Table 3 Mean and standard deviations of classification accuracies for two deep-learning models with respect to the different window lengths for each CV method.

From: Risk of data leakage in estimating the diagnostic performance of a deep-learning-based computer-aided system for psychiatric disorders

  

60 s (original)

20 s

15 s

10 s

5 s

CNN-13

sCV

62.06 ± 2.86

64.77 ± 3.23

67.54 ± 3.34

67.96 ± 2.82

74.24 ± 3.94

osCV

65.36 ± 3.80

75.70 ± 4.50

77.16 ± 2.93

80.33 ± 2.66

81.54 ± 2.25

tCV

65.12 ± 3.50

71.98 ± 2.56

73.17 ± 2.59

74.19 ± 3.27

79.60 ± 1.83

otCV

61.26 ± 4.54

80.36 ± 1.11

82.28 ± 1.02

83.61 ± 1.04

85.61 ± 0.65

EEGNet

sCV

74.79 ± 3.60

79.81 ± 2.33

79.57 ± 2.25

79.85 ± 1.41

78.72 ± 2.05

osCV

75.83 ± 3.48

79.32 ± 2.89

80.13 ± 2.16

78.06 ± 3.79

79.86 ± 2.80

tCV

74.30 ± 3.27

93.53 ± 0.85

96.38 ± 0.83

97.79 ± 0.46

98.12 ± 0.42

otCV

76.41 ± 2.92

99.89 ± 0.09

99.92 ± 0.05

99.90 ± 0.03

99.77 ± 0.08