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

Correction for multiple comparisons in neuroimaging data is an important area of research. Recently, wavelet-based methods have gained popularity and have been reported to achieve better sensitivity compared to spatial domain methods. However, these techniques produce smoothed statistical maps which are difficult to interpret. The generated maps have to be thresholded again in the spatial domain to delineate active from inactive regions. The selection of a proper threshold satisfying the required error rate control is not straightforward. In this paper, a framework is proposed for thresholding wavelet-denoised maps in which a rejection region is fixed, and the achieved false discovery rate (FDR) is estimated. This approach provides a meaningful strategy to choose thresholds for wavelet-denoised statistical parametric maps (SPMs). Two FDR estimation algorithms were used to assess the achieved error rate control when thresholding wavelet filtered SPMs at various rejection regions. Their performance was evaluated using both simulated and resting fMRI data. The proposed framework was also applied on in vivo data.