Incorporation of spatial- and connectivity-based cortical brain region information in regularized regression:...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Cognitive Neuroscience

Background:

• Analyzing the link between brain structure, function, and neurocognition necessitates integrating data from multiple brain-imaging modalities.
• Regularization methods offer a novel way to enhance linear regression models by incorporating structural brain information.

Purpose of the Study:

• To develop and validate a novel regularization method that integrates structural connectivity and cortical distance for improved neuroimaging analysis.
• To assess the association between brain structure and neurocognitive outcomes, specifically vocabulary comprehension.

Main Methods:

• A specialized Tikhonov regularization method was employed, incorporating structural connectivity from Diffusion Weighted Imaging and cortical distance into the penalty term.
• A Laplacian matrix based on cortical surface proximity was utilized to inform regression coefficient estimation.
• The method's performance was evaluated through extensive simulation studies and application to Human Connectome Project data.

Main Results:

• The proposed regularization technique effectively integrated structural and functional brain data.
• The study identified associations between the cortical properties of the left hemisphere and vocabulary comprehension.
• Simulation studies demonstrated the robustness and efficacy of the developed method in realistic brain-imaging scenarios.

Conclusions:

• The novel regularization approach provides a principled and effective way to combine structural and functional brain information for neurocognitive studies.
• This method enhances the analysis of brain-behavior relationships, offering insights into cognitive functions like language.
• The findings highlight the utility of integrating multimodal neuroimaging data for a comprehensive understanding of brain function.