Within-subject reproducibility varies in multi-modal, longitudinal brain networks
Johan Nakuci1,2, Nick Wasylyshyn3,4, Matthew Cieslak5
1Neuroscience Program, University at Buffalo, SUNY, Buffalo, NY, 14260, USA. jnakuci3@gatech.edu.
Scientific Reports
|April 24, 2023
Summary
Reproducibility in brain network analysis is crucial. Structural brain networks derived from diffusion MRI are more reliable for identifying individuals than functional networks, especially when accounting for state-dependent fluctuations.
Area of Science:
- Neuroscience
- Network Science
- Neuroimaging
Background:
- Network neuroscience leverages neuroimaging (dMRI, fMRI, E/MEG) to study brain function.
- Understanding within- and between-subject variability is essential for reproducible research over time.
Purpose of the Study:
- To assess the reproducibility and reliability of brain networks across multiple modalities and sessions.
- To compare the stability of structural versus functional brain networks.
- To investigate the impact of state-dependent variability on network reproducibility.
Main Methods:
- Analysis of a longitudinal, 8-session, multi-modal dataset (dMRI, simultaneous EEG-fMRI) with tasks.
- Comparison of within-subject and between-subject reproducibility across modalities.
- Evaluation of network statistics reliability and individual identification using fingerprinting analysis.
Main Results:
- Within-subject reproducibility consistently exceeded between-subject reproducibility across all modalities.
- Alpha-band connectivity in EEG networks showed higher reproducibility during rest and task.
- Structural networks demonstrated higher reliability than functional networks; synchronizability and eigenvector centrality were less reliable.
- Structural dMRI networks were superior for individual identification compared to functional networks.
Conclusions:
- Brain network reproducibility varies significantly across modalities, connections, and network measures.
- Functional networks capture state-dependent variability, unlike structural networks.
- The choice of network analysis should consider the need to account for state-dependent fluctuations.
