Flexible statistical modelling detects clinical functional magnetic resonance imaging activation in partially compliant subjects

Improving the reliability of single-subject fMRI by weighting intra-run variability

At present, functional magnetic resonance imaging (fMRI) is one of the most useful methods of studying cognitive processes in the human brain in vivo, both for basic science and clinical goals. Although neuroscience studies often rely on group analysis, clinical applications must investigate single subjects (patients) only. Particularly for the latter, issues regarding the reliability of fMRI readings remain to be resolved. To determine the ability of intra-run variability (IRV) weighting to consistently detect active voxels, we first acquired fMRI data from a sample of healthy subjects, each of whom performed 4 runs (4 blocks each) of self-paced finger-tapping. Each subject's data was analyzed using single-run general linear model (GLM), and each block was then analyzed separately to calculate the IRV weighting. Results show that integrating IRV information into standard single-subject GLM activation maps significantly improved the reliability (p=0.007) of the single-subject fMRI data. This suggests that taking IRV into account can help identify the most constant and relevant neuronal activity at the single-subject level.