A transformer-based convolutional neural network model can accurately classify pediatric heart sounds to detect congenital heart diseases using a minimum signal duration of 5 seconds.
