A hybrid CPU-GPU accelerated framework for fast mapping of high-resolution human brain connectome

Area of Science:

• Neuroscience
• Computational Biology
• Graph Theory

Background:

• Understanding the human brain connectome relies on neuroimaging and graph theory.
• High-resolution connectome research demands significant computational power.
• Existing methods face computational challenges with large datasets.

Purpose of the Study:

• To propose and evaluate a hybrid CPU-GPU framework for accelerating human brain connectome computation.
• To enhance the analysis of high-resolution voxel-based brain networks.
• To facilitate the mapping of the human brain connectome in various states.

Main Methods:

• Developed a hybrid CPU-GPU framework for computational acceleration.
• Applied the framework to a resting-state functional MRI dataset (197 participants).
• Computed Pearson's Correlation coefficients and constructed brain networks (58k nodes).
• Quantified and analyzed graph properties of functional brain networks.

Main Results:

• The framework significantly accelerated the computation of high-resolution functional brain networks (80-150 minutes per network).
• Functional brain networks exhibit efficient small-world properties and modular structures.
• Identified highly connected nodes in medial frontal and parietal cortical regions.
• Results align with previous human brain network studies.

Conclusions:

• The proposed hybrid framework substantially improves the efficiency of high-resolution brain network analysis.
• This approach accelerates the mapping of the human brain connectome.
• The framework has potential applications in studying brain connectomes in normal and disease states.