Comprehensive Cardiac Magnetic Resonance to Detect Subacute Myocarditis

Abstract (1) Background: Compared to acute myocarditis in the initial phase, detection of subacute myocarditis with cardiac magnetic resonance (CMR) parameters can be challenging due to a lower degree of myocardial inflammation compared to the acute phase. (2) Objectives: To systematically evaluate non-invasive CMR imaging parameters in acute and subacute myocarditis. (3) Methods: 48 patients (age 37 (IQR 28–55) years; 52% female) with clinically suspected myocarditis were consecutively included. Patients with onset of symptoms ≤2 weeks prior to 1.5T CMR were assigned to the acute group (n = 25, 52%), patients with symptom duration >2 to 6 weeks were assigned to the subacute group (n = 23, 48%). CMR protocol comprised morphology, function, 3D-strain, late gadolinium enhancement (LGE) imaging and mapping (T1, ECV, T2). (4) Results: Highest diagnostic performance in the detection of subacute myocarditis was achieved by ECV evaluation either as single parameter or in combination with T1 mapping (applying a segmental or global increase of native T1 > 1015 ms and ECV > 28%), sensitivity 96% and accuracy 91%. Compared to subacute myocarditis, acute myocarditis demonstrated higher prevalence and extent of LGE (AUC 0.76) and increased T2 (AUC 0.66). (5) Conclusions: A comprehensive CMR approach allows reliable diagnosis of clinically suspected subacute myocarditis. Thereby, ECV alone or in combination with native T1 mapping indicated the best performance for diagnosing subacute myocarditis. Acute vs. subacute myocarditis is difficult to discriminate by CMR alone, due to chronological connection and overlap of pathologic findings.


Introduction
The diagnosis of subacute myocarditis remains challenging due to various reasons. In contrast to acute myocarditis, patients with subacute myocarditis may present with rather mild or non-specific symptoms. Primary clinical workup may reveal non-pathologic ECG, preserved or just slightly impaired left-ventricular function, as well as more discrete laboratory findings than in the acute phase of inflammation [1,2].
In clinical practice, CMR has emerged as a diagnostic tool to confirm clinically suspected myocarditis due to its ability of non-invasive multiparametric tissue characterization [2][3][4][5][6][7]. However, as myocardial edema decreases and CMR lesions become more diffuse, the decrease of myocardial inflammation during the transition from the acute to the subacute phase still poses a challenge for diagnosing subacute myocarditis [8][9][10][11][12].
Yet, the correct diagnosis is of high importance at this stage of myocarditis and might be crucial regarding patients' recovery. Patients may both need general supportive therapy, heart failure medication and abstinence from competitive sports in order not to risk a transition to chronic myocarditis or even dilated cardiomyopathy (DCM), considerably worsening the course [13,14].
The purpose of this study was to systematically evaluate non-invasive CMR parameters for detection of subacute myocarditis and to test their diagnostic performance compared to acute myocarditis and healthy controls.

Study Population
In this single center study, 332 consecutive patients with clinically suspected myocarditis underwent CMR imaging from October 2019 to May 2022 and were prospectively evaluated. N = 265 patients were excluded due to a symptom duration >6 weeks or other final CMR diagnosis of non-ischemic cardiomyopathy. N = 14 patients were excluded due to a history of coronary artery disease, angiographic evidence of coronary artery disease (CAD) or pre-existing valve disease. N = 5 patients were excluded due to incomplete dataset acquisition. Datasets of n = 48 myocarditis patients (age 37 (IQR 28-55) years; 52% female) were finally evaluable.
Additional endomyocardial biopsy for reference standard diagnosis confirmation was performed according to current clinical indications [15] in 6 patients of the acute group and 2 patients of the subacute group within median 1 day of CMR.
Patients' symptoms, cardiovascular risk profiles and laboratory values were recorded; 15 healthy volunteers served as a control group (Supplementary File S1). All subjects gave written informed consent, and the Institutional Review Board approved the study protocol.

CMR Image Aqcuisition
CMR examinations were performed on a 1.5T scanner (MAGNETOM Aera, SIEMENS Healthcare, Erlangen, Germany). CMR protocol comprised late gadolinium enhancement (LGE) imaging and mapping (T 1 , ECV, T 2 ) as well as additional detection of pericardial effusion and acquisition of morphology, volumetry, and strain. For functional assessment, steady state free precession (SSFP) CINE loops in vertical and horizontal long-axis as well as short-axis orientation were performed. T 1 mapping was performed native and 15-20 min post contrast agent administration with a MOLLI sequence 5(3)3 with generation of 3 short axis T 1 maps (apical, mid, basal). T 2 mapping was performed before contrast media application with a T 2 prepared SSFP sequence in 3 short axis slices (apical, mid, basal) with 2D inversion recovery (IR) gradient recovery echo (GRE) sequence for late enhancement imaging 10 min after intravenous administration of 0.15 mmol Gadobutrol (Gadovist, Bayer Healthcare, Leverkusen, Germany) per kg body weight.
Detailed sequence parameters for CMR imaging are given in detail in Supplementary File S1.

CMR Image Analysis
Image analysis was performed by two experienced CMR readers in consensus using dedicated software (cvi42 Version 5.13, CVI Circle Cardiovascular Imaging, Calgary, AB, Canada) and according to the Society for Cardiovascular Magnetic Resonance (SCMR) recommendations [16,17]. Functional assessment was performed in a stack of SAX slices with semi-automated contouring of endocardial and epicardial borders, with manual re-adjustment if necessary; cutoff values according to [18]. 3D-Strain analysis for global radial (GRS), circumferential (GCS) and longitudinal strain (GLS) was performed using post-processing CMR feature tracking of 4CV and SAX cine loops; cutoff values according to [19].
LGE imaging was evaluated qualitatively and semi-quantitatively: LGE patterns (linear vs. patchy) were qualitatively assessed and localized (septal mid-myocardial vs. subepicardial) and assigned to the myocardial segments according to the 17 segment-model of the American Heart Association [20]. Semi-quantitative evaluation of LGE fraction of left-ventricular (LV) myocardial mass was performed with a threshold of ≥2 standard deviations (SD) above remote myocardium [16].
T 1 and T 2 mapping was evaluated in a segmental and global approach; values were considered elevated if above 2 standard deviations of a healthy in-house control group performed at the same 1.5T scanners (T 1 > 1053; T 2 > 51 ms). For descriptive statistics, ECV values above 30% were considered as definitely increased [21][22][23]. For diagnostic evaluation, criterions derived from receiver-operating (ROC) curves were used as cutoff values. A combination of 2018 expert recommendations for updated CMR criteria in acute and subacute myocardial inflammation ('Lake Louise criteria'; 2018 LLC) were evaluated in each case: (1) myocardial edema (elevated T 2 ) and (2) non-ischemic injury (elevated T 1 , and/or ECV, and/or LGE) [4].

Endomyocardial Biopsy Protocol
Endomyocardial biopsies were performed in selected patients according to current ESC diagnostic guidelines [5]. At least five right-ventricular biopsies were taken followed by a comprehensive cardiopathological workup. For details see Supplementary File S1.

Statistical Analysis
Normality of data was tested using the Kolmogorov-Smirnov test. Continuous, nonparametric variables are indicated as median (interquartile range). Categorical data are indicated as frequency (percentage %). For unpaired group comparison Mann-Whitney U test was performed in continuous nonparametric data; Fisher's exact test was performed in categorical data (JMP, Version 16, SAS Institute Inc., Heidelberg, Germany). Receiver operating characteristic (ROC) curves were generated for comparison of LGE and mapping parameters in patients with acute and subacute myocarditis as well as in controls using the method of Delong et al. [24] (MedCalc, Version 18, MedCalc Software Ltd., Ostend, Belgium). Global level of significance α was set to 5%. Local level of significance (α loc ) for each test with dependent variables was corrected according to the Bonferroni equation according to k = 85 performed comparisons: α loc = α glob /k = 0.0006. Table 1 depicts the study's patient characteristics. All patients were symptomatic (at least 1 symptom) at the time of diagnostic work-up: 20 (42%) demonstrated dyspnea, 17 (35%) chest pain, 16 (33%) fever, 16 (33%) fatigue, 12 (25%) angina pectoris and 2 (4%) peripheral edema. About one quarter of all patients (n = 11, 23%) were ≥NYHA III. One female patient of the acute group had COVID-19.

Subacute Myocarditis vs. Controls
For discrimination of subacute myocarditis from healthy controls, LGE imaging and ECV mapping demonstrated the highest AUCs with 0.96 (p < 0.0001) for LGE and 0.90 (p < 0.0001) for ECV; T 2 and T 1 mapping performed slightly inferior 0.79 (p < 0.001) for T 2 and 0.76 (p = 0.002) for T 1 . AUCs revealed a criterion of >1015 ms for T 1 and of >49 ms for T 2 , Figure 1.  In the discrimination of subacute myocarditis from healthy controls, LGE and ECV performed best with AUCs of 0.96 (p < 0.0001) and 0.90 (p < 0.0001) respectively; 0.79 (p < 0.001) for T 2 with a criterion of >49 ms; 0.76 (p = 0.002) for T 1 with a criterion of >1015 ms. (C) For comparison of acute from subacute myocarditis, the areas under the curve (AUCs) were 0.76 (p < 0.001) for LGE with a criterion of >2, 8% of LV myocardial mass; 0.66 (p = 0.049) for T 2 with a criterion of >51 ms; T 1 and ECV showed no significant differences. The diagonal line course indicates difficult discrimination of acute vs. subacute myocarditis by CMR alone.
The best diagnostic performance in the detection of subacute myocarditis and the discrimination from healthy controls was achieved by both ECV evaluation alone or in combination with T 1 mapping (applying a segmental or global increase of native T 1 > 1015 ms and ECV > 28%), demonstrating a sensitivity of 96% (CI 78-100) and an accuracy of 91% (CI 77-98), see Figure 2. LGE with a criterion of >2, 8% of LV myocardial mass; 0.66 (p = 0.049) for T2 with a criterion of ms; T1 and ECV showed no significant differences. The diagonal line course indicates difficult d crimination of acute vs. subacute myocarditis by CMR alone.
The best diagnostic performance in the detection of subacute myocarditis and discrimination from healthy controls was achieved by both ECV evaluation alone or combination with T1 mapping (applying a segmental or global increase of native T1 > 10 ms and ECV > 28%), demonstrating a sensitivity of 96% (CI 78-100) and an accuracy 91% (CI 77-98), see Figure 2.

Figure 2. Diagnostic Performance of CMR Criteria Combination for Discrimination of Subac
Myocarditis from Healthy Controls. The best diagnostic performance in the detection of subac myocarditis and the discrimination from healthy controls was achieved by both ECV evaluat alone or in combination with T1 mapping, demonstrating a sensitivity of 96% (CI 78-100) and accuracy of 91% (CI 77-98). A segmental or global increase of native T1 > 1015 ms and ECV > 2 was applied, derived from ROC analysis. LLC = Lake Louise Criteria.

Figure 2. Diagnostic Performance of CMR Criteria Combination for Discrimination of Subacute
Myocarditis from Healthy Controls. The best diagnostic performance in the detection of subacute myocarditis and the discrimination from healthy controls was achieved by both ECV evaluation alone or in combination with T 1 mapping, demonstrating a sensitivity of 96% (CI 78-100) and an accuracy of 91% (CI 77-98). A segmental or global increase of native T 1 > 1015 ms and ECV > 28% was applied, derived from ROC analysis. LLC = Lake Louise criteria. T 1 mapping had a sensitivity of 100% (CI 85-100) with lowest specificity of 50% (CI 21-79). T 2 mapping showed a sensitivity of 87% (CI 66-97) and an accuracy of 83% (66-93); 2018 expert recommendations (LLC) resulted in a sensitivity of 87% (CI 66-97) and an accuracy of 86% (CI 70-95). LGE had the lowest sensitivity of 61% (CI 39-80), but the highest specificity with 100%. Diagnostic performances of CMR parameters are depicted in Table 3 and Figure 2.

Acute Myocarditis vs. Subacute Myocarditis
For discrimination of acute from subacute myocarditis, the areas under the curve (AUCs) were 0.76 (p < 0.001) for LGE with a criterion of >2.8% of LV myocardial mass; 0.66 (p = 0.049) for T 2 with a criterion of >51 ms. T 1 and ECV did not differ significantly, as shown in Figure 1. Typical CMR examples of acute and subacute myocarditis are illustrated in Figure 4.

Acute Myocarditis vs. Subacute Myocarditis
For discrimination of acute from subacute myocarditis, the areas under the curve (AUCs) were 0.76 (p < 0.001) for LGE with a criterion of >2.8% of LV myocardial mass; 0.66 (p = 0.049) for T2 with a criterion of >51 ms. T1 and ECV did not differ significantly, as shown in Figure 1. Typical CMR examples of acute and subacute myocarditis are illustrated in Figure 4.

Discussion
This prospective study systematically evaluated non-invasive CMR imaging parameters for the detection of subacute myocarditis compared to acute myocarditis and healthy controls. Clinically suspected subacute myocarditis could be sensitively detected by (1) ECV evaluation alone or (2) in combination with T1 mapping or (3) T2 mapping.

Subacute Myocarditis vs. Controls
ECV and native T1 mapping. Regarding the detection of subacute myocarditis, both ECV alone or in combination with T1 mapping yielded the best sensitivity (96%) and accuracy (91%). Radunski et al. reported the best diagnostic accuracy for global myocardial ECV >27% in diagnosis of acute myocarditis [25]; Luetkens et al. applied an ECV cutoff of 28.8% achieving an accuracy of 74% in diagnosing myocarditis [26]. This is in line with the ECV cutoff of 28% for subacute myocarditis which we found in our study. ECV has shown to be capable of detecting subtle myocardial alternations including fibrosis and therefore is especially beneficial when LGE is not attainable [25,27]. Furthermore, ECV has proven to be an independent CMR parameter robustly associated with outcome in myocardial fibrosis [28], and to be the best imaging biomarker of acute myocarditis burden also in dual-energy computed tomography (DECT), allowing an early prediction of the occurrence of cardiac complications [29,30].

Discussion
This prospective study systematically evaluated non-invasive CMR imaging parameters for the detection of subacute myocarditis compared to acute myocarditis and healthy controls. Clinically suspected subacute myocarditis could be sensitively detected by (1) ECV evaluation alone or (2) in combination with T 1 mapping or (3) T 2 mapping.

Subacute Myocarditis vs. Controls
ECV and native T 1 mapping. Regarding the detection of subacute myocarditis, both ECV alone or in combination with T 1 mapping yielded the best sensitivity (96%) and accuracy (91%). Radunski et al. reported the best diagnostic accuracy for global myocardial ECV >27% in diagnosis of acute myocarditis [25]; Luetkens et al. applied an ECV cutoff of 28.8% achieving an accuracy of 74% in diagnosing myocarditis [26]. This is in line with the ECV cutoff of 28% for subacute myocarditis which we found in our study. ECV has shown to be capable of detecting subtle myocardial alternations including fibrosis and therefore is especially beneficial when LGE is not attainable [25,27]. Furthermore, ECV has proven to be an independent CMR parameter robustly associated with outcome in myocardial fibrosis [28], and to be the best imaging biomarker of acute myocarditis burden also in dual-energy computed tomography (DECT), allowing an early prediction of the occurrence of cardiac complications [29,30].
Segmental or global elevation of T 1 values above the cutoff of 1015 ms showed an excellent sensitivity of 100%, however, with only low specificity of 50% (Table 3). Specificity improved to 83%, additionally including ECV. Moreover, the combination of T 1 mapping with ECV has shown high sensitivity (96%) and accuracy (91%) with an additional benefit of diversification compared to ECV evaluation as a single parameter. T 1 elevation above 990 ms has been previously proposed for detection of inflammation in acute myocarditis with good diagnostic performance [31,32]. In our study, this cutoff would have resulted in a substantial rise of the false positive rate and a decrease of the positive predictive value. T 2 mapping. Evaluation of T 2 showed lower diagnostic performance in single parameter analysis compared to T 1 and ECV with a sensitivity of 87% and accuracy of 83%, applying a segmental or global increase of >49 ms for T 2 . Performing ECV in addition to T 2 improved the specificity. This result supports the approach that T 1 -based ECV and T 2 mapping have complementary diagnostic value as reflected by current expert recommendations for the diagnostic management of patients with suspected myocarditis [33,34].
2018 Expert Recommendations (Lake Louise criteria). Application of 2018 LLC resulted in a sensitivity of 87% and an accuracy of 86% and thus provided less accurate diagnosis compared to the duo of ECV + T 1 mapping. However, 2018 LLC showed decent diagnostic performance, considering being designed for detection of myocarditis in the acute and subacute phase [35].
LGE had the lowest single parameter detection rate with a sensitivity of 61% and an accuracy of 74%. Lagan et al. reported similar compiled sensitivity and accuracy for LGE in myocarditis of 63 and 72%, respectively [7]. A metaanalysis by Kotanidis et al. also stated comparable numbers with a sensitivity of 68% [36]. Considering the presented superior sensitivity of mapping techniques, the question arises whether implication of LGE is of additional benefit and whether LGE may also be omitted in a comprehensive CMR protocol in suspected subacute myocarditis. Hereby, two important upsides of LGE are of note: First, LGE with its characteristic non-ischemic patterns showed the highest specificity in this study in line with previous studies [36]. Second, occurrence of LGE has also shown to be of predictive value for major adverse cardiac events and outcome [37]. Hereby, especially septal LGE is associated with worsening of the course compared to subepicardial lateral LGE [38][39][40][41].

Acute vs. Subacute Myocarditis
Pathological ECG findings and blood results were less conspicuous and less common in patients with subacute myocarditis. Pericardial effusion, elevated T 2 values as well as LGE prevalence were more frequent and more pronounced in the acute myocarditis group. The AUCs of LGE and T 2 showed the best discrimination of both groups with an LGE extent > 2.8% of LV myocardial mass and T 2 > 51 ms. This might be explained by a decrease of inflammation from the acute to the subacute phase [42]. T 1 mapping has proven to be a sensitive marker for myocardial disease of different entities [33,43], but seems limited for discrimination of acute from chronic processes [44]. ROC analysis indicated difficult discrimination of acute vs. subacute myocarditis by CMR alone due to an overlap of the pathologic findings.

Limitations
As a limitation of this study, endomyocardial biopsy (EMB) was not performed in all patients, but only for diagnosis confirmation in eight ambiguous cases after a careful risk-benefit analysis according to current indications [15]. The overall sample size is limited. T 1 and T 2 mapping values are applicable for the specific scanner and sequence type used in this study and not generalizable over all vendors, scanners, and sequence types.

Clinical Implications
Many patients with inflammatory cardiomyopathy tend to present with a latency of several weeks since symptom onset. According to 2013 ESC recommendations, EMB is recommended for definite diagnosis of myocarditis [5]. However, in clinical routine, EMB is not always performed due to various limitations (e.g., availability, invasiveness, sampling error) and CMR has emerged as a tool to non-invasively characterize myocardial tissue [3,4,6]. Therefore, the 2013 ESC recommendations are expected to be revised in the next years [5,12].
This study provides CMR data on detection of clinically suspected subacute myocarditis. In view of the fact that CMR has recently been accepted as a workflow for non-invasive confirmation of clinically suspected myocarditis [3,14], this may help to detect ongoing myocardial inflammation.
In contrast, unrecognized inflammation in CMR may falsely underestimate subacute myocarditis and lead to incorrect declaration of myocarditis as cured. As a consequence, subacute myocarditis may progress to chronic myocarditis or dilated cardiomyopathy instead of healing to complete restitutio ad integrum.

Conclusions
A comprehensive CMR approach allows reliable diagnosis of clinically suspected subacute myocarditis. Thereby, ECV alone or in combination with native T 1 mapping indicated the best performance for diagnosing subacute myocarditis.
Acute vs. subacute myocarditis is difficult to discriminate by CMR alone, due to chronological connection and overlap of pathologic findings.