Empirical and substantive models, the Bayesian paradigm, and meta-analysis in functional brain imaging

Area of Science:

• Neuroimaging
• Statistical Modeling
• Brain Function Analysis

Background:

• Functional neuroimaging research utilizes established statistical methods.
• Methods include experimental design, general linear models, random field theory, and signal processing.
• Adapting these methods for neuroimaging data is an ongoing process.

Purpose of the Study:

• To compare the performance of five distinct statistical models on fMRI data.
• To analyze spatial activation patterns in the motor cortex.
• To evaluate the incorporation of prior information in statistical models for neuroimaging.

Main Methods:

• Application of five statistical models: t-statistic, Kolmogorov-Smirnov, principal component/canonical variates, artificial neural network, and frequency domain regression.
• Analysis of fMRI data from a motor cortex study.
• Comparison of spatial activation patterns generated by each model.

Main Results:

• Similar spatial activation patterns were observed across the five empirical statistical models.
• All models demonstrated a lack of explicit incorporation of substantive prior information.
• Distinctions were made between exploratory and confirmatory statistical models.

Conclusions:

• Empirical statistical models yield comparable results in fMRI spatial activation.
• Integrating substantive prior information is crucial for advancing neuroimaging analysis.
• Bayesian paradigms and meta-analysis offer frameworks for combining empirical and prior information, mitigating publication bias.