fig1

Figure 1. A Review of the Digital Network Architecture of ECG. (A) Overall encoder-decoder architecture and the downstream steps from waveform segmentation to digital-signal reconstruction; (B) Schematic illustration of Encoder5, Encoder6 and Decoder5 in CU²-NeXt; (C) Schematic of the low-level encoder and decoder components in CU²-NeXt, which employ ConvNeXt blocks for feature extraction; (D) Structure of the convolutional module, which utilizes depthwise convolution for feature extraction while capturing spatial and channel information from the feature maps; (E) Detailed architecture of the CBAM attention module; (F) Schematic of the channel attention structure; (G) Schematic of the spatial attention mechanism. ECG: Electrocardiogram; CBAM: convolutional block attention module; AMP: adaptive max pooling; AAP: adaptive average pooling; GELU: Gaussian error linear unit; BN: batch normalization.