Single scan, subject-specific component extraction in dynamic functional connectivity using dictionary learning.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Connectivity Analysis

Background:

• Individual differences in brain activity are crucial for understanding healthy controls.
• Functional connectivity (FC) patterns from fMRI are used for subject identification.
• Temporal variability in FC, or dynamic FC (dFC), is an emerging area for subject identifiability.

Purpose of the Study:

• To propose a novel method for subject identification using dynamic functional connectivity (dFC).
• To extract subject-specific components from a single fMRI scan using dictionary learning (DL).
• To evaluate the reusability of learned dictionaries for new subjects.

Main Methods:

• Utilized dynamic functional connectivity (dFC) derived from fMRI data.
• Applied a dictionary learning (DL) algorithm to extract subject-specific components.
• Validated the method on Human Connectome Project (HCP) and Nathan Kline Institute (NKI) datasets.

Main Results:

• Achieved a significant increase in subject identification accuracy from 89.19% to 99.54%.
• Demonstrated successful application using the Schaefer atlas with subcortical nodes from the HCP atlas.
• Found no significant differences in identification accuracy between groups of twins and unrelated subjects.

Conclusions:

• The proposed DL-based method effectively extracts subject-specific dFC components.
• The method enhances subject identification accuracy using single fMRI scans.
• Learned dictionaries can be stored and reused for identifying new subjects.