Functional connectivity network estimation with an inter-similarity prior for mild cognitive impairment...

Area of Science:

• Neuroscience
• Medical Imaging Analysis
• Graph Theory

Background:

• Functional connectivity network (FCN) analysis is crucial for understanding brain patterns and diagnosing neurological disorders like Alzheimer's disease (AD) and Mild Cognitive Impairment (MCI).
• Challenges in FCN analysis stem from low-quality functional magnetic resonance imaging (fMRI) data and incomplete knowledge of brain structure.
• The inherent inter-similarity of FCNs, where similar brain regions exhibit similar connection patterns, offers a promising avenue for improvement.

Purpose of the Study:

• To develop an advanced functional brain network modeling scheme.
• To incorporate the inter-similarity prior into a graph-regularization term for more accurate FCN estimation.
• To enhance the discriminative power of FCNs for neurological disorder diagnosis.

Main Methods:

• Proposed a novel functional brain network modeling scheme.
• Integrated an inter-similarity prior into a graph-regularization term.
• Employed an efficient optimization algorithm for solving the model.
• Conducted experiments to differentiate Mild Cognitive Impairment (MCI) from normal controls using FCNs.

Main Results:

• The proposed method achieved a classification accuracy of 88.19% in distinguishing MCI from normal controls.
• Outperformed baseline and state-of-the-art methods in classification tasks.
• Post hoc analysis revealed that the method identifies more biologically meaningful functional brain connectivity patterns.

Conclusions:

• The developed functional brain network modeling scheme effectively utilizes inter-similarity priors.
• The method demonstrates superior performance in diagnosing Mild Cognitive Impairment.
• This approach yields more biologically relevant functional brain connectivity, advancing neurological disorder research.