Hypertrophic cardiomyopathy (HCM) is the most common inherited cardiac disorder, characterized by an unexplained ventricular hypertrophy with heterogeneous clinical expression, including angina, heart failure, syncope and arrhythmias [1,2]. Even though the majority of patients have a normal or near normal life expectancy, a relevant percentage of them develops ventricular arrhythmias and may even suffer sudden cardiac death (SCD). For this reason, the estimation of the SCD risk to recommend an implantable cardiac defibrillator (ICD) in primary prevention is essential to reduce mortality in this population [1]. Along with the variables included in the SCD risk calculator proposed by the European Society of Cardiology (ESC) [2] (including age, maximum left ventricle [LV] wall thickness, left atrium size, LV outflow tract obstruction, family history of SCD, non-sustained ventricular tachycardia and unexplained syncope), some additional risk factors and risk modifiers are lately acquiring great importance in this context [3]. Among them the presence of myocardial fibrosis, assessed by late gadolinium enhancement (LGE) on cardiac magnetic resonance (CMR), is of special interest, since it is a well-known marker of arrhythmic events in patients with HCM, and different meta-analysis have shown its relevance as a predictor of SCD [4,5]. Nevertheless, despite the fact of being recommended in all patients with HCM, CMR is not readily accessible in real world clinical practice. The electrocardiogram (ECG) is, on the other hand, one of the most worldwide available complementary tests in patients with HCM. For this reason, there is growing interest in studying the correlation between ECG parameters and CMR features, in particular, the presence of fibrosis, despite the limited evidence available in this regard [[6], [7], [8], [9], [10]]. Moreover, in the last years, machine learning techniques have been implemented providing a new promising approach in the analysis of hundreds of computed ECG variables, potentially outperforming the conventional visual analysis, and incrementing the diagnostic and prognostic performance of the ECG [11,12, [13]].

With these premises, the main aim of our study was to apply a machine learning approach to computer assisted ECG analysis, to identify ECG predictors of fibrosis in patients with HCM. This could help to prioritize access to CMR for patients at higher risk, in which the detection of extensive areas of fibrosis can orient the clinician in the decision-making process involved in the recommendation of ICD.