Advancing temporal dynamics in spatial Random Field Theory: A framework for fMRI signal detection with simulation...

Area of Science:

• Neuroimaging
• Statistical analysis
• Brain activity mapping

Background:

• Functional magnetic resonance imaging (fMRI) data possess complex spatiotemporal dependencies.
• Traditional statistical models often overlook these dynamics, limiting the detection of evolving neural activity crucial for task-based studies.

Purpose of the Study:

• To develop a novel statistical framework for fMRI data analysis that accounts for temporal dependencies.
• To enhance sensitivity to dynamic and distributed neural signals in neuroimaging.

Main Methods:

• Proposed a time-adaptive random field framework integrating a functional autoregressive process (FAR(1)).
• Introduced a new statistic, Xmax, for spatiotemporal signal detection.
• Compared Xmax against existing methods like Ymax (Location-Scale RFT) and GLM.

Main Results:

• Xmax demonstrated significantly higher detection power in simulations compared to Ymax and GLM, exceeding 0.85 power.
• Maintained stable Type I error control.
• Real fMRI data analysis revealed Xmax yields more spatially extensive and coherent activation patterns than GLM.

Conclusions:

• Incorporating temporal dependence in fMRI analysis improves sensitivity to dynamic neural signals.
• The proposed framework enables more effective detection of spatially distributed brain activity.
• Offers a powerful approach for spatiotemporal analysis in neuroimaging.