Topological Data Analysis Captures Task-Driven fMRI Profiles in Individual Participants: A Classification Pipeline...

Area of Science:

• Neuroscience
• Data Science
• Mathematics

Background:

• Blood-oxygen-level-dependent (BOLD)-based functional magnetic resonance imaging (fMRI) is a primary tool for brain function research.
• BOLD fMRI signals have limitations including low signal-to-noise ratio and limited temporal/spatial resolution.
• The high dimensionality of BOLD signals offers opportunities for advanced data analysis techniques.

Purpose of the Study:

• To investigate the application of Topological Data Analysis (TDA) for characterizing functional brain signals.
• To compare the efficacy of TDA-based methods versus standard vectorization for analyzing fMRI data.
• To assess the utility of TDA in classifying task- and condition-induced brain activity patterns.

Main Methods:

• fMRI data were acquired from the anterior cingulate cortex (ACC) during a motor control task.
• fMRI signals were summarized using TDA methods (persistent homology, persistence landscapes) and standard vectorization.
• Machine learning (support vector classifiers) was employed to test classification accuracy of both data types.

Main Results:

• TDA-based classification accuracy consistently outperformed non-TDA based classification within each participant.
• The TDA analytic pipeline demonstrated a superior ability to characterize task- and condition-induced structure in ACC fMRI signals.
• These findings highlight the potential of TDA for revealing complex structures within regional fMRI data.

Conclusions:

• TDA provides a valuable framework for analyzing the inherent structure within regional fMRI signals.
• TDA methods can enhance the characterization of functional brain activity, particularly in high-dimensional datasets.
• The study suggests TDA's utility for exploring individual differences in brain signal structure in both healthy and clinical populations.