Iterative Data-adaptive Autoregressive (IDAR) whitening procedure for long and short TR fMRI

Area of Science:

• Neuroimaging
• Statistical analysis
• Brain function research

Background:

• Functional magnetic resonance imaging (fMRI) is crucial for brain studies.
• Serial correlations in fMRI data pose significant analytical challenges and can lead to erroneous conclusions.
• Conventional statistical methods struggle with the complex serial correlations inherent in fMRI data.

Purpose of the Study:

• To evaluate the inadequacy of conventional whitening procedures for short-TR fMRI data.
• To introduce and assess an iterative whitening approach, IDAR, for addressing serial correlations.
• To compare the performance of IDAR against traditional methods in analyzing fMRI data.

Main Methods:

• Investigated shortcomings of existing whitening methods for fMRI data.
• Developed and implemented an iterative data-adaptive autoregressive (IDAR) model.
• Employed high-order autoregressive models with data-driven orders to capture complex serial correlations.

Main Results:

• Conventional whitening methods (AR(1), ARMA(1,1)) were ineffective for short-TR fMRI data.
• IDAR significantly improved serial correlation reduction, statistical power, and Type-I error control compared to conventional methods, particularly for short-TR data.
• IDAR faced limitations in simultaneously managing residual correlations and inflated Type-I error rates.

Conclusions:

• There is a critical need to address serial correlations in short-TR fMRI data due to their increasing prevalence.
• IDAR offers a viable solution for various fMRI datasets but requires further refinement for optimal performance.
• Future research should focus on innovative approaches to simultaneously mitigate serial correlations and control Type-I error without sacrificing analytical power.