GEM-pRF: GPU-empowered mapping of population receptive fields for large-scale fMRI analysis

Area of Science:

• Neuroscience
• Computational Neuroscience
• Neuroimaging

Background:

• Population receptive field (pRF) mapping is crucial for understanding visual system organization.
• Existing pRF mapping methods face computational bottlenecks, limiting scalability and speed.
• Current techniques often compromise precision for speed or use slow iterative updates.

Purpose of the Study:

• To present a novel mathematical reformulation of the General Linear Model (GLM) for accelerated pRF estimation.
• To introduce GPU-Empowered Mapping of population Receptive Fields (GEM-pRF) software.
• To overcome the computational limitations of existing pRF mapping techniques.

Main Methods:

• Reformulated the General Linear Model (GLM) by orthogonalizing the design matrix.
• Enabled direct and fast computation of the objective function's derivatives, eliminating iterative refinement.
• Developed a GPU-Empowered Mapping of population Receptive Fields (GEM-pRF) software implementation.

Main Results:

• GEM-pRF dramatically accelerates pRF estimation with high accuracy.
• Validation with empirical and simulated data confirmed GEM-pRF's precision.
• Benchmarking showed a reduction in computation time by nearly two orders of magnitude compared to established tools.

Conclusions:

• GEM-pRF offers a significant advancement for large-scale fMRI retinotopic mapping.
• The reformulated GLM and GPU implementation provide a broadly applicable solution for neuroimaging.
• This approach has the potential to accelerate computational modeling across various neuroscience domains.