Skip to main content

Table 3 Performance of transformer model for EEG signals with eyes close on different methods for five class (ADHD, MDD, OCD, SMC, and Healthy) classification

From: Transformers for autonomous recognition of psychiatric dysfunction via raw and imbalanced EEG signals

Decision

Method

Accuracy

F1-score

Precision

Recall

Window-level

Trans + CCE

61.74

46.57

51.24

47.83

Trans + Focal Loss (\(\gamma\) = 2)

55.28

31.34

35.77

32.70

Trans + Focal Loss (\(\gamma\) = 0.5)

65.85

45.60

51.99

45.03

Trans + WCCE

49.61

36.94

37.06

37.72

Patient-level

Trans + CCE

67.12

49.84

63.55

49.54

Trans + Focal Loss (\(\gamma\) = 2)

57.53

32.00

40.79

32.84

Trans + Focal Loss (\(\gamma\) = 0.5)

68.49

44.59

49.51

45.11

Trans + WCCE

52.05

36.89

35.79

38.25

  1. The best performance is in bold