Persistent homology for MCI classification: a comparative analysis between graph and Vietoris-Rips filtrations

Area of Science:

• Neuroscience
• Data Analysis
• Computational Topology

Background:

• Mild cognitive impairment (MCI) is linked to early neurodegeneration and subtle brain connectivity disruptions.
• Accurate classification of MCI subtypes is crucial for timely intervention and treatment.

Purpose of the Study:

• To explore the application of persistent homology for classifying mild cognitive impairment (MCI) subtypes.
• To compare the efficacy of Vietoris-Rips filtration and graph filtration in analyzing brain network topology.

Main Methods:

• Brain network topology was analyzed using functional magnetic resonance imaging (fMRI) time series data.
• Two persistent homology methods were employed: Vietoris-Rips filtration and graph filtration.
• Brain network structures were quantified using Wasserstein distance.

Main Results:

• Vietoris-Rips filtration demonstrated superior performance compared to graph filtration in brain network analysis.
• A maximum accuracy of 85.7% was achieved using Vietoris-Rips filtration for MCI classification within the Default Mode Network.
• The study utilized an in-house dataset for validation.

Conclusions:

• Vietoris-Rips filtration is a powerful tool for capturing complex brain network patterns.
• This topological data analysis technique offers a robust approach for the early diagnosis and precise classification of MCI subtypes.