Group analysis in functional neuroimaging: selecting subjects using similarity measures

Area of Science:

• Neuroimaging
• Brain Imaging Analysis
• Statistical Neuroscience

Background:

• Standard functional MRI (fMRI) group analyses average individual data.
• Averaging assumes group homogeneity, which is often not met, potentially skewing results.
• Assessing inter-subject variability is critical for reliable population-level conclusions in fMRI.

Purpose of the Study:

• To develop and validate a method for quantifying inter-subject similarity in fMRI data.
• To enable multivariate comparison of temporal and spatial patterns across subjects.
• To identify outliers and define homogeneous subgroups for more robust fMRI group analyses.

Main Methods:

• Adaptation of the RV coefficient to measure spatial and temporal similarities between fMRI time series.
• Multidimensional scaling (MDS) for visualizing subject relationships and group structure.
• Development of outlier detection measures within the group.

Main Results:

• The proposed method effectively quantifies meaningful spatial and temporal similarities between subjects.
• Multidimensional scaling provides visual insights into subject positioning and group variability.
• Outlier detection successfully identifies subjects with distinct patterns.
• Analysis restricted to homogeneous subgroups maintains statistical sensitivity in fMRI.

Conclusions:

• The developed method is a powerful tool for assessing fMRI group homogeneity.
• Identifying and analyzing homogeneous subgroups enhances the reliability of population-level findings.
• This approach improves the validity of group fMRI studies by accounting for inter-subject variability.