Neural network-derived electrocardiographic features have prognostic significance and important phenotypic and genotypic associations

Background Subtle prognostically-important ECG features may not be apparent to physicians. In the course of supervised machine learning (ML), many thousands of ECG features are identified. These are not limited to conventional ECG parameters and morphology. Hypothesis Novel neural network (NN)-derived ECG features can predict future cardiovascular disease and mortality Methods and Results We extracted 5120 NN-derived ECG features from an AI-ECG model trained for six simple diagnoses and applied unsupervised machine learning to identify three phenogroups. In the derivation cohort (CODE, 1,558,421 subjects), the three phenogroups had significantly different mortality profiles. After adjusting for known covariates, phenogroup B had a 20% increase in long-term mortality compared to phenogroup A (HR 1.20, 95% CI 1.17-1.23, p < 0.0001). The predictive ability of the phenogroups was retained in a group with physician confirmed normal ECGs. We externally validated our findings in five diverse cohorts (Figure) and found phenogroup B had a significantly greater risk of mortality in all cohorts. Phenome-wide association study (PheWAS) showed phenogroup B had a higher rate of future AF, ischaemic heart disease, AV block, heart failure, VT, and cardiac arrest. Phenogroup B had increased cardiac chamber volumes and decreased cardiac output. A single-trait GWAS yielded four loci. SCN10A, SCN5A and CAV1 have roles in cardiac conduction and arrhythmia. ARHGAP24 does not have a clear cardiac role and may be a novel target. Gradient-weighted Class Activation Mapping (Grad-CAM) identified the terminal QRS and terminal T wave as important regions of the ECG for identification of phenogroup B. Conclusion NN-derived ECG features can be used to predict all-cause mortality and future cardiovascular diseases. We have identified biologically plausible and novel phenotypic and genotypic associations that describe mechanisms for the increased risk identified. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement AS is funded by a British Heart Foundation (BHF) clinical research training fellowship (FS/CRTF/21/24183). FSN and NSP are supported by the BHF (RG/F/22/110078). FSN is also supported by the National Institute for Health Research Imperial Biomedical Research Centre. DO'R is supported by the Medical Research Council (MC\_UP\_1605/13); National Institute for Health Research (NIHR) Imperial College Biomedical Research Centre; and the British Heart Foundation (RG/19/6/34387, RE/18/4/34215). IA, TN and partially MM have been supported by MH CZ - DRO (FNBr, 65269705). ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: This study complies with all relevant ethical regulations. The Clinical Outcomes in Digital Electrocardiography (CODE) study was approved by the Research Ethics Committee of the Universidade Federal de Minas Gerais, protocol 49368496317.7.0000.5149. The Whitehall II study was approved by the Joint University College London/University College London Hospitals Committees on the Ethics of Human Research. The UK Biobank has approval from the North West Multi-Centre Research Ethics Committee as a Research Tissue Bank (application ID 48666). The Longitudinal Study of Adult Health (ELSA-Brasil) was approved by the Research Ethics Committees of the participating institutions and by the National Committee for Research Ethics (CONEP 976/2006) of the Ministry of Health. The Sao Paulo-Minas Gerais Tropical Medicine Research Center (SaMi-Trop) study was approved by the Brazilian National Institutional Review Board (CONEP), No. 179.685/2012. For the Beth Israel Deaconess Medical Center (BIDMC) cohort ethics review and approval was provided by the Beth Israel Deaconess Medical Center Committee on Clinical Investigations, IRB protocol # 2023P000042. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes SaMi-Trop cohort was made openly available (https://doi.org/10.5281/zenodo.4905618). The CODE-15% cohort was also made openly available (https://doi.org/10.5281/ zenodo.4916206). Restrictions apply to additional clinical information on the CODE-15% and SaMi-Trop cohorts; to the full CODE cohort, the ELSA-Brasil cohort, and the Whitehall II cohort. UK Biobank data are available upon application (http://www.ukbiobank.ac.uk/). The BIDMC dataset is restricted due to ethical limitations. Researchers affiliated to educational, or research institutions may make requests to access the datasets. Requests should be made to the corresponding author of this paper. They will be forwarded to the relevant steering committee.