Estimation of connectional brain templates using selective multi-view network normalization.

Area of Science:

• Neuroscience
• Network Science
• Medical Imaging Analysis

Background:

• The brain connectome, representing brain structure and function as a network, captures individual signatures.
• Extracting a shared, representative signature from multiple brain networks (multi-view) across a population is challenging.
• Existing methods often integrate network views equally, lacking a selective approach for population-level analysis.

Purpose of the Study:

• To introduce netNorm, a novel method for normalizing populations of multi-view brain networks.
• To develop a technique for extracting a representative population brain signature from diverse neuroimaging data.
• To create a robust method for identifying deviations in brain connections indicative of disorders.

Main Methods:

• NetNorm employs a selective, local-scale fusion technique for multi-view brain networks.
• It selects representative cross-view feature vectors for pairwise connectivity between regions of interest.
• A population representative tensor is estimated and non-linearly fused into a single brain connectional template (CBT).

Main Results:

• NetNorm generates a centered and representative CBT that captures unique population traits from multi-view networks.
• The method successfully identifies disordered brain connections by comparing templates from healthy and disordered groups.
• Demonstrated broad applicability across four connectomic datasets, highlighting the discriminative power of estimated CBTs.

Conclusions:

• NetNorm effectively normalizes multi-view brain networks to produce a representative population connectional template.
• The generated CBTs accurately reflect population-specific brain connectional characteristics.
• NetNorm offers an efficient method for spotting atypical brain deviations without requiring machine learning for feature identification.