Functional Mesh Model with Temporal Measurements for brain decoding.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Machine Learning

Background:

• Brain decoding aims to infer cognitive states from neural activity.
• Blood Oxygenation Level Dependent (BOLD) signals are commonly used to measure brain activity.
• Existing methods often rely on raw voxel intensities or simpler spatial relationships.

Purpose of the Study:

• To develop a novel method, Functional Mesh Model with Temporal Measurements (FMM-TM), for enhanced brain decoding.
• To estimate functional relationships among voxels using temporal data.
• To improve brain decoding accuracy by utilizing these estimated relationships.

Main Methods:

• Constructing a functional mesh around seed voxels using nearest neighbors based on correlation and distance metrics.
• Representing seed voxel BOLD responses as a linear combination of neighbors' responses.
• Estimating voxel-to-neighbor relationships using linear regression and employing weights as features for Support Vector Machine and k-Nearest Neighbor classifiers.

Main Results:

• The FMM-TM features demonstrated superior performance compared to raw voxel intensities.
• The proposed method outperformed features derived from various pairwise distance metrics.
• FMM-TM showed better results than local mesh model features using stationary and temporal measurements.

Conclusions:

• The Functional Mesh Model with Temporal Measurements (FMM-TM) offers a powerful approach for brain decoding.
• Leveraging temporal functional relationships between voxels significantly enhances decoding accuracy.
• This method provides a more sophisticated feature representation for neural decoding tasks.