Adaptive control of the false discovery rate in voxel-based morphometry

Area of Science:

• Neuroimaging
• Statistical analysis
• Brain morphology

Background:

• Voxel-based morphometry (VBM) is a high-resolution neuroimaging technique for studying brain structure-function relationships.
• Controlling the false discovery rate (FDR) is crucial for thresholding statistical maps in VBM and fMRI studies.
• Nonadaptive FDR procedures, like Benjamini-Hochberg (B-H), may be overly conservative in VBM due to the nature of anatomical data.

Purpose of the Study:

• To evaluate the effectiveness of adaptive FDR control procedures in VBM analyses.
• To demonstrate the limitations of nonadaptive FDR control (e.g., B-H) in VBM.
• To advocate for the adoption of adaptive FDR methods in VBM research.

Main Methods:

• Mathematical derivations and simulations were used to assess FDR control procedures.
• Two adaptive FDR procedures were examined: Benjamini, Krieger, and Yekutieli (2006) and Storey and Tibshirani (2003).
• A real VBM data example was used for empirical validation.

Main Results:

• Nonadaptive FDR control (B-H) can lead to substantial conservatism in VBM, reducing statistical power.
• Adaptive FDR procedures demonstrated improved performance over the B-H procedure.
• The examined adaptive methods offer a more powerful approach for detecting true associations in VBM.

Conclusions:

• Adaptive FDR control is recommended for VBM analyses to enhance statistical power.
• Researchers should consider adaptive FDR methods over nonadaptive ones for more robust findings in neuroimaging.
• The study provides evidence for improved VBM analysis through adaptive statistical thresholding.