Extendable supervised dictionary learning for exploring diverse and concurrent brain activities in task-based fMRI.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Machine Learning

Background:

• Diverse brain activities, including task-evoked, delayed, and intrinsic ones, coexist during cognitive tasks.
• Traditional task-based functional magnetic resonance imaging (tfMRI) analysis using the general linear model (GLM) may not fully capture these concurrent activities.
• The GLM's subtraction-based approach primarily focuses on direct task responses, potentially missing other simultaneous brain processes.

Purpose of the Study:

• To propose a novel hybrid framework, extendable supervised dictionary learning (E-SDL), for exploring diverse and concurrent brain activities in tfMRI data.
• To address the limitations of existing methods in identifying multifaceted brain responses under task conditions.
• To enhance the analysis of brain activity by accounting for variations in task-related regressors.

Main Methods:

• Developed the extendable supervised dictionary learning (E-SDL) framework.
• Systematically extended basic task paradigm regressors into meaningful groups to capture regressor variations.
• Applied the E-SDL framework to five independent Human Connectome Project (HCP) tfMRI datasets.

Main Results:

• The E-SDL framework successfully identified more meaningful group-wise consistent task-evoked networks.
• The method also revealed common intrinsic connectivity networks (ICNs) concurrently.
• Demonstrated the framework's advantage in uncovering the diversity of concurrent brain activities.

Conclusions:

• The proposed E-SDL framework offers an improved approach for analyzing tfMRI data.
• It effectively captures diverse and concurrent brain activities, going beyond traditional GLM limitations.
• This advancement aids in a more comprehensive understanding of brain function during tasks.