Statistical parametric mapping of brain morphology: sensitivity is dramatically increased by using brain-extracted...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Imaging Analysis

Background:

• Voxel-based morphometry (VBM) enables large-scale neuroimaging data exploration with minimal manual input.
• The accuracy of Statistical Parametric Mapping 2 (SPM2) for VBM is debated due to segmentation errors.
• Common errors include misclassifying non-brain tissue as gray matter and excluding cortical gray matter.

Purpose of the Study:

• To evaluate the impact of using brain-extracted images in SPM2 for VBM analysis.
• To develop and describe modifications to the SPM2 pipeline for processing brain-extracted images.
• To assess the reduction in statistical errors and its effect on statistical power.

Main Methods:

• Visual inspection of SPM2 gray matter segmentations for 101 participants.
• Development of a modified SPM2 pipeline accepting brain-extracted images (non-brain tissues removed).
• Comparison of VBM analysis results using standard vs. brain-extracted images.

Main Results:

• Standard SPM2 processing incorrectly included non-brain tissue and excluded cortical gray matter in many subjects.
• Using brain-extracted images in the modified SPM2 pipeline eliminated these segmentation errors.
• Residual mean square errors (RMSEs) were reduced by over 30%.

Conclusions:

• Processing brain-extracted images significantly improves the accuracy of SPM2-based VBM analysis.
• This improved accuracy reduces statistical errors and enhances the power of neuroimaging studies.
• Utilizing brain-extracted images may allow for sample sizes to be halved, increasing research efficiency.