire LV volume. This cut-off was based on the previous prognostic study of Aquaro et al, who used the same threshold, and a significant LGE change from 6.2 [IQR: 2; 10] % to 4.1 [IQR: 2; 8] % was found over 6 months follow-up [6].

For semiquantitative LGE assessment, the 17-segment model of the American Heart Association was applied to the entire LV myocardium [14]. LGE presence and transmurality were coded (0 = no LGE, 1 = < 25%, 2 = 26–50%, 3 = 51–75%, 4 = 76–100%), and visual presence score (VPS) and visual transmurality score (VTS) were calculated, each multiplied by the number of affected myocardial segments, resulting in VPS (0–17) and VTS (0–68). Change in LGE extent was defined as the change of VPS or VTS score during follow-up.

For a simplified assessment of LGE change, baseline and follow-up short-axis LGE images were displayed side by side. Readers estimated the percentage LGE difference of the entire LV myocardium, consolidated into a five-point Likert scale: 0 = no change, 1 = 1–25%, 2 = 26–50%, 3 = 51–75%, and 4 = 76–100%, termed the ‘visual change score’ (VCS). Positive values denote LGE progression, while negative values indicate regression.

Baseline and follow-up CMR studies were independently evaluated by three observers (M.K., M.P., J.M.S.) with over 5 years of CMR experience and Level 3 certification from the European Association of Cardiovascular Imaging or equivalent authority. Consensus readings were used to estimate the correlation between each technique. To test interobserver reliability, one observer (M.K.) repeated measurements with a 6-month interval.

Continuous variables are presented as median and interquartile range (IQR). Categorical variables are expressed as counts and percentages. The Wilcoxon signed-rank test compared patient characteristics between baseline and follow-up for continuous variables, and the McNemar test compared categorical data. Friedman’s test assessed the extent of LGE among different quantitative methods. Spearman’s correlation evaluated the relationship between LGE assessment techniques, with correlation coefficients (ρ) indicating a very strong (≥ 0.70), strong (0.40 to < 0.70), moderate (0.30 to < 0.40), weak (0.20 to < 0.30), and no or negligible (0.01 to < 0.20) relationship [15, 16]. Intra- and interobserver repeatability were assessed using intraclass correlation (ICC), categorized as poor (< 0.5), moderate (0.5 to < 0.75), good (0.75 to < 0.9), or excellent (0.9–1.0) agreement [17]. A 2-sided p < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS (version 23).