Enhanced design matrix for task-related fMRI data analysis.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Signal Processing

Background:

• Task-related functional magnetic resonance imaging (fMRI) analysis typically relies on conventional General Linear Models (GLM).
• Standard GLM approaches face challenges in accurately modeling the hemodynamic response function (HRF) and accounting for unmodeled sources of variance.
• These unmodeled sources include physiological noise, scanner artifacts, and residual head motion, which can compromise the reliability of fMRI results.

Purpose of the Study:

• To introduce a novel methodology for analyzing task-related fMRI data using Information-Assisted Dictionary Learning (IADL).
• To enhance the GLM framework by providing a more robust approach to design matrix construction.
• To improve the handling of uncertainties and unmodeled signals in fMRI data analysis.

Main Methods:

• Development of the Information-Assisted Dictionary Learning (IADL) method for constructing the fMRI design matrix.
• Integration of IADL within the conventional GLM framework.
• Evaluation of the methodology using a synthetic fMRI-like dataset and a challenging real fMRI study.

Main Results:

• The proposed IADL method effectively handles uncertainties in HRF modeling.
• It accommodates unmodeled brain-induced sources, scanner artifacts, and motion residuals.
• The approach integrates prior knowledge of brain activity sparsity, leading to more consistent results compared to standard methods.

Conclusions:

• The IADL methodology offers a powerful alternative for analyzing task-related fMRI data.
• It enhances the robustness and reliability of fMRI analyses by addressing common sources of noise and uncertainty.
• A provided toolbox extension for SPM facilitates the application and reproducibility of this novel approach.