CONECT: Novel Weighted Networks Framework Leveraging Angle-Relation Connection (ARC) and Metaheuristic Algorithms for...

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Dementia subtype classification using electroencephalography (EEG) is challenging.
• Traditional EEG analysis often overlooks crucial geometric and structural signal information.
• Existing methods may not fully capture the complexity of neural dynamics in dementia.

Purpose of the Study:

• Introduce CONECT (Complex Network Conversion and Topology), a novel framework for EEG analysis.
• Enhance the accuracy and interpretability of dementia subtype classification.
• Explore novel EEG biomarkers based on network topology and signal geometry.

Main Methods:

• Transformed EEG time series into weighted networks using a novel Angle-Relation Connection (ARC) rule.
• Developed a tunable edge-weighting function integrating amplitude, temporal, and angular components.
• Proposed new graph-based features: Weighted Angular Irregularity Index (WAII) and Curvature-Based Edge Feature Index (CBEFI).
• Applied Ant Colony Optimization (ACO) for feature selection on the OpenNeuro ds004504 dataset.

Main Results:

• The CONECT framework demonstrated potential for accurate dementia subtype classification.
• Novel graph-based features (WAII, CBEFI) showed promise as dementia biomarkers.
• Ant Colony Optimization improved classification performance and model transparency.
• The geometry-informed approach captured localized irregularity and signal geometry effectively.

Conclusions:

• CONECT offers a promising, interpretable, and geometry-informed framework for dementia diagnosis.
• The novel network-based approach advances EEG signal analysis in neuroscience.
• This method has the potential for practical application in clinical settings for dementia subtype identification.