Estimation of false discovery rates for wavelet-denoised statistical parametric maps

Area of Science:

• Neuroimaging analysis
• Statistical signal processing

Background:

• Multiple comparison correction is crucial in neuroimaging.
• Wavelet methods offer improved sensitivity but yield smoothed, difficult-to-interpret statistical maps.
• Spatial domain thresholding of wavelet-denoised maps requires careful selection to control error rates.

Purpose of the Study:

• To propose a novel framework for thresholding wavelet-denoised statistical parametric maps (SPMs).
• To provide a strategy for selecting thresholds that control the false discovery rate (FDR).

Main Methods:

• Developed a framework for thresholding denoised SPMs by fixing a rejection region and estimating the achieved FDR.
• Utilized two FDR estimation algorithms to evaluate error rate control.
• Applied the framework to simulated and resting-state functional MRI (fMRI) data.

Main Results:

• The proposed framework offers a meaningful strategy for threshold selection in wavelet-denoised SPMs.
• Performance evaluation demonstrated effective error rate control using the proposed method.
• The framework was successfully applied to real in vivo neuroimaging data.

Conclusions:

• The developed framework facilitates more interpretable statistical maps in neuroimaging.
• This approach aids researchers in appropriately thresholding wavelet-processed data to maintain statistical rigor.
• The method is validated on both simulated and real fMRI datasets.