Skip to main content
[Preprint]. 2024 Nov 19:2024.11.18.24317513. [Version 1] doi: 10.1101/2024.11.18.24317513

Figure 3. Within-cohort (ADNI) prediction performance.

Figure 3

(a) Left: Boxplots represent variability across 20 test sets for clinical diagnosis prediction measured using mAUC. Right: Statistical difference between all models. “***” indicates p < 0.00001 and statistical significance after multiple comparison correction (FDR q < 0.05). “**” indicates p < 0.001 and statistical significance after multiple comparison correction (FDR q < 0.05). “n.s.” indicates no statistical significance (p ≥ 0.05) or did not survive FDR correction. (b) Same as (a) but for MMSE prediction error (MAE). (c) Same as (a) but for ventricle volume prediction error (MAE).