Distance-dependent consensus thresholds for generating group-representative structural brain networks.

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• Large-scale structural brain networks map white matter connectivity, crucial for cognition.
• Compromised brain networks are linked to neurocognitive deficits and maladaptive behaviors.
• Group-representative networks enhance signal-to-noise for studying brain organization.

Purpose of the Study:

• To identify limitations in current methods for generating sparse group-representative brain networks.
• To introduce and validate an alternative approach for constructing more accurate group-representative brain networks.
• To improve the fidelity of group brain network models to individual subject-level properties.

Main Methods:

• Developed a novel approach to generate group-representative brain networks that preserves individual connection-length distributions.
• Benchmarked the new method against existing approaches for network construction.
• Evaluated network statistics to compare the accuracy of different methods in recapitulating subject-level networks.

Main Results:

• Current sparse network generation methods overestimate short-range connections, misrepresenting brain network structure.
• The novel approach successfully preserves the connection-length distribution found in individual subjects.
• Networks generated by the new method better match subject-level network statistics and properties.

Conclusions:

• The proposed method offers a more accurate representation of large-scale structural brain networks compared to existing techniques.
• This improved network modeling preserves crucial features that support integrative brain function.
• The developed method is promising for future research into the fundamental organizational principles of brain networks.