Multi-level bootstrap analysis of stable clusters in resting-state fMRI

Area of Science:

• Neuroscience
• Computational Neuroscience
• Brain Imaging Analysis

Background:

• Resting-state functional magnetic resonance imaging (fMRI) is crucial for identifying intrinsic brain activity.
• Quantifying the stability of resting-state networks (RSNs) is essential for robust analysis.
• Existing methods for RSN identification lack a principled measure of stability.

Purpose of the Study:

• To introduce a generic statistical framework for quantifying the stability of resting-state networks (RSNs).
• To implement this framework using k-means clustering and a bootstrapping approach.
• To validate the method's performance in both simulated and real fMRI data.

Main Methods:

• Developed the Bootstrap Analysis of Stable Clusters (BASC) framework.
• Utilized k-means clustering and extensive bootstrapping for robust RSN identification.
• Applied a multi-level approach to analyze RSN stability at individual and group levels.

Main Results:

• The multi-level BASC demonstrated good performance on synthetic data.
• Seven stable RSNs were identified at the group level from real fMRI data, aligning with prior literature.
• BASC successfully mapped individual RSNs to group-level networks while preserving subject-specific details.

Conclusions:

• BASC provides a principled and reliable method for quantifying RSN stability in fMRI data.
• The multi-level analysis capability of BASC bridges individual and group-level network findings.
• BASC facilitates the identification of stable brain networks and their subject-specific variations.