Complexity of resting cortical activity predicts neurophysiological responses to theta-burst stimulation but fails to generalize: A rigorous machine-learning approach
Fig 5
Schematic Diagram of Feature Grouping and Model Generation.
In the top panel, feature groups were iteratively generated by generating all combinations of different features (S2 Table), ROIs (S3 Table and data transformations (normalization in top or distance to the median in bottom). An example of a feature group would be normalization of the alpha band powers from the central region of interested and PreTBS Features. Each table represents one feature group. In the middle panel, feature groups with labeling of modulation in corticospinal or cortical excitability are further iteratively generated by generating all combinations of feature groups and categorization methods, represented as a green column appended to the feature group table (different windows of LMFP (S1 Table) or t-tests of peak-to-peak MEP amplitudes between post and pre-TBS protocol). In the bottom panel, each feature group with labeling is trained by different classifiers, represented as gears, to form models. An example of a model would be a logistic regression with L2 regularization trained on the same example feature group above to the iTBS responses as facilitation or suppression based on the ratio of the LMFP between post- and preTBS sessions for the 55-85 ms window. See the main text for the total number of models tested in MEP and LMFP Ratios Experiments.