12-lead ECG synthesis from reduced lead set using multi-lead convolutional autoencoder

Abstract

Abstract Accurate reconstruction of the standard 12-lead electrocardiogram (ECG) from a reduced number of leads has significant potential for simplifying cardiac monitoring systems and enabling wearable healthcare solutions. This study proposes a novel multi-lead convolutional autoencoder (MLCAE) for reconstructing standard 12-lead ECG signals using only three input leads (namely, I, II, and V2) without feature extraction. The model leverages inter-lead correlations through a deep convolutional architecture to learn compact latent representations and effectively map partial observations to complete cardiac signals. Experimental results demonstrate that the proposed ML-CAE achieves superior performance in reconstructing 12-lead ECG signals, with an average correlation of 0.993, outperforming CAE (0.972) and U-Net (0.983), while maintaining low reconstruction error. These findings highlight the effectiveness of the proposed approach in preserving ECG morphology, making it suitable for wearable cardiac monitoring applications.