al. (2012) also studied the inter-scanner reliability of the FIRST software for segmentation (Patenaude et al. 2011), and found it to be high. However, it is overly optimistic to always expect high reproducibility from automated segmentations of brain MRI, and the reproducibility is region specific. For example, both FSL and FreeSurfer tend to do less well in segmenting small structures relative to larger structures. There are also differences in accuracy and reliability among different methods for automated segmentation of the brain (Shokouhi et al. 2011). Cortical thickness and other local gray matter density measures can show reduced reliability owing to sensitivity to image contrast variability, which becomes particularly challenging in multicenter studies (Schnack et al. 2010). The volumes of some structures, such as the caudate, may even show systematic biases in certain populations because of tissue class ambiguity that arises as a consequence of white matter degradation near gray matter structures. Some cortical regions are also difficult to delineate accurately due to the large intersubject variability. As noted below, one goal of ENIGMA has been to screen brain measures for reproducibility, heritability, and association with disease, to see which ones are likely to be promising for genetic analysis (we return
