Group-ICA Model Order Highlights Patterns of Functional Brain Connectivity.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Psychiatry

Background:

• Resting-state networks (RSNs) are reliably detected using independent component analysis (ICA).
• ICA model order influences RSN spatial characteristics and sub-network parcellation.
• Brain organization exhibits a modular hierarchy potentially mirrored by ICA model orders.

Purpose of the Study:

• To investigate how ICA model order selection affects the detection of functional connectivity differences between individuals with seasonal affective disorder (SAD) and healthy controls.
• To determine if specific model orders optimize the identification of disease-related functional connectivity alterations.

Main Methods:

• Utilized dual regression to assess functional connectivity differences.
• Compared unmedicated SAD patients with healthy controls.
• Systematically varied ICA model order to evaluate its impact on between-group differences.

Main Results:

• Disease-related functional connectivity differences varied significantly with ICA model order.
• Maximum between-group differences were observed at model order 70, suggesting an optimal point.
• Higher model orders revealed finer RSN details but increased the risk of false positives.

Conclusions:

• Fine-grained RSNs detected with higher ICA model orders enhance sensitivity to subtle functional connectivity changes in disorders like SAD.
• Multilevel ICA exploration is recommended for optimizing sensitivity to brain disorders while managing false positive risks.