Table 2 Averaged single-trial classification accuracy (%) in physiological, pharmacological, and pathological conditions for TMS–EEG: this represents the accuracy ± standard deviation.

From: Quantifying arousal and awareness in altered states of consciousness using interpretable deep learning

Target domain

Source domain

Arousal

Awareness

Sleep

Sleep

87.79 ± 2.50

91.95 ± 4.74

Sleep + Ane

87.23 ± 2.99

89.96 ± 5.48

Sleep + DoC

80.73 ± 5.05

89.60 ± 4.26

Sleep + Ane + DoC

84.01 ± 3.46

91.14 ± 4.29

Ane

Ane

79.01 ± 10.61

80.20 ± 10.06

Ane + Sleep

82.58 ± 6.92

87.78 ± 6.46

Ane + DoC

69.99 ± 11.89

82.22 ± 10.66

Ane + Sleep + DoC

72.68 ± 17.22

85.61 ± 9.09

DoC

DoC

–

75.84 ± 14.71

DoC + Sleep

75.94 ± 18.14

79.44 ± 15.51

DoC + Ane

83.12 ± 12.79

75.30 ± 11.99

DoC + Sleep + Ane

66.29 ± 19.02

78.78 ± 12.98

  1. Ane anesthesia domain, DoC patients with disorders of consciousness domain.
  2. The target domain implies the condition with the target participant to be tested for calculating explainable consciousness indicator (ECI) using convolutional neural network (CNN) with spatiotemporal information, and the source domain implies the conditions included in training for learning classifiers.
Back to article page