Skip to main content
. 2016 May 30;11:802–812. doi: 10.1016/j.nicl.2016.05.017

Fig. 1.

Fig. 1

Performance of individual ROIs included in meta-ROIs for each of the five methods. For each method, regions in each row are ordered best to worst by their performance in the diagnostic separability (CN vs. MCI/AD) criteria, measured as an area under the receiver operating characteristic curve (AUROC), and plotted in the leftmost column. We show the six regions with the highest AUROC for each method, which together form the meta-ROIs. The second column plots the Spearman rank correlation (rho) between the ROI and total intracranial volume (TIV). The following two columns plot reliability of the measure in each ROI across successive scans, in the two Mayo and ADNI datasets respectively, as measured by intraclass correlation coefficients (ICC). In the ADNI reliability dataset, the ICCs are shown by manufacturer and indicated by colored points. The rightmost column plots the Spearman rank correlation between the ROI and negative Braak stage at autopsy. 95% confidence intervals are shown for all but the per-manufacturer Reliability-ADNI data. AUROC, rho, and ICC values are shown to the right of each panel. Numeric volumes listed for the ADNI reliability panel are for all manufacturers combined. Larger values (toward the right) are preferred in all columns with exception of the second, correlation with TIV, for which values near zero are preferred. Plots of performance for the top 20 ROIs by each method are included in the Supplementary material.