Spurious correlations in surface-based functional brain imaging.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Functional MRI (fMRI) data analysis often involves mapping volumetric data to surface meshes.
• Common processing pipelines create meshes with non-uniform vertex spacing, denser in sulci than gyri.

Purpose of the Study:

• To explain the origins of spatial bias in fMRI surface mapping.
• To illustrate the impact of this bias on fMRI data analysis using in-silico models.
• To identify affected fMRI analyses and propose solutions.

Main Methods:

• Analysis of fMRI data processing pipelines for vertex spacing.
• Development and use of in-silico models simulating fMRI data.
• Evaluation of common fMRI analysis techniques for susceptibility to spatial bias.

Main Results:

• Uneven vertex spacing in surface meshes leads to artificially strong correlations between neighboring sulcal fMRI time series.
• This spatial bias, termed "gyral bias", causes anatomical folding information to leak into functional data.
• Common fMRI analyses like test-retest reliability, fingerprinting, and functional parcellations are shown to be affected.

Conclusions:

• The
• gyral bias
• is a significant artifact in surface-based fMRI analysis.
• This bias can lead to spurious findings and misinterpretations in neuroimaging research.
• Recommendations are provided to mitigate or correct for this spatial bias in fMRI data processing and analysis.