Fig. 9
From: Assessing brain-muscle networks during motor imagery to detect covert command-following
Subject-level classifiers based on brain-muscle networks, EEG power, and heart activity features. Left: average AUC across folds with 95% bootstrap confidence interval for each subject and classifier. Subjects are sorted considering the AUC for the full classifier. Right: ratio between the mean AUC for the classifiers lacking groups of features (or the LDA-CSP) and the mean AUC for the full classifier. ~ BMN: without the brain-muscle networks features, ~ HR: without the heart rate, ~ HRV: without heart variability, ~ mu-alpha: without the mu-alpha power for the 64 channels, ~ mu-beta: without the mu-beta power for the 64 channels, only BMN: classifier fed with only the brain-muscle networks features, and LDA-CSP: LDA classifier based on common spatial patterns