Table 3 Comparative study of CNN, transformer, and S-transform contributions to model performance (MIT-BIH and Icentia 11k).

Model	CNN		1D Transformer		2D CNN		2D Transformer		2D hybrid model
Model	MIT-BIH	Icentia 11k	MIT-BIH	Icentia 11k	MIT-BIH	Icentia 11k	MIT-BIH	Icentia 11k	MIT-BIH	Icentia 11k
Overall Acc (%)	31.73	47.33	67.26	57.23	99.41	96.19	80.81	60.81	99.58	97.80
N
F1	0.3289	0.3519	0.6696	0.5962	0.9881	0.9638	0.6782	0.5814	0.9896	0.9792
Se	0.4065	0.3417	0.6884	0.6924	0.9852	0.9749	0.6379	0.6333	0.9881	0.9913
PPV	0.2762	0.3628	0.6516	0.5235	0.9910	0.9530	0.7239	0.5374	0.9910	0.9674
S
F1	0.0220	0.5736	0.6400	0.6615	0.9956	0.9758	0.7334	0.7906	0.9970	0.9707
Se	0.0119	0.4591	0.6190	0.5932	1.0000	0.9699	0.7410	0.7926	0.9970	0.9593
PPV	0.1428	0.7703	0.6624	0.7475	0.9912	0.9817	0.7259	0.7886	0.9970	0.9823
V
F1	0.1203	0.5391	0.4250	0.6657	0.9896	0.9610	0.7001	0.6628	0.9926	0.9857
Se	0.0801	0.5936	0.3026	0.6877	0.9881	0.9610	0.7032	0.6777	0.9941	0.9880
PPV	0.2410	0.4938	0.7132	0.6451	0.9910	0.9610	0.6970	0.6485	0.9911	0.9834
Q
F1	0.4360	0.4429	0.6849	0.3473	0.9970	0.9471	0.9255	0.3710	1.0000	0.9763
Se	0.6261	0.5010	0.8189	0.3156	0.9970	0.9419	0.9584	0.3287	1.0000	0.9733
PPV	0.3343	0.3968	0.5884	0.3860	0.9970	0.9524	0.8947	0.4260	1.0000	0.9792
F
F1	0.4122	–	0.8521	–	1.0000	–	0.9912	–	1.0000	–
Se	0.4613	–	0.9345	–	1.0000	–	1.0000	–	1.0000	–
PPV	0.3725	–	0.7830	–	1.0000	–	0.9824	–	1.0000	–
Kappa	0.1465	0.2978	0.5908	0.4297	0.9926	0.9492	0.7601	0.4774	0.9948	0.9707
MCC	0.1559	0.3013	0.5988	0.4330	0.9926	0.9493	0.7605	0.4791	0.9948	0.9707

Quick links

Search