A low-variance subspace underlies individual differences in resting state fMRI.

Area of Science:

• Neuroscience
• Brain Imaging
• Individual Differences Research

Background:

• Isolating stable individual differences in human brain activity from complex, noisy resting-state fMRI (rs-fMRI) data is a significant challenge.
• The existence and drivers of reliable dimensions of individual variation in brain activity are not well understood.

Purpose of the Study:

• To determine if highly reliable, low-dimensional subspaces of rs-fMRI activity exist.
• To investigate the origins of reliability in these brain activity dimensions.
• To explore the potential of these dimensions as personal fingerprints for identification and biomarker discovery.

Main Methods:

• Developed and applied Reliability Component Analysis (RCA), a novel dimensionality reduction technique.
• RCA prioritizes test-retest reliability over explained variance, unlike traditional methods like Principal Component Analysis (PCA).
• Analyzed rs-fMRI data to identify reliable dimensions of brain activity.

Main Results:

• Identified a low-dimensional linear subspace of highly reliable rs-fMRI activity.
• These reliable dimensions function as personal fingerprints, enabling accurate participant identification.
• Reliability of many dimensions is linked to morphological, demographic, or behavioral properties and predictable from cortical anatomy.

Conclusions:

• Stable individual signatures can be successfully isolated from rs-fMRI data.
• These signatures reflect persistent anatomical and physiological differences between individuals.
• The findings provide a principled, low-dimensional framework for discovering neuroimaging biomarkers.