Our study population consisted of consecutive patients with NICM enrolled prospectively for CMR studies presenting to the Canberra Heart Rhythm Centre, in the period between November 2019 and November 2021. The patients underwent invasive electrophysiological studies with voltage mapping and radiofrequency ablation when deemed necessary. The diagnosis of NICM was made with corroborative evidence of LV dysfunction (LV ejection fraction <50%), in the absence of significant coronary artery disease (>50% stenosis as assessed by coronary angiography). The following categories of NICM were excluded: congenital heart diseases; hypertrophic cardiomyopathy; arrhythmogenic right ventricular cardiomyopathy (ARVC); and LV noncompaction. In addition, pacing-dependent patients, persistent/long standing atrial fibrillation patients were excluded for the study. The study was conducted as per the ethical guidelines of the Declaration of Helsinki and was approved by the Human Research and Ethics Committee of The Canberra Hospitals (2019/ETH13256).

2.1 CMR acquisition protocol

CMR was performed on a 1.5 T scanner (Ingenia; Philips Healthcare, the Netherlands) with a cardiac phased-array receiver surface coil and electrocardiogram (ECG) gating. For the assessment of LV functions, cine imaging was performed by using a steady-state free precession (SSFP) sequence in the vertical long axis, horizontal long axis and short axis. In patients with devices, Turbo Spin echo for cine imaging was used when significant susceptibility artifact was present on SSFP imaging. Patients with significant artifacts due to devices were excluded from the study. Pacing-dependent patients were excluded from the study. In patients with cardiac resynchronization devices, biventricular pacing was on. Standard parameters were repetition time/echo time 3.6/1.8 ms; sense factor 2, flip angle — 60°; section thickness — 8 mm; field of view — 300 mm. For scar assessment, LGE images covering the entire LV were acquired approximately 15 min after an intravenous injection of 0.2 mmol/kg gadobenate dimeglumine contrast agent. The LGE-images were acquired using a magnitude inversion-recovery (IR) or phase sensitive inversion recovery (PSIR) gradient–recalled echo sequence with 8.0 mm slice thickness. A wideband LGE sequence was used to minimize artifacts from the battery pack in subjects with device implants.

2.2 Myocardial strain and LGE-scar assessment by CMR

All the CMR studies were analysed offline by using a dedicated software called the Segment Medviso version 3.3 RX [8] (www.medviso.com/segment). The base and the apex of the LV were defined from the short-axis slices. The endocardial and epicardial borders were traced manually in the short axis, apical 3-chamber long-axis, apical 2-chamber long-axis and apical 4-chamber long-axis images in end-diastole and end-systole. Adequate precaution was exercised to avoid blood pool contamination and to exclude the papillary muscles. CMR studies with poor image quality or missing slices were excluded from the analysis. Segmentation of the LV into the standard 17-segments was carried out by the software [9]. LV dimensions, volume, mass and LVEF were estimated automatically. Myocardial strain was measured throughout the cardiac cycle by myocardial Feature-Tracking (FT) [10]. The peak measurements of three strain parameters — circumferential strain, radial strain, and longitudinal strain — were considered for analysis. Global and segment-wise strain values were extracted from the software. Circumferential strain and radial strain were measured from 16 segments (excluding apical segment). Longitudinal strain was measured from all 17 segments. The strain measurements were performed in a blinded fashion by two experienced analysts.

Inter-observer reproducibility of all the three strain parameters was studied. Due to high inter-observer variability with radial strain, only circumferential strain and longitudinal strain were considered for localization of abnormalities in each segment. Peak circumferential strain and longitudinal strain values of >−17% were considered abnormal. This cut-off was chosen arbitrarily based on multiple studies reporting outcomes with the same cut-offs [7, 11-13]. However, there are no existing large-scale studies which have validated abnormal segmental strain values. The number of abnormal segments were counted for circumferential strain, longitudinal strain and composite of circumferential strain and longitudinal strain (circumferential + longitudinal strain). Percentage abnormal myocardium was derived as a proportion of abnormal segments. LGE-scar was determined in each segment by the EWA-algorithm [14]. Automatic delineation of scar borders was performed and was verified by an experienced CMR-analyst. The percentage area of LGE-scar for the total LV and in each segment was extracted from the software. Segmental circumferential strain, longitudinal strain and percentage area of LGE-scar was displayed on a 17-segment color polar plots.

2.3 Electro-anatomical mapping and catheter ablation

A systematic protocol for EAM was followed uniformly in all patients. All antiarrhythmic drugs were discontinued routinely at least 5 half-lives before the procedure. Three-dimensional (3D) left ventricular geometry was reconstructed by intracardiac echocardiography (ICE; 64-element, 5.5 to 10 Hz; SOUNDSTARTM, CARTOSOUNDTM module Biosense Webster, La Jolla, CA, USA). EAM of the endocardial LV was performed using the CARTO 3 Version 7 mapping system (Biosense Webster) using a multi-electrode mapping catheter (PENATRAYTM, Biosense Webster). The geometry created using ICE was registered to an endocardial 3D shell of LV acquired by the mapping catheter. High density mapping of the LV was performed at all the segments of the LV. The low voltage zones were addressed further by point-by-point mapping using a deflectable 3.5-mm irrigated-tip mapping catheter with contact force (THERMOCOOL SMARTTOUCH-SFTM, Biosense Webster) during sinus rhythm. In patients with cardiac resynchronization devices, biventricular pacing was on. Geometry, bipolar and unipolar electrograms (EGMs) were simultaneously recorded and all segments of the ventricle were sampled. The mitral and aortic annuli were defined by ICE. In addition, the mitral annulus was verified as that with a 1:1 ratio between atrial and ventricular electrograms. Low voltage points acquired with <3 g contact force, <10 mm from the endocardial shell, points with unstable cycle length, points within 1 cm of the aortic and mitral valve annulus were all excluded from analysis. Bipolar signals were filtered at 30 to 400 Hz. Unipolar signals were measured between the tip electrode and Wilson-central terminal and were filtered at 1 to 240 Hz. The fill threshold was set to 10 mm.

Ventricular tachycardia (VT) induction was attempted in all patients with programmed ventricular stimulation with triple extrastimuli from at least two right ventricular or LV sites with at least two drive cycle lengths. Induced VTs were identified as clinical if they matched the cycle length and morphology of the stored electrograms from the ICD or the 12-lead ECG when available. VT entrainment was performed if the VTs were hemodynamically stable. Pace mapping at threshold was performed to match the inducible VT, in case the VT was hemodynamically unstable or non-sustained and repetitive. Substrate modification was performed at the regions of good pace map, aiming at elimination of local abnormal ventricular activity potentials, late potentials, and low amplitude fractionated electrograms. The contact force catheter was also used for ablation. The primary endpoint for ablation was elimination of the clinical VT and monomorphic nonclinical VT.

2.4 Segmental analysis of EAM data

The raw EAM datasets were exported from the CARTO system and imported into an EP Lab Research Works application (www.eplabworks.com). Automatic annotation of all EGMs and automatic segmentation of the LV into 17 segments were performed. The landmarks for LV segmentation were set in cooperation with 2 experienced electrophysiologists. The EGM analysis was performed in each segment. Annotation of all electrograms were individually reviewed. Low voltage and scar regions were defined based on standard abnormal values for bipolar low voltage zones (Bi-LVZ; <1.5 mV), bipolar scar (Bi-Scar; <0.5 mV) and unipolar low voltage zone (Uni-LVZ; <8.3 mV) [15, 16]. The bipolar and unipolar low voltage maps were displayed on 17-segment color polar plots. The extent of the low voltage zones in the entire endocardial surface as well as in each segment was quantified as endocardial surface area (cm2) and proportion (%) of abnormal to the total LV endocardial surface area and proportion (%) of the abnormal area within each segment.

2.5 Clinical follow-up

The duration of follow-up was calculated from the time of the EAM study. The patients were followed up every 6 months at the heart failure out-patient clinic, or earlier if symptomatic. All patients with device implants were followed up with remote monitoring-based device interrogations every month. Device therapies were reviewed for appropriateness by experienced cardiac electrophysiologists.

2.6 Statistical analysis

Continuous variables which were normally distributed were expressed as mean±SD. Categorical variables were presented as proportions in percentages. The inter-observer variability was assessed with Bland-Altman analysis, coefficient of variation (CV) and the 95% limits of agreement (LOA) were studied for the dispersion around the mean. The correlation was tested by Spearman’s rank coefficients for parameters detected by CMR and % bipolar LVZs and by Pearson’s correlation for parameters detected by CMR and % unipolar LVZs. Correlation values were ranked as mild (0–0.3), moderate (0.4–0.6) and high (0.7–1.0). Linear regression models were studied for each variable, to determine the increase of LVZ per-unit increase in the independent variable. Student-t test was carried out to identify the differences in the mean LVZ in the patients with and without VT. The 17-segment bull’s eye maps of the CMR circumferential strain, CMR longitudinal strain, CMR LGE, bipolar voltage EAM and unipolar voltage EAM were used for side-by-side comparisons. The segmental abnormalities detected by LGE, abnormal circumferential strain and abnormal longitudinal strain in each patient were classified as concordant with EAM, if the segments also had bipolar/unipolar voltage abnormalities. The concordance rates were thus presented as proportion of correct classifications. Differences, correlation-coefficients, and the odds ratio were considered statistically significant at the two-sided p < 0.05 level. All the analyses were performed using STATA 17.0 (STATA Corporation, Texas, USA).