MEDUSA: A GPU-based tool to create realistic phantoms of the brain microstructure using tiny spheres

Area of Science:

• Neuroscience
• Biophysics
• Computational Biology

Background:

• Realistic modeling of brain microstructure is crucial for understanding neural function.
• Existing methods for generating microstructural phantoms can be computationally intensive and struggle with accurate packing and collision avoidance.

Purpose of the Study:

• To develop a novel, efficient GPU-based tool for generating realistic brain microstructure phantoms.
• To accurately simulate white matter components, including axonal fibers, oligodendrocytes, and astrocytes.

Main Methods:

• A spherical meshing technique was employed to decompose microstructural elements into overlapping spheres.
• This approach facilitates rapid phantom construction while preventing inter-object collisions.
• The method utilizes Graphics Processing Units (GPUs) for accelerated computation.

Main Results:

• The tool successfully generated realistic human brain white matter microstructural phantoms.
• High packing density and angular dispersion were achieved for axonal fibers, even with multiple fiber populations.
• Complex biomimicking geometries, including myelinated and beaded axons and glial cells, were constructed.

Conclusions:

• The developed GPU-based tool offers a fast and reliable method for creating complex brain microstructure phantoms.
• The technique is adaptable for modeling both white and gray matter microstructures.
• This advancement facilitates more accurate in silico neuroscience research.