A Multi-layer CNN Using the ECG, Age and Sex Predicts Ventricular Arrhythmias in the General Population

Ramírez, Julia (Universidad de Zaragoza) ; Miguel, Antonio (Universidad de Zaragoza) ; van Duijvenboden, Stefan ; Orini, Michele ; Young, William J. ; Tinker, Andrew ; Lambiase, Pier D. ; Munroe, Patricia B. ; Martínez, Juan Pablo (Universidad de Zaragoza)

Resumen: Life-threatening ventricular arrhythmias (LTVA) prediction in individuals without cardiovascular disease remains a major challenge. We tested the performance of a multilayer convolutional neural network (CNN) using ECG signals, age and sex. We split 86,603 individuals from the UK Biobank study into a training (90%) and a test (10%) set. In the training set, we trained a multilayer CNN using 15-second ECGs at rest from lead I, age and sex as inputs. The output was the probability of LTVA within a 12-year follow-up. The CNN model consisted of a four-layer CNN (128, 128, 256 and 256 channels, kernel sizes of 3) and a single attention layer. Age and sex were included as external inputs to the final layer. In the test set (0.9% LTVA events), the CNN's prediction led to a median AUC of 0.601, and a specificity of 0.287 for a sensitivity of 0.750. We set a threshold at the CNN's prediction value maximising the sum of specificity and sensitivity in the training set. Survival analyses showed a hazard ratio (HR) of 1.396(P=0.021) for individuals with a CNN's prediction value > threshold, versus those with a CNN's prediction value < threshold. A multilayer CNN model using 10-second ECG data from lead I, together with information on age and sex, can stratify individuals at risk of LTVA. Our findings support the potential utility of wearables for accessible screening in the general population.

Idioma: Inglés

DOI: 10.22489/CinC.2023.342

Año: 2023

Publicado en: Computing in Cardiology 50 (2023), [4 pp.]
ISSN: 2325-8861

Factor impacto CITESCORE: 1.1 - Computer Science (all) (Q4) - Cardiology and Cardiovascular Medicine (Q4)

Factor impacto SCIMAGO: 0.227 - Computer Science (miscellaneous) - Cardiology and Cardiovascular Medicine

Financiación: info:eu-repo/grantAgreement/ES/AEI/PID2021-128972OA-I00
Financiación: info:eu-repo/grantAgreement/ES/MICINN/PID2022-140556OB-I00
Financiación: info:eu-repo/grantAgreement/ES/MICINN/RYC2021-031413-I
Financiación: info:eu-repo/grantAgreement/EUR/MICINN/TED2021-130459B-I00

Tipo y forma: Artículo (Versión definitiva)

Área (Departamento): Área Teoría Señal y Comunicac. (Dpto. Ingeniería Electrón.Com.)