(A) Features for MF vs. MN classification were selected based on an F-test of the contrast MF>MN (p<0.05, k=30: 6,248 total features), respectively. (B) Classifications were performed similar to main analyses in the text but over the range of 1 to 6000 ranked features. Note that feature selection was biased by using a single mask based on the F-test conducted over the complete data set. In spite of this, MF vs. MN classification only reached a peak accuracy of 79% (p< 0.0001, not significant after Bonferroni correction for multiple comparisons) with top 8 to 21 features. Shaded grey region represents 95% CI. Brain images are displayed using Neurological convention (i.e. L=R).