Modeling state-related fMRI activity using change-point theory.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Statistical Modeling

Background:

• The general linear model (GLM) is standard for functional magnetic resonance imaging (fMRI) analysis but struggles with imprecise psychological event timing.
• Accurate modeling of fMRI data requires methods that can handle uncertainty in the onset and duration of cognitive processes.

Purpose of the Study:

• To introduce a new non-linear analysis approach for fMRI data, accommodating uncertainty in psychological event timing and duration.
• To develop a multi-subject extension of the exponentially weighted moving average (EWMA) method for fMRI, termed Hierarchical EWMA (HEWMA).

Main Methods:

• Developed HEWMA, a hierarchical model extending EWMA for change-point analysis to multi-subject fMRI data.
• Utilized concepts from statistical control theory and change-point theory to model slowly varying processes with uncertain event onsets/durations.
• Validated HEWMA's false-positive rate control and estimated statistical power through simulations using real fMRI data.

Main Results:

• HEWMA provides an exploratory yet inferentially capable alternative to the GLM for fMRI analysis.
• The method demonstrated effective control of false-positive rates and provided reliable power estimates in simulations.
• Applied HEWMA to an fMRI study of state anxiety in 24 participants.

Conclusions:

• HEWMA is a robust and flexible method for analyzing fMRI data, particularly when psychological event timing is not precisely known.
• The HEWMA approach can be applied voxel-wise, to regions of interest, or to components from independent component analysis (ICA).
• A freely available MATLAB toolbox facilitates the application of HEWMA in fMRI research.