A Joint Analysis of Multi-Paradigm fMRI Data With Its Application to Cognitive Study

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Machine Learning

Background:

• Advancements in neuroimaging yield vast multi-modal brain data, necessitating sophisticated analytical approaches.
• Understanding individual cognitive behaviors requires integrated analysis of diverse brain imaging datasets.

Purpose of the Study:

• To introduce a novel multi-view learning algorithm, structure-enforced collaborative regression (SCoRe), for analyzing functional magnetic resonance imaging (fMRI) data.
• To enhance the biological relevance and predictive power of brain imaging analysis by incorporating anatomical structure.

Main Methods:

• Developed SCoRe, a multi-view learning algorithm that integrates anatomical priors into collaborative regression.
• Applied SCoRe to fMRI data from the Philadelphia Neurodevelopmental Cohort.
• Utilized Wide Range Achievement Test (WRAT) scores to assess cognitive skills.

Main Results:

• SCoRe demonstrated superior prediction performance compared to traditional collaborative regression (CoRe) by leveraging brain structure.
• The algorithm proved less sensitive to hyper-parameters than CoRe.
• Identified 14 brain regions significantly predicting WRAT scores, consistent with independent research.

Conclusions:

• SCoRe offers a biologically meaningful and effective approach for analyzing multi-modal brain imaging data.
• Incorporating anatomical structure improves the identification of brain regions associated with cognitive abilities.
• The identified brain regions provide valuable insights into the neural underpinnings of cognitive skills.