Figure 5.

Discriminate maps generated from the four different feature selection algorithms employed. Linear discriminant weights were averaged across the 10 cross-validation iterations to produce the discriminant maps for t-test filter (a) and reliability filter (b) methods. Discriminate maps for recursive feature elimination and reliability recursive feature elimination were calculated by quantizing linear discriminant maps extracted each iteration of cross-validation to −1 if the feature had a negative weight and +1 if its weight was positive. Summary maps were generated by averaging the quantized maps across the iterations of cross validation. dmThalamus: dorsomedial thalamus, vMF10: ventral medial prefrontal cortex (BA10), sgACC24∣25: subgenual cingulate cortex (BA 24∣25), aINS: anterior insula, rACC24: rostral anterior cingulate cortex (BA 24), dlPFC9: dorsolateral prefrontal cortex (BA 9), OFC11: orbitofrontal cortex (BA 11), dMF10: dorsomedial prefrontal cortex (BA10), MCC24: midcingulate cortex (BA24), vPCC: ventral posterior cingulate cortex, scACC25: subcallosal cingulate cortex (BA 25), BA: Brodmann area