Skip to main content

Table 1 Optimal classification results and performance comparisons in the three MI paradigms

From: A novel channel selection method for optimal classification in different motor imagery BCI paradigms

Subjects Dataset 1 Dataset 2 Dataset 3
MI tasks Parad Two-class control Parad Four-class control Parad
All Ch C3, C4, Cz Opt Ch All Ch C3, C4, Cz Opt Ch All Ch C3, C4, Cz Opt Ch
Acc (%) Acc (%) Acc(%) Acc (%) Acc (%) Acc (%) Acc (%) Acc (%) Acc (%)
Sub1 67.8 88.9 92.8 95.6 92.7 99.3 75.5 69.4 87.1
Sub2 62.2 65.6 86.7 82.7 65.1 91.2 74.1 76.4 90.6
Sub3 62.2 61.7 82.8 85.9 51.8 94.1 76.3 72.7 83.3
Sub4 70.0 74.4 90.0 67.9 52.2 85.3 58.2 67.0 80.2
Sub5 71.1 79.4 90.0 93.5 76.2 96.1 63.4 63.4 77.6
Sub6 62.8 54.4 80.0 96.2 95.2 98.6 55.8 73.4 80.3
Sub7 64.4 69.4 80.0 92.2 93.8 98.1 77.7 71.2 81.8
Sub8 57.2 60.6 78.9 82.4 82.3 90.4 74.8 66.9 84.9
Mean 64.7 69.3 85.2 87.1 76.2 94.1 69.5 70.1 83.2
Std 4.6 11.2 5.4 9.5 18.0 4.9 8.8 4.2 4.2
p value p = 7.0e−5 p = 0.024 p = 2.4e−4
  1. In each dataset, the p-value is computed from one-way ANOVA among Opt Ch Acc, All Ch Acc and C3, C4, Cz Acc. It has to been noted that the subjects 1–8 in all the three datasets are not the same group of subjects
  2. Parad paradigm, Ch channel, Acc accuracy, Opt optimal