This document describes a deep convolutional neural network called DeepArrNet for automatic arrhythmia detection and classification from electrocardiogram (ECG) signals. It proposes a novel structural unit called PTP (Pointwise-Temporal-Pointwise) that uses depthwise temporal convolutions with varying kernel sizes. DeepArrNet stacks multiple PTP units together with increasing kernel sizes and residual connections between units. It introduces multiple depthwise convolutions per unit with different kernels. Experiments on two public datasets show DeepArrNet outperforms other state-of-the-art methods with outstanding performance in evaluation metrics while providing efficient computation.

DeepArrNet: An Efficient Deep CNN Architecture For Automatic Arrhythmia Detection and Classification From Denoised ECG Beats PDF

The list of books you might like