Table 2 Single trial cross-validated classification accuracies, expressed as percentages, for each movement condition and EEG channel set.

From: Detection of attempted movement from the EEG during neuromuscular block: proof of principle study in awake volunteers

 

30 EEG channels

9 EEG channels

Training & test conditions

Training & test conditions

Actual movement

Isometric movement

Imagined movement

Attempted movement

Actual movement

Isometric movement

Imagined movement

Attempted movement

Subject #

1

81 (73–89)

83 (75–91)

73 (64–82)

78 (68–88)

78 (70–86)

80 (72–88)

69 (59–79)

81 (72–90)

2

83 (75–91)

68 (58–78)

55 (44–66)

73 (62–84)

81 (73–89)

68 (58–78)

56 (46–66)

82 (73–91)

3

93 (89–97)

89 (83–95)

73 (64–82)

92 (87–97)

96 (93–99)

88 (82–94)

78 (69–87)

92 (87–97)

4

79 (71–87)

81 (73–89)

75 (66–84)

81 (72–90)

92 (87–97)

77 (68–86)

80 (72–88)

79 (69–89)

Mean

84 (74–94)

80 (66–94)

69 (54–84)

81 (68–94)

87 (73–100)

78 (65–91)

71 (53–88)

84 (74–93)

  1. First, separate classifiers were trained on each of the named movement conditions, using a subset of the recorded trials. Then the classification performances were estimated for each classifier on another set of trials belonging to that condition. For this procedure of performance estimation ten-fold cross-validation was used. Results are given as classification accuracy (95% CI): for each individual classification accuracy the binomial 95% CI is given, for the mean classification accuracy the standard 95% CI is given.
Back to article page