Use of the new Lake Louise Criteria improves CMR detection of atypical forms of acute myocarditis

The purpose of our study was to compare diagnostic performance of old and new Lake Louise Criteria (oLLC and nLLC) among different clinical presentations: infarct-like (IL), cardiomyopathic (CM) and arrhythmic (AR). 102 patients with clinical suspicion of acute myocarditis underwent cardiac magnetic resonance (CMR) on a 1.5 T scanner. Protocol included cine-SSFP, T2-weighted STIR, T2 mapping, early and late gadolinium enhancement and T1 mapping acquired before and after gadolinium administration. The degree of agreement has been calculated with Cohen’s K test. 42 patients also underwent endomyocardial biopsy (EMB). IL onset was present in 54/102 patients, CM in 28/102 and AR in 20/102. nLLC were positive in 58.3% of the patients, while oLLC in 37.9%, k = 0.57 (IC: 0.428–0.713). The degree of agreement between nLLC and oLLC was 0.49 (IC: 0.111–0.876) for AR onset (nLLC positive in 35% vs oLLC in 15%), 0.25 (IC: 0.035–0.459) for CM pattern (nLLC positive in 60.7% vs oLLC 17.9%) and 0.73 (IC: 0.543–0.912) for IL presentation (nLLC positive in 66.7% vs oLLC in 57.4%). Diagnostic accuracy was 75% for both nLLC and oLLC among IL onset, and 41.6% for oLLC vs 66.7% for nLLC, as regards CM clinical presentation. nLLC have improved diagnostic performance of CMR for the diagnosis of acute myocarditis, in particular for atypical clinical presentation.


Introduction
Clinical presentations of acute myocarditis (AM) vary through a wide range of manifestations [1,2], affecting either children [3] or adults [4]. The infarct-like onset, the most common clinical pattern [5] which seems to be mostly correlated to Parvovirus B19 infection [6], is characterized by flu-like symptoms followed by chest pain, typical electrocardiogram (ECG) features, myocardial necrosis biomarkers release (in particular T-troponin) and normal coronary angiogram [7].
In other patients, myocarditis shows a cardiomyopathic presentation with acute or chronic heart failure, systolic impairment with or without left ventricle dilation or with new onset ventricular arrhythmias, high-degree atrio-ventricular block or bundle branch block [2,8]. Other confounding features are the variable clinical course (myocarditis can be acute, subacute, or chronic [9]), the underlying pathological substrate and the aetiology of the disease (viral or bacterial infections [10], toxins or drugs or autoimmune disorders [10][11][12]).
Endomyocardial biopsy (EMB) still represents the gold standard for the diagnosis of acute myocarditis [13], that is based on the Dallas criteria [14], which have been improved by immunohistochemical analysis and Polymerase Chain Reaction (PCR) [15]. However, EMB is an invasive procedure with several limits, such as the lack of standard protocols, sampling errors and the low sensitivity [16][17][18].
In 2009, Friedrich et al. defined the Consensus Criteria for CMR in myocardial inflammation, the so-called Lake Louise Criteria (LLC). According to LLC, AM diagnosis was established with the presence of at least 2 out of 3 of the following CMR features: oedema (with T2-weighted sequences), hyperaemia (with Early Gadolinium Enhancement-EGE) and necrosis or fibrosis (with Late Gadolinium Enhancement-LGE) [19].
New LLC (nLLC) were recently published following the introduction of myocardial mapping and redefined imaging diagnosis according to the combined presence of a T1 criterion (presence of LGE or increased T1 mapping or extracellular volume values) and a T2 criterion (hyperintensity in T2 weighted STIR or increased T2 mapping values) [20].
According to the literature, the sensitivity of old Lake Louise Criteria (oLLC) varied with different clinical presentations (ranging from an 80% for the infarct-like onset to a 40% for the arrhythmogenic pattern [21]).
To the best of our knowledge, the clinical performance of nLLC has not been tested yet in relation to different clinical presentations and in particular for atypical onsets. Our hypothesis was that parametric imaging could recognize subtle or diffuse forms of the disease, by quantitatively measuring T1 and T2 relaxation time variations.
The purpose of our study was therefore to compare the diagnostic performance of old and new LLC among the different clinical patterns: infarct-like (IL), cardiomyopathic (CM) and arrhythmic (AR).
Patient's categorization was based on well-defined clinical features, as previously addressed by our group [21]: • Those who presented with chest pain, increase in T-Troponin serum level (≥ 0.014 μg/ml) and ST-T tract elevation were grouped in the IL onset • Patients with systolic impairment (ejection fraction < 54%), left ventricle (LV) dilation and heart failure symptoms in the absence of ECG ST-T abnormalities were categorized as CM; • Subjects with new onset arrhythmias (atrial fibrillation, ventricular arrhythmias, high-degree atrio-ventricular block or bundle branch block) with or without heart failure symptoms were grouped in the AR pattern.
The categorization of clinical presentations was performed by a blinded radiologist at the time of the database preparation. Local ethic committee approved our study.

CMR protocol
After having obtained written informed consent, CMR was performed on a 1.5 T scanner (Magnetom Avanto, Siemens medical systems, Erlangen, Germany), using body and phased array coils.
For T1-mapping, we used a Modified Look Locker Inversion Recovery (MOLLI) sequence, with a 5(3)3 scheme. MOLLI were acquired in three axial slices (basal, mid-ventricular and apical planes) and in four-chambers view, before contrast agent administration and 15 min after Gadolinium injection (

Images analysis
Images were analysed in blind using a dedicated software (Cvi42, Calgary, Canada) by two radiologists with 2 years and 10 years of experience, respectively. Global and regional ventricular function were assessed using cine-SSFP. Endocardial and epicardial contours were traced on end-diastolic and end-systolic images acquired in short axis planes, in order to determine left ventricular volumes and ejection fraction (EF).
The presence of oedema was evaluated both with T2w-STIR sequences and with T2 mapping. As regards STIR images, a visual assessment of oedema was performed with a signal intensity (SI) > 2 standard deviation (SD) above the remote myocardium. For the semi-quantitative analysis, we used the T2 ratio technique in order to relate the SI of the myocardium to that of the skeletal muscle. We outlined endocardial and epicardial contours in the short axis planes and we drew a Region of Interest (ROI) within a visible skeletal muscle. A ratio ≥ 2 was considered positive for oedema [19]. For T2 mapping sequences, we manually traced endocardial and epicardial contours throughout basal, mid and apical slice, and positioned a reference point in the anterior interventricular junction. Thanks to our dedicated software, we were able to identify T2 values for each AHA segment [22]. Then, we calculated mean slice T2 values oedema for each plane. We considered the presence of oedema for mean slice T2 values of more than 50 ms.
Hyperaemia was evaluated with T1-TSE sequences. We traced endocardial and epicardial contours in axial slices acquired before gadolinium injection and then copied contours in slices acquired immediately after contrast media administration. We also drew a ROI within a visible skeletal muscle in order to calculate EGE ratio (EGEr). Hyperaemia was considered to be positive with an EGEr ≥ 4 or with an absolute increase in myocardial SI of 45% in the contrast-enhanced T1-TSE as compared to the pre-contrast images [20].
As regards necrosis/fibrosis, LGE was identified with a signal intensity > 5 SD [23] above the remote myocardium and a non-ischemic pattern of distribution (subepicardial or mesocardial or patchy enhancement) [19]. T1 mapping analysis was performed tracing endocardial and epicardial contours on the basal, mid and apical slices acquired before and 15 min after Gadolinium injection. A ROI was drawn within the LV cavity in order to obtain T1 value of the blood pool. Finally, a reference point was positioned in the anterior interventricular junction.
Our software identified native T1 values and calculated extracellular volume (ECV) for each AHA segment [22] according to the formula described elsewhere [24]. Haematocrit value was acquired within 24 h before the scan.
Native T1 mapping was considered to be pathological with mean slice T1 values of more than 1050 ms; ECV was identified as pathological with a mean slice percentage ≥ 28%.

Endomyocardial biopsy
42 (42/102, 41.1%) patients also underwent EMB, in accordance with the American Heart Association, the American College of Cardiology and the European Society of Cardiology recommendations [16]. EMB procedure performed in our Institution is explained elsewhere [21].

Statistical analysis
Statistical analysis was performed using Medcalc software Ltd (Ostend, Belgium).
The clinical and demographic features were compared using a two-way Chi-Squared Test. Kruskal-Wallis test was used for comparing ventricular functional parameters between the three population (IL, CM and AR).
Inter-observer agreement and the level of agreement between old and new LLC were assessed with Cohen's k test. Agreement based on k values was interpreted as follows: below 0.4 as poor, between 0.41 and 0.60 as moderate agreement, between 0.61 and 0.80 as substantial agreement, and between 0.81 and 1 as almost perfect. McNemar's test (MNT) was used to compare the diagnostic proportion between oLLC and nLLC among different clinical presentations. Diagnostic accuracy was calculated only in the selected subcohort of patients who underwent EMB, as the proportion between true positive and true negative and is expressed as a percentage followed by Clopped-Pearson Confidence Interval.
The net reclassification improvement (NRI) and the net absolute reclassification improvement (NARI) were also calculated.

3
A p-value < 0.05 was considered statistically significant.
The average time between the onset of symptoms and CMR was 6 ± 1.7 days, with a range of 3-20 days following ER admission. 54 (54/102, 52.9%) patients presented with IL onset, 28 (28/102, 27.5%) with CM and 20 (20/102, 19.6%) with AR. Within the IL forms, three individuals presented with severe heart failure symptoms associated with remarkable T-troponin release and ST-T abnormalities, and were categorized as "fulminant myocarditis". In one of those patients, ventricular mechanical support was needed, due to the occurrence of cardiogenic shock.
For ventricular function, we found statistically significant differences between CM group compared to IL and AR as regards end diastolic volume (EDV) normalized to Body Surface Area (BSA) and EF. EDV/BSA and EF are expressed as median and confidence interval (CI  Table 1).

Comparison between oLLC vs nLLC
Inter-observer agreement was excellent for both oLLC and nLLC, resulting in a Cohen's k of 0.94 and 0.98, respectively. In detail, the agreement was substantial for EGE ratio (k = 0.66) and excellent for LGE and edema assessment with T2w STIR (k was 0.94 and 0.90 respectively). As regards relaxometric images analysis, the agreement was almost perfect either for native T1 and for T2 mapping and for ECV (k was 0.96 and 0.96 and 0.89, respectively). Doubtful cases were discussed until an agreement between the two operators was found.
Diagnostic performances of oLLC vs nLLC was also compared in a subcohort of 42 individuals with EMB.
Histological confirmation was available in 2 patients with AR onset ( Fig. 2). The NRI and NARI were 0.13 and 0.00 for IL, − 0.64 and 0.25 for CM and 0.07 and 0.19 for all clinical presentations, respectively.

Infarct-like presentations
Intra-myocardial accumulation of water, together with endothelial dysfunction, represent the hallmarks of active inflammation which are usually found in IL forms of myocarditis [25] [26]. Corresponding CMR features are characterized by a typical non-ischemic distribution of myocardial damage consistent with tissue oedema and necrosis [6,27]. This was confirmed in our patient's cohort, where the vast majority of patients had a local, typically mid-basal lateral left ventricular involvement associated with preserved or slightly reduced ejection fraction, [6] (mean EF was 56.04% [54. .87%] in IL group).
In this specific setting, use of conventional imaging (T2w-STIR, PSIR-LGE and T1-TSE) provides sufficient information to display typical inflammation features like oedema [28], necrosis/fibrosis and hyperaemia, providing a visual and a semi-quantitative assessment of myocardial injury (Fig. 3).
As a consequence, we found no added value from the use of relaxometry-based sequences which yielded to comparable results between oLLC vs nLLC, for diagnostic performances with a substantial degree of agreement [0.73 (0.543-0.912)] and a DA of 75% [47.62%-92.73%] for both Criteria using EMB as a reference standard.
We also observed three cases of diffuse myocardial damage who were clinically categorized as fulminant myocarditis (FM). FM has been described as an acute disease due to a severe inflammatory response that requires mechanical or electrical support and the eventual management of threatening arrhythmias. Although a typical pattern of CMR myocardial damage could not be identified in this subgroup, oLLC well performed in those patients, showing an equal diagnostic performance as compared to nLLC. Our results are supported by the literature [29]. In fact, myocardial oedema, evaluated by T2w STIR, is usually well represented and involves different myocardial segments, with an elevated SI as compared to not injured myocardium [29], whereas LGE is associated with a worsening of cardiac function and development of heart failure after myocarditis [30,31].

Atypical forms: cardiomyopathic and arrhythmic presentations
CMR appearance of cardiomyopathic forms reflects the different evolution of myocardial damage which has been described as the result of a prolonged subacute inflammatory response [6] and myocytes apoptosis [21].
LV dysfunction and clinical symptoms are usually preceded by viral persistence in myocardial tissue, which triggers the activation of autoimmune processes inducing subtle and diffuse myocardial damage [32]. Accordingly, there is a progressive oedema reabsorption, which corresponds to mild and diffuse signal intensity abnormalities that are poorly detectable with conventional sequences [33]. In this specific clinical setting, tissue relaxometry-based imaging allows quantification of T1 and T2 relaxations changes which yields to superior diagnostic performances as compared to conventional imaging (Fig. 4). Myocardial mapping Table 1 The table summarizes main population clinical features,  CMR functional parameters, median native T1, T2 mapping and ECV  values both for the suspected myocarditis and for the biopsy-proven  positive cases, the level of agreement between oLLC and nLLC,   endomyocardial biopsy histological diagnosis and the diagnostic  accuracy of oLLC and nLLC in relation to different clinical presentations (IL and CM) IL infarct-like, CM cardiomyopathic, AR arrhythmic, CMR cardiac magnetic resonance, EDV end diastolic volume, BSA body surface area, CI confidence interval, EF ejection fraction, LLC Lake Louise Criteria, EMB endomyocardial biopsy A similar potential mechanism has been hypothesized in arrhythmic disease onset, which pathogenesis still needs to be clarified. In some cases, arrhythmias originates from re-entrant mechanisms due to foci of fibrosis and extracellular matrix expansion that occur in the chronic phase of the infection [35]. It has also been speculated that AR pattern could be associated to Human Herpes virus 6, a neurotropic virus which directly affects conduction system cells, generating arrhythmias [6]. But ventricular arrhythmias are also attributed to ion channel dysfunction in the acute phase and to the pro-arrhythmic effect of cytokines released in the chronic phase of viral infection [36]. These micro-molecular alterations [37] in the absence of typical inflammation features are hardly detected even by mapping sequences. In addition, CMR in patients with AR onset often reveals low diagnostic quality images and motion artefacts due to arrhythmias [21]. However, nLLC

Study limitations
Main limitations of our study concerns the small sample size and, in particular, the low number of patients who underwent EMB. Moreover, biopsy was only performed in those subject whose CMR and clinical diagnosis was unclear: according to clinical guidelines EMB was used as a diagnostic tool only in patients with heart failure with or without LV dilation and/or new onset arrhythmias [16] (thus, mainly CM and AR groups) and this could have introduced a selection bias for the evaluation of DA of oLLC and nLLC, affecting mostly IL presentation group. Indeed, IL patients who underwent the invasive procedure had inadequate clinical response to therapy or persistently depressed LV systolic function and this brought to a variable delay in EMB procedure with Fig. 1 The degree of agreement between old and new LLC was good for the IL presentation (57.4% vs 66.7%, respectively) and fair for CM and AR onsets (17.9% vs 60.7% and 15% vs 35%, respectively). Cohen's k was 0.57 considering the overall clinical patterns (38.2% vs 58.8%). LLC Lake Louise Criteria, IL infarct-like, CM cardiomyopathic, AR arrhythmic

Conclusions
Our study has shown a significant added diagnostic value of nLLC in patients with CM forms of myocarditis, reflecting the superior detection of subtle/diffuse forms of myocardial involvement characterizing myocardial mapping imaging. Conventional sequences remain highly accurate in IL forms, as a consequence of the predominant active inflammatory manifestations associated with acute symptoms with troponin raise and typical ECG-changes. No significant advantages were found in the application of nLLC in patients with AR onsets. Pathological substrate triggering arrhythmias in myocarditis remains debated, but likely depends on a combination of direct viral infection and/or structural damage to the conduction system which is poorly depictable with CMR imaging. As an additional peculiarity, image quality is highly influenced by the presence of arrhythmias, further limiting accuracy of the method in this specific clinical setting.
Larger prospective studies, also investigating on other imaging parameters such as myocardial strain, intra-cavitary fluid analysis or early myocardial T1 mapping [38], are expected to further clarify the role of CMR in different forms of myocarditis with the aim of tailoring specific scanning protocols to patient's clinical presentations. Fig. 3 43-year-old male patient with acute chest pain and T-Troponine release (0.05 mcg/L). The EGE ratio (a) was 5.4 with hyperemia of the anterior and infero-septal segments on mid-ventricular plane (a, red arrows). Oedema (b) with a patchy distribution on the antero-lateral and infero-septal segments (b, white arrows), was confirmed by a T2 ratio of 2.4 on mid-ventricular plane. A non-ischemic LGE (c) was found in the inferior and infero-septal segments (c, yel-low arrows) with a subepicardial distribution. Native T1 (d, mean value of 1073 ms), ECV (e, mean value of 29%) and T2 mapping (f, mean value of 51 ms) were all increased, confirming the positivity of both old and new LLC. EGE early gadolinium enhancement, STIR short tau inversion recovery, LGE late gadolinium enhancement, ECV extracellular volume fraction, LLC Lake Louise Criteria No hyperaemia (a, EGE ratio < 4), neither oedema (b, T2 ratio < 2) nor LGE (c) was found: oLLC were negative for suspected acute myocarditis. Native T1 was severely increased (d) with a mid-ventricular mean value of 1190 ms, ECV (e) mean value was 35% and T2 mapping was slightly increased (f, mean value: 52 ms). Both T1 and T2 criterion were positive according to nLLC and the diagnosis of viral myocarditis was confirmed by EMB. EF ejection fraction, EDV end diastolic volume, BSA body surface area, EGE early gadolinium enhancement, STIR short tau inversion recovery, LGE late gadolinium enhancement, oLLC old Lake Louise Criteria, ECV extracellular volume, nLLC new Lake Louise Criteria, EMB endomyocardial biopsy