Mortality and SCD risk increase when LVEF decreases below 50% [22, 23], and particularly when it becomes 40% [24]. On the other hand, due to the improvement in reperfusion therapies, there are relatively few patients with very low LVEF after an acute coronary event [25, 26, 27]. However, there is no evidence of a cause-effect relationship between an abnormal LVEF and the occurrence of SCD. In fact, although the absolute risk of SCD is higher in patients with lower LVEF, the total number of SCDs is higher in patients with higher LVEF [28]. In addition, current LVEF values used to define high-risk populations (typically 35%) have poor sensitivity, failing to identify up to 50% of patients at risk for SCD [29].

Current evidence for the choice of LVEF as the main criterion for ICD implantation in primary prevention in patients with ICM is based on the classical trials MADIT-I and II [11, 30], CABG-PATCH [31], MUSTT [32], DINAMIT [33], and SCD-HeFT [12]. Although in the latest clinical practice guidelines the presence of LVEF 35% remains the criterion that currently determines implantation [11, 12], it has not been shown to be a parameter either sensitive or specific for the prediction of SCD in ICM [34, 35, 36, 37, 38, 39]. In the recent PRESERVE EF study [40] it was found that 9 of 575 patients (1.6%) with ICM and LVEF $>$40% (mean 50.8%) had appropriate ICD therapies due to sustained VA after a mean clinical follow-up of 32 months. These patients, who did not have a standard indication for ICD implantation in primary prevention, had been indicated for ICD because they had some non-invasive additional risk factor (see below) and were inducible during EPS.

Traditionally, other clinical variables apart from LVEF have been tested in order to improve the risk stratification of VA/SCD in patients with ICM. Some of these variables have been early detection of potential triggers (premature ventricular complexes and non-sustained VTs), presence of late potentials on the signal-averaged ECG, T wave alternans, inducibility during EPS, analysis of autonomic functional markers, or the use of multivariable clinical models. The available evidence has confirmed that none of them was sufficiently reliable for arrhythmic risk stratification [19, 41, 42, 43, 44, 45, 46, 47, 48, 49]. Therefore, their use cannot be recommended for making relevant clinical decisions.

As previously mentioned, LVEF is a suboptimal risk marker for VA and SCD, and yet it constitutes the main criterion on which current clinical practice guidelines base the indication for ICD implantation in primary prevention of SCD [11, 12]. LVEF has been classically obtained using echocardiography, which has shown to generally overestimate LVEF by 3–7% compared to cardiac magnetic resonance (CMR), a more reproducible technique [50]. These differences might justify a better risk stratification using CMR [51, 52], despite no published ICD trial has included the use of CMR-derived LVEF. On the other hand, there is evidence [53] suggesting that the use of a weighted CMR score after a ST-segment–elevation myocardial infarction, including CMR-LVEF and other variables (myocardial salvage index, microvascular obstruction, and myocardial hemorrhage), was independently associated with major adverse cardiovascular events, and may provide incremental prognostic stratification as compared with GRACE score and echo-LVEF [53].

Far beyond LVEF, other clinical predictors for arrhythmia events may be found; for example, in a recent meta-analysis including 17 studies with LVEF values ranging from 25 to 59% [54], the presence of a coronary chronic total occlusion (CTO), particularly when affecting the infarct-related artery, was associated with a 1.68-fold increase in the occurrence of VT/VF or appropriate ICD therapy. Aiming to non-invasively characterize the post-infarction myocardial injury, cardiac magnetic resonance with late gadolinium enhancement (CMR-LGE) has made it possible to answer clinically relevant questions that are currently not assessable with LVEF alone or with the use of other non-invasive imaging modalities. In fact, the diagnostic and prognostic role of CMR in ICM has already been broadly recognized in more recent articles, guidelines and consensus documents [55, 56, 57, 58]. Numerous studies have shown a link between the presence of macroscopic fibrosis evaluated with CMR and the appearance of VA in patients with ICM. In a recent meta-analysis by Disertori et al. [59] including 2850 patients from 19 different studies, patients without LGE had an annual arrhythmia/SCD event rate of 1.7%, compared with an event rate of 8.6% per year in patients with LGE. Moreover, 100% of those with events had macroscopic fibrosis on CMR-LGE [59, 60]. This has been supported in modern, large prospective series of patients with ICM and non-ischemic etiologies [61].

The ability of CMR-LGE to detect myocardial fibrosis has been supported by histological correlation studies [62, 63]. The presence of slow and heterogeneous electrical conduction associated with fibrosis may favor reentry, increasing vulnerability to VT/VF onset [6, 64, 65, 66]. Areas with different levels of fibrosis coexist in the gray zone, heterogeneous tissue, or border zone (BZ), resulting in simultaneous regions of viable and non-viable myocardium, which can be identified with CMR-LGE [64]. In addition, heterogeneous tissue channels (HTC), or border zone channels (BZC), have been correlated with functional, slow conduction channels identifiable by endocardial voltage maps during electroanatomical mapping (EAM) [67, 68, 69, 70], as well as with the critical isthmuses of VT reentry circuits [71].

Despite the evidence that size and heterogeneity of the myocardial scar could be predictors of VA and mortality, it remains unknown whether CMR, beyond the use of LVEF, could facilitate the clinical decision-making in relation to arrhythmia risk stratification. In this sense, a previous prospective study by Klem et al. [72], based on a cohort of ischemic patients with a wide range of LVEF values, analyzed the risk of arrhythmic events: Patients with LVEF $>$30% and fibrosis $>$5% of total myocardial mass had a higher VA risk than those patients with LVEF $>$30% and small scars (5%), but similar to that of patients with LVEF 30%. Similarly, those patients with LVEF 30% and fibrosis 5% had a risk of arrhythmic events similar to that of patients with LVEF $>$30%. More recently [73], the presence of myocardial fibrosis and the mass of heterogeneous tissue (a.k.a. BZ), both analyzed with CMR-LGE, were evaluated in relation to the occurrence of SCD and the composite event of SCD or VA. Of 947 patients analyzed retrospectively [73], there were 29 cases of SCD (2.96%) and 80 cases of SCD/VA (8.17%) after a median follow-up of 5.82 years. The visual presence of myocardial fibrosis or heterogeneous tissue on CMR was strongly associated with the appearance of SCD and the composite event. In contrast, the associations between LVEF 35% and the development of SCD, or the composite SCD/VA, were much weaker in a competing risk analysis [73]. In addition, all patients having events (SCD/VA) showed evident fibrosis on CMR. Furthermore, the cut-off point of LVEF 35% was very weakly associated with both the occurrence of SCD or the composite SCD/VA [73]. Additionally, from all the scar variables that can be assessed with CMR, the amount (mass) of BZ and, particularly, the amount of BZ mass being part of tissue corridors (BZC) appear to be the most powerful variables associated with the development of reentrant VA [74, 75].

Including CMR-derived variables, the currently ongoing European PROFID project will aim to develop a clinical prediction model for SCD in ICM in order to improve the selection of ICD candidates. Two randomized trials will validate the usefulness of this prediction model according to the LVEF (PROFID-Preserved, NCT04540289; and PROFID-Reduced, NCT04540354). The preliminary results presented in EHRA Congress 2022 indicate a potentially relevant role of CMR-LGE; particularly regarding the extent of scar and BZ. Therefore, un updated model including these variables is to be expected.

Apart from evaluating macroscopic fibrosis using LGE, CMR is also capable of assessing diffuse interstitial myocardial fibrosis, which is related to adverse remodeling in patients with ICM [76, 77]. This is usually performed using T1-mapping techniques. Diffuse fibrosis has been shown to be an independent predictor of VA [78]. However, it should be remarked that T1-mapping is limited to just a single slice of myocardium, assuming that the diffuse fibrosis affects uniformly the remote myocardium. Therefore, more studies would be required to explore the possibilities of this method.

In addition to the aforementioned, it has recently been shown that the QRS morphology of potential inducible VTs and the location of the responsible circuit in each case can be accurately predicted using computer simulation models trained with CMR-LGE images from infarcted pigs [79]. This may suggest that our current post-processing methods for BZ analysis could eventually be complemented with even more sophisticated tools (computer models, machine learning, etc.) to identify reentrant circuits through CMR findings. Other studies have strongly supported the usefulness of CMR in identifying and characterizing the arrhythmogenic substrate in secondary prevention. Integrated 3D reconstructions of the scar obtained from CMR images are capable of accurately depicting not only the location of the arrhythmogenic substrate, but also its structure and the presence of BZC, which constitute the therapeutic target during substrate ablation procedures [80, 81, 82]. These studies have demonstrated improved clinical outcomes and better procedural efficiency when taking advantage of the CMR-derived data on the arrhythmogenic substrate.