Learning image derived PDE-phenotypes from fMRI data.

Area of Science:

• Neuroscience
• Computational Biology
• Applied Mathematics

Background:

• Partial differential equations (PDEs) are crucial for modeling physical systems but are underexplored in functional magnetic resonance imaging (fMRI) analysis.
• Existing methods like sparse identification of nonlinear dynamics (SINDy) and PDE-Net 2.0 offer data-driven PDE discovery.
• Understanding brain activity through PDEs could reveal novel insights into neurological disorders.

Purpose of the Study:

• To apply PDE modeling to fMRI data for identifying biomarkers of Attention Deficit Hyperactivity Disorder (ADHD).
• To explore the potential of PDEs in uncovering hidden brain activity patterns and essential components.
• To investigate the role of oxygen transport in brain activity related to neurological conditions.

Main Methods:

• Functional magnetic resonance imaging (fMRI) data from the ADHD200 dataset were analyzed.
• Dimensionality reduction was performed using canonical independent component analysis (CanICA) and uniform manifold approximation (UMAP).
• Sparse ridge regression was employed to identify significant partial differential equations (PDEs) from the reduced fMRI data.

Main Results:

• Identified significant PDE features from reduced fMRI data.
• Achieved high accuracy in classifying individuals with Attention Deficit Hyperactivity Disorder (ADHD).
• Demonstrated the utility of PDE-based feature extraction for neurological disorder analysis.

Conclusions:

• PDE modeling offers a novel and effective approach for analyzing fMRI data in the context of neurological disorders.
• The identified PDE features provide meaningful insights into brain activity, particularly concerning oxygen transport.
• This methodology holds promise for advancing the understanding and diagnosis of conditions like ADHD and other intracranial pathologies.