Figure 5.

MNI152 template overlaid by mean relevance maps (highest relevance in yellow) obtained for all nine models. Unmasked and masked MRI classifiers obtain relevant image features from volumetric information (left and center columns). In contrast, the proposed ${\text{Graz}}^{+}$-method bases the classifier’s decision on deep brain image features, virtually independently of the registration method (right column). Heat maps are thresholded to the top 50% of the overall relevance. See “Methods” for description.