Geometry Reduced Order Modeling (GROM) with application to modeling of glymphatic function.

Area of Science:

• Computational neuroscience
• Biomedical engineering
• Medical imaging analysis

Background:

• Computational brain modeling is crucial for understanding metabolic waste clearance but faces scalability challenges.
• Patient-specific brain models are currently computationally prohibitive for large-scale simulations.
• Existing methods struggle to meet the demands of high-throughput, personalized brain modeling.

Purpose of the Study:

• To develop and validate a novel computational approach for accelerating patient-specific brain models.
• To leverage model order reduction (MOR) techniques to decrease computational costs in brain simulations.
• To enable more feasible and scalable computational modeling of brain functions like glymphatic transport.

Main Methods:

• Utilized image registration based on magnetic resonance imaging (MRI) to create inter-brain mappings.
• Applied model order reduction techniques to computational models of brain geometries.
• Mapped previously computed simulation solutions onto new patient-specific brain geometries.
• Investigated the approach on two glymphatic function modeling problems using a dataset of 101 human MRIs.

Main Results:

• Achieved a speedup factor exceeding 750 times compared to full-order simulations for both example problems.
• Introduced minimal additional system assembly overhead.
• Maintained accuracy, with reduced solutions recovering full-order solutions within a 10% error margin in most cases.
• Demonstrated applicability across different neurological conditions, including idiopathic Normal Pressure Hydrocephalus.

Conclusions:

• Model order reduction, integrated with inter-brain mapping from MRI data, significantly accelerates patient-specific brain simulations.
• This novel technique makes high-throughput computational modeling of brain functions, such as glymphatic transport, more feasible.
• The method holds promise for advancing personalized medicine in neuroscience by enabling efficient patient-specific computational analysis.