A simple and objective method for reproducible resting state network (RSN) detection in fMRI

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biostatistics

Background:

• Spatial Independent Component Analysis (ICA) is used to decompose functional MRI (fMRI) data into spatial independent components (ICs).
• When applied to resting-state fMRI (rsfMRI), ICA identifies spatially independent components (ICs) that represent resting-state networks (RSNs) with potential biological relevance.
• ICA's contrast function can lead to run-to-run variability in IC estimates due to data limitations and non-convex optimization.

Purpose of the Study:

• To address the run-to-run variability in ICA estimates for rsfMRI.
• To develop an enhanced algorithm, RAICAR-N, for objective assessment of RSN reproducibility within and across subjects.
• To identify reproducible RSNs in publicly available human rsfMRI data.

Main Methods:

• We propose RAICAR-N, an enhancement to the RAICAR algorithm, which assigns reproducibility p-values to each IC.
• RAICAR-N enables objective assessment of within-subject and across-subject reproducibility.
• The algorithm was applied to publicly available human rsfMRI datasets.

Main Results:

• Reproducibility analyses using RAICAR-N indicated that many published RSNs are highly reproducible.
• The study identified several highly reproducible RSNs that are not commonly reported in the literature.
• RAICAR-N provides a robust method for evaluating the reliability of RSNs derived from rsfMRI data.

Conclusions:

• The enhanced RAICAR-N algorithm offers an objective framework for assessing RSN reproducibility in rsfMRI.
• Many commonly reported RSNs demonstrate high reproducibility, validating their biological relevance.
• RAICAR-N also highlights novel, reproducible RSNs that warrant further investigation.