Fixed and random effect analysis of multi-subject fMRI data using wavelet transform.

Area of Science:

• Neuroimaging
• Statistical analysis
• Signal processing

Background:

• Group analysis of functional MRI (fMRI) data is crucial for understanding brain function.
• Accurate estimation of random effect variance is essential for robust group-level inference.
• Existing methods for random effect variance estimation in fMRI may have limitations in sensitivity and specificity.

Purpose of the Study:

• To propose and validate a new wavelet-based method for estimating random effect variance in fMRI group analysis.
• To enhance the detection of brain activation in group fMRI studies.
• To compare the performance of the proposed method against existing techniques.

Main Methods:

• Application of discrete wavelet transform to individual subject parameter maps from the general linear model (GLM).
• Noise reduction using vertical energy thresholding (VET).
• Identification of significant random effects based on sample variance of wavelet coefficients.
• Reconstruction of random effect maps using inverse wavelet transform for variance estimation.

Main Results:

• The proposed method demonstrated improved group activation detection on simulated fMRI data, as evidenced by Receiver Operating Characteristic (ROC) curves.
• Application to experimental fMRI data revealed high sensitivity, detecting activation in the visual cortex, cuneus, precuneus, thalamus, and cerebellum.
• Notably, the precuneus and cerebellum activations were identified by the proposed method but missed by most previous methods.

Conclusions:

• The novel wavelet-based method offers a sensitive and effective approach for estimating random effect variance in fMRI group analysis.
• This method enhances the ability to detect subtle or widespread brain activations.
• The findings suggest a significant advancement in statistical methodologies for neuroimaging research.