Luan G. Cotini, Renato Candido, Magno T. M. Silva Detecção automática de arritmias cardíacas utilizando redes neurais convolucionais. Simpósio Brasileiro de Telecomunicações – SBrT’2026, 2026, Salvador, Anais do SBrT 2026, 2026, pp. 1–3. (in portuguese)

Abstract

This work compares two convolutional neural network (CNN) architectures for the automatic detection of cardiac arrhythmias. Based on the electrocardiogram (ECG) signal, two approaches were considered: a one-dimensional model (1D-CNN) operating on the raw signal and a two-dimensional model (2D-CNN) fed with spectrograms derived from the ECG signal. To mitigate class imbalance, class weights in the cost function and random oversampling were employed. The results indicate that the 2D-CNN outperforms the 1D-CNN, reducing false alarms and improving the F1-score.

Keywords

Machine learning, convolutional neural networks, electrocardiogram, spectrogram.

Downloads

• PDF