Comparison of Fourier and wavelet resampling methods

Area of Science:

• Neuroimaging
• Signal Processing
• Statistical Analysis

Background:

• Resampling is crucial for calculating null distributions to assess statistical significance.
• Functional Magnetic Resonance Imaging (fMRI) generates time series data requiring robust analytical methods.
• Understanding spatial and temporal correlations in fMRI data is key to accurate interpretation.

Purpose of the Study:

• To evaluate the preservation of spatial and temporal correlations in fMRI time series using Fourier and wavelet resampling.
• To compare the null distributions generated by these resampling techniques.
• To assess the accuracy of Fourier versus wavelet resampling in localizing activated voxels in fMRI data.

Main Methods:

• Simulated and experimental fMRI time series data were analyzed.
• Fourier resampling and wavelet resampling methods were applied.
• Null distributions of test statistics were computed and compared.
• Receiver Operating Characteristic (ROC) analysis was used to evaluate voxel localization accuracy.

Main Results:

• Wavelet resampling demonstrated superior preservation of spatial and temporal correlations compared to Fourier resampling.
• The null distributions generated by both methods were compared.
• ROC analysis indicated that wavelet resampling achieved higher accuracy in identifying activated voxels.

Conclusions:

• Wavelet resampling is a more effective method than Fourier resampling for analyzing fMRI time series data.
• The findings suggest wavelet resampling enhances the accuracy of brain activation localization in fMRI studies.
• This improved accuracy has implications for the statistical validity of fMRI analyses.