FIGURE 4.

Experiment #3: Group differences in normalization accuracy. Normalization accuracy was assessed in healthy controls, patients with middle cerebral artery (MCA) or posterior cerebral artery (PCA) strokes, and a third set of patients with strokes in various locations for validation. (a) For the standard ANTs_mni method, normalization accuracy worsens when lesion size (in number of voxels) increases (lower line). For a cohort‐specific template, normalization accuracy shows less dependence on lesion size, and normalized mutual information (NMI) is greater overall (upper line). (b) No differences in normalization accuracy were found when a cohort specific template was used, even for stroke subjects who were not included in the template. (c) When ANTs_mni was used, normalization was most accurate for healthy controls compared to stroke patients. ***p < .001; n.s. = nonsignificant (one‐way ANOVA). Shaded regions around the correlations represent 95% confidence intervals