Arrhythmia Classification Algorithm Based on a Two-Dimensional Image and Modified EfficientNet

Table 3

Comparison of sensitivity, specificity, and precision at L = 180.

Evaluation indicator	Method	Type
		APC	LBBB	NOR	PAB	PVC	RBBB	VEB	VFW
Sensitivity	Method 1	0.89	0.995	0.998	1	0.989	0.997	1	0.979
	Method 2	0.906	0.996	0.998	1	0.99	0.999	1	0.987
	Method 3	0.909	0.998	0.998	1	0.989	0.997	1	0.979
	Method 4	0.902	0.998	0.999	1	0.987	0.997	1	0.991
Specificity	Method 1	0.89	0.995	0.998	1	0.989	0.997	1	0.979
	Method 2	0.906	0.996	0.998	1	0.99	0.999	1	0.987
	Method 3	0.909	0.998	0.998	1	0.989	0.997	1	0.979
	Method 4	0.902	0.998	0.999	1	0.987	0.997	1	0.991
Precision	Method 1	0.954	1	0.996	1	0.99	0.994	1	0.958
	Method 2	0.962	1	0.997	1	0.986	0.993	1	0.979
	Method 3	0.967	1	0.996	1	0.989	0.994	1	0.939
	Method 4	0.974	1	0.996	1	0.989	0.993	1	0.987

Notes: Method 1, Method 2, Method 3, and Method 4 represent EfficientNet-B0, EfficientNet-B0 + data augmentation, AFF-EfficientNet-B0, and AFF-EfficientNet-B0 + data augmentation, respectively.