Illustration of the results of the classification performed on the dyslexic subjects. In A and B, the right cerebellar and lentiform nucleus clusters observed for each classification are presented (classification performed on CI computed after 3000, 3500, 4000, 4500 and 5000 resamples) as well as the final classification (the average of previous classifications). Each cluster was homogenous as illustrated in C: Among the 21 voxels showing 100% of differences between groups, two clusters can be observed with a correlation of ~1, i.e. voxels in each cluster have identical classification values (Spearman rank correlation), and identical (or nearly identical) grey matter volume values (Pearson correlations). As illustrated in D, all dyslexic subjects (red dots) were located outside the bootstrapped 95% CI (blue lines) of local grey matter volumes (LGMV). Those values are the observed value in each voxel and derived from preprocessing (step 1). Simulations (5000 resamples) of the dyslexics' distributions (red histograms) compared to the control distributions (blue histograms) show a clearer separation over the right cerebellum compared to the lentiform nucleus.