Accurate module induced brain network construction for mild cognitive impairment identification with functional MRI

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• Functional brain networks (FBNs) from fMRI are crucial for diagnosing neurological disorders like mild cognitive impairment (MCI).
• Traditional Pearson's correlation (PC) methods create overly dense FBNs, not reflecting sparse brain connectivity.
• Existing sparse FBN methods often overlook important topological structures like modularity.

Purpose of the Study:

• To develop an accurate module induced PC (AM-PC) model for estimating FBNs with clear modular structures.
• To improve the construction of sparse and modular FBNs by incorporating Laplacian matrix constraints.
• To enhance the diagnostic capability for neurological disorders using improved FBN analysis.

Main Methods:

• Proposed the accurate module induced PC (AM-PC) model incorporating sparse and low-rank constraints on the graph Laplacian matrix.
• Utilized the property of zero eigenvalues in the Laplacian matrix to control the number of network modules.
• Applied the AM-PC model to resting-state fMRI data for FBN estimation.

Main Results:

• The AM-PC model successfully estimated FBNs with distinct modular structures.
• The method achieved better classification performance in distinguishing MCI subjects from healthy controls compared to previous approaches.
• Validation was performed on 143 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Conclusions:

• The AM-PC model provides a more biologically plausible and effective method for FBN construction.
• This approach shows significant potential for improving computer-aided diagnosis of MCI and related neurological conditions.
• The findings highlight the importance of modularity in FBNs for understanding brain function and disease.