Quantification of retinotopic maps with a Gaussian process modeling.

Area of Science:

• Neuroscience
• Human Brain Mapping
• Visual Cortex Research

Background:

• Retinotopic mapping using fMRI is crucial for understanding visual cortex organization.
• Current methods face challenges with limited resolution, low signal-to-noise ratio, and lack automated visual area segregation.
• Population receptive field (PRF) models aid estimation but struggle with topological accuracy.

Purpose of the Study:

• To develop an improved method for retinotopic map quantification in the human visual cortex.
• To address challenges in modeling cortical topology and automatically segregate visual areas.
• To enhance the accuracy and reliability of visual area boundary delineation.

Main Methods:

• Implemented an extended polar angle parametrization.
• Introduced cortical anchor point identification.
• Utilized a Gaussian process model for map estimation, outperforming linear regression.

Main Results:

• Reduced topological violations in retinotopic maps from 49.2% to 31.5%.
• Automatically defined precise boundaries between six discrete visual areas with a mean 95% credible interval width of 0.104 π rad.
• Estimated foveal confluence location to be systematically more dorsal and medial than the occipital pole.

Conclusions:

• The Gaussian process modeling approach offers a more accurate and reliable method for quantifying retinotopic maps.
• This method improves the delineation of visual areas and their boundaries.
• Findings provide a more precise understanding of visual field representations on the human cortex.