Topological graph kernel on multiple thresholded functional connectivity networks for mild cognitive impairment...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Brain connectivity networks are increasingly utilized for classifying neurological conditions like Alzheimer's disease and mild cognitive impairment (MCI).
• Traditional methods often extract local network features, potentially losing crucial global topological information.
• Accurate classification of MCI is vital for early intervention and treatment.

Purpose of the Study:

• To develop and evaluate a novel connectivity-networks-based classification framework for accurately identifying mild cognitive impairment (MCI) patients from normal controls (NC).
• To address the limitation of losing global topological information in conventional feature extraction methods.

Main Methods:

• Proposed a new graph-kernel-based approach to directly measure topological similarity between brain connectivity networks.
• Applied the framework to functional connectivity networks derived from 12 MCI and 25 NC subjects.
• Utilized classification and connectivity analysis to evaluate the method's performance and identify discriminative brain regions.

Main Results:

• Achieved a classification accuracy of 91.9% for MCI detection.
• Demonstrated high sensitivity (100.0%), balanced accuracy (94.0%), and an AUC of 0.94.
• Connectivity analysis revealed reduced functional connectivity in MCI patients compared to normal controls, consistent with existing literature.

Conclusions:

• The proposed graph-kernel-based framework shows significant potential for accurate MCI classification using brain connectivity networks.
• The method effectively captures global topological information, outperforming traditional approaches.
• Findings highlight altered functional connectivity patterns as a key biomarker for MCI.