An evaluation of thresholding techniques in fMRI analysis

Area of Science:

• Neuroimaging
• Statistical analysis
• Brain activity mapping

Background:

• Identifying brain regions with significant activity in functional magnetic resonance imaging (fMRI) is crucial for neuroscience research.
• Single-subject analysis requires robust methods for voxel-wise thresholding to detect activation.
• Various statistical approaches exist, differing in their handling of spatial correlation and error rates.

Purpose of the Study:

• To review and compare different voxel-wise thresholding methods for single-subject fMRI data.
• To evaluate the impact of various error rates and spatial correlation handling on activation detection sensitivity.
• To provide recommendations for effective thresholding in fMRI studies.

Main Methods:

• Review of individual voxel-wise thresholding techniques.
• Description of different error rates for threshold calibration.
• Discussion of methods controlling error rates at a prespecified level alpha.
• Simulation study to assess operating characteristics of thresholding methods.
• Illustration with a real bilateral finger tapping fMRI experiment.

Main Results:

• The choice of error rate significantly impacts the sensitivity of fMRI thresholding methods.
• Accounting for spatial correlation among test statistics had minimal impact on method performance.
• Simple thresholding procedures that ignore spatial correlation performed well.
• Simulation results demonstrated the effectiveness of simpler approaches.

Conclusions:

• Simple voxel-wise thresholding methods are effective and recommended for most single-subject fMRI experiments.
• The impact of spatial correlation is less critical than the choice of error rate for thresholding.
• These findings offer practical guidance for analyzing fMRI data and identifying brain activation.