Diagnosis of Mild Cognitive Impairment With Ordinal Pattern Kernel

Area of Science:

• Neuroimaging
• Machine Learning
• Computational Neuroscience

Background:

• Mild cognitive impairment (MCI) is a precursor to Alzheimer's disease (AD), making early diagnosis crucial for intervention.
• Graph kernels are used for brain disease diagnosis by comparing brain connectivity networks.
• Existing graph kernels often ignore edge weight information in connectivity networks, potentially limiting diagnostic accuracy.

Purpose of the Study:

• To propose a novel graph kernel, the ordinal pattern kernel, for enhanced similarity measurement between brain connectivity networks.
• To apply the ordinal pattern kernel to the classification of MCI.
• To leverage the weighted information within brain connectivity networks for improved diagnostic performance.

Main Methods:

• Development of the ordinal pattern kernel to capture topological and edge weight information in brain networks.
• Application of the ordinal pattern kernel to functional magnetic resonance imaging (fMRI) data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• Comparison of the ordinal pattern kernel's performance against state-of-the-art graph kernels for MCI classification.

Main Results:

• The proposed ordinal pattern kernel demonstrated superior performance in classifying MCI compared to existing graph kernels.
• The method effectively utilizes the ordinal patterns of edge weights in brain connectivity networks.
• Experimental results validate the effectiveness of the ordinal pattern kernel in a real-world neuroimaging dataset.

Conclusions:

• The ordinal pattern kernel offers a more comprehensive approach to analyzing brain connectivity networks by incorporating edge weight information.
• This novel kernel shows significant promise for the accurate and early diagnosis of mild cognitive impairment.
• The findings suggest a potential advancement in using graph-based machine learning for neurodegenerative disease detection.