Variance in clinical outcomes explained by different models. Predictive utility of the R-map calculated in each target (left) and an agreement map, in which only regions predictive for both targets were retained (right). Using the the R-map calculated on either target alone, a significant portion of variance in outcomes in both cohorts could be explained (R = 0.27 at P < 0.001 for STN; R = 0.32 at P = 0.042 for GPi). The agreement map was able to explain additional variance in each of the two cohorts (R = 0.34 at P < 0.001 for STN; R = 0.39 at P = 0.022). Single target maps on the left correspond to renderings in Fig. 3 and are shown as volumetric cuts at z = −50, −30, −10, 10, 30 and 50 mm. The agreement map is shown both in volumetric and surface fashion. The BigBrain atlas served as the backdrop for volumetric representations.²⁵