Figure 6

Pairwise classification accuracy of SVM using nodes (a) and boundaries (b) across all threshold levels. The highest accuracy for nodes was achieved by R1 versus R2 comparison for the threshold value of 99.7%, corresponding to 764 voxels. In case of boundaries, R1 versus R3 yielded the highest accuracy for the threshold value of 63% (174 voxels). The second highest accuracy was obtained by R1 versus R2 comparison for the threshold value of 67% resulted in 346 voxels being highly modulated by 10 Hz rTMS. Dashed line represents random-choice accuracy, one of two R categories.