Estimating Functional Brain Networks by Low-Rank Representation With Local Constraint

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• Alzheimer's disease (AD) alters brain functional architecture, particularly in preclinical stages.
• Functional connectivity networks (FCNs) from resting-state fMRI are key for early AD detection.
• Existing methods often overlook network modularity, a critical aspect of brain function.

Purpose of the Study:

• To propose a novel method for constructing Functional Brain Networks (FBNs) that incorporates modularity information.
• To address the limitations of current sparse learning methods that focus solely on functional connectivity.

Main Methods:

• Introduced a local similarity-constrained low-rank sparse representation (LSLRSR) method.
• Formulated the problem as a low-rank sparse graph learning task solved via an efficient optimization algorithm.
• Utilized manifold regularization and the alternating direction method of multipliers (ADMM) for solving nonconvex optimization problems.

Main Results:

• The proposed LSLRSR method generates FBNs with enhanced modularity compared to state-of-the-art techniques.
• Demonstrated the ability to encode and leverage modularity information within the network construction framework.

Conclusions:

• The developed algorithm provides a more comprehensive delineation of FBNs by integrating modularity.
• This work offers a foundation for investigating disease-related alterations in brain network modularity, particularly in Alzheimer's disease.