Table 6 Pairwise classification by PPA subtype with TRF beta weights as input.

From: Application of machine learning and temporal response function modeling of EEG data for differential diagnosis in primary progressive aphasia

Task

Feature

Frequency band

Algorithm

Participant group

Precision

Recall

F1

Accuracy

nfvPPA vs. lvPPA

Semantic dissimilarity/Word frequency (tied)

Delta

Decision tree

nfvPPA

0.67/0.79

0.72/0.56

0.73

0.73/0.75

lvPPA

0.78/0.73

0.76/0.89

svPPA vs. nfvPPA

Multiband envelope

Theta

Decision tree

svPPA

0.75

0.66

0.74

0.75

nfvPPA

0.74

0.81

svPPA vs. lvPPA

Semantic dissimilarity

Delta

Naive Bayes

svPPA

0.69

0.83

0.74

0.74

lvPPA

0.81

0.66

  1. F1 refers to the macro average of both class’ F1 scores.