Automatic Seizure Detection Using Multi-Input Deep Feature Learning Networks for EEG Signals

Table 4

The calssification results of the proposed MDFLN based on two public datasets.

Dataset	PID	Base network				Fine-tuning network
		Accuracy	Specificity	Sensitivity	AUC	Accuracy	Specificity	Sensitivity	AUC
CHB-MIT	1	99.76	99.52	100	100	100	100	100	100
	2	99.06	98.75	99.38	98.63	100	100	100	100
	3	95.71	93.82	97.6	97.01	98.52	97.31	99.73	97.74
	4	94.34	91.72	96.97	99.13	97.11	96.14	98.08	99.11
	5	94.88	91.42	98.33	98.79	97.49	96.66	98.33	99.82
	6	91.56	98.26	84.86	94.42	97.75	97.27	98.22	99.17
	7	98.23	99.37	97.1	98.54	97.91	99.05	96.78	98.67
	8	97.78	97	98.55	99.65	97.66	97	98.33	99.61
	9	98.65	97.69	99.62	99.22	98.46	97.31	99.62	99.37
	10	99.15	99.04	99.28	99.23	96.9	96.9	96.9	99.03
	11	96.14	93.8	98.48	90.69	98.04	97.98	98.1	95.8
	12	96.6	96.48	96.72	99.49	98.53	98.98	98.07	99.75
	13	98.38	98.25	98.5	99.88	99.13	99.5	98.75	99.85
	14	96.37	97.8	94.95	97.12	99.27	98.54	100	100
	17	96.8	98.58	95.03	98.92	96.62	99.64	93.59	97.03
	18	97.79	95.91	99.66	99.76	99.15	99.66	98.64	99.79
	19	96.4	94.59	98.21	94.89	96.18	94.15	98.22	95.51
	20	98.29	97.72	98.87	99.74	96.4	93.96	98.85	99.82
	21	97	96.2	97.81	97.99	96.43	93.96	98.9	94.6
	22	96.66	94.86	98.46	100	99.74	99.49	100	100
	23	98	96.74	99.25	99.92	98.99	99.24	98.74	99.7
	24	98.22	98.44	98	99.73	97.77	98.21	97.32	99.27
	Average	97.08	96.63	97.53	98.31	98.09	97.77	98.42	98.8
Bonn	Case 1	98.32	99.03	95.36	99.14	98.4	98.71	97.02	99.07
	Case 2	92.91	94.46	89.36	—	93.12	94.84	89.68	—
	Case 3	89.76	93.6	74.4	—	91.28	94.55	78.2	—