Table 4.

Tabulated summary-signal conversion, algorithm design, and modification.

Ref.	Model	Modification
[59]	KecNet-CNN based model	Domain knowledge-optimized 1st CNN layer merged with parallel ECG quantitative features based on clinical knowledge
[64]	Tenary neural network (TNN)	ECG sequences converted to binary images
[6]	2D-CNN	ECG signals converted to scalograms
[51]	2D-CNN	ECG sequences converted to binary images
[63]	Fussing transformer	Decoder eliminated-modified input embedding (CNN architecture) and replaced self-attention with a depth-wise convolution
[79]	CNN	Partitioned layers-First 2 layers in the wearable device-Last 3 layers in cloud