Figure 7
From: Jumping over baselines with new methods to predict activation maps from resting-state fMRI
Subject-wise Pearson r score benchmark results for 3 selected (poor, mediocre, good), 4 high performing models (MMP-RR-PCR, Rest-Task GICA-RR, MMP-RR-DR, AF-Mod) and two baseline models (Mean (Baseline), Group Z-stat). Poor (left): Motor-Right Hand; Mediocre (middle): Emotion: Faces; Good (right): Language: Math-Story. Experiment included 3–100 subjects for training. The group z-statistic baseline results are considerably worse than the group mean baseline. The 4 models largely resemble each other’s performance when sample sizes increase past 40 subjects.
