Uncovering Dynamic Brain Reconfiguration in MEG Working Memory n-Back Task Using Topological Data Analysis.

Area of Science:

• Neuroscience
• Data Science
• Computational Biology

Background:

• High temporal resolution neuroimaging data, including functional magnetic resonance imaging (fMRI), electroencephalogram (EEG), and magnetoencephalography (MEG), is increasingly available, driving efforts to understand neural function dynamics.
• Current state-of-the-art methods often collapse data spatially and temporally, leading to information loss during the analysis of complex neural systems.
• Generative models for neural system classification and prediction are less explored compared to architectural descriptions.

Purpose of the Study:

• To address the information loss in neuroimaging data analysis by applying a Topological Data Analysis (TDA) method, specifically Mapper.
• To visualize evolving patterns of brain activity as a mathematical graph using MEG data.
• To examine how variations in brain state dynamics, visualized through Mapper graphs, relate to performance measures like response time and accuracy during a working memory task.

Main Methods:

• Analysis of preprocessed MEG data from 83 subjects from the Human Connectome Project (HCP) during an n-back task.
• Application of the Mapper algorithm to visualize brain activity dynamics as a mathematical graph.
• Correlation of Mapper graph features with behavioral measures such as response time and task performance.

Main Results:

• Identification of a novel neuroimaging marker using the Mapper method on MEG data, which effectively explains participant performance and response times.
• Distinction of two task-positive brain activations during 0-back and 2-back tasks, which are difficult to discern with traditional methods requiring data collapse.
• Observation of a distinct group in Mapper graphs indicating high brain engagement with fine temporal resolution, suggesting potential for enhanced spatiotemporal resolution through multimodal imaging.

Conclusions:

• Topological Data Analysis (TDA) methods, particularly Mapper, are effective for extracting subtle dynamic properties from high temporal resolution MEG data without temporal and spatial collapse.
• The Mapper approach offers a valuable tool for understanding brain dynamics and identifying performance-related neuroimaging markers.
• This study highlights the potential of TDA in advancing neuroimaging analysis and potentially merging different imaging modalities for improved spatiotemporal resolution.