Using CNN Saliency Maps and EEG Modulation Spectra for Improved and More Interpretable Machine Learning-Based Alzheimer’s Disease Diagnosis

Table 7

Performance comparison achieved with best threshold-cluster settings and top-24 features. Bold values indicate the highest accuracy achieved for a given classification task.

Tasks	Benchmark [29]		Proposed (exp. 1)		Proposed (exp. 2)		Proposed (exp. 3)		Proposed (exp. 4)
	Acc (%)	F1 (%)	Acc (%)	F1 (%)	Acc (%)	F1 (%)	Acc (%)	F1 (%)	Acc (%)	F1 (%)
T1	50  3	50  3	54  2	55  2	54  2	55  2	54  2	55  2	47  2	49  2
T2	70  1	65  1	60  2	49  3	60  2	49  3	71  2	61  2	71  2	61  2
T3	64  3	63  3	48  2	48  2	46  2	46  2	46  4	46  4	65  2	64  2
T4	73  2	71  2	83  2	83  2	82  0	82  0	83  1	83  1	79  1	79  1
T5	73  2	72  2	86  3	86  3	89  3	89  3	85  3	84  3	82  2	81  2

Bold values indicate the highest accuracy achieved for a given classification task.