Bootstrap generation and evaluation of an fMRI simulation database.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Statistical Modeling

Background:

• Functional magnetic resonance imaging (fMRI) research relies on computer simulations for validating data analysis methods.
• Existing fMRI simulation approaches lack a standardized framework for assessing their realism.
• Evaluating simulation realism is crucial for advancing fMRI data interpretation.

Purpose of the Study:

• To introduce and validate a parametric bootstrap framework for generating realistic fMRI simulations.
• To assess the strengths and limitations of two popular fMRI simulation models using the proposed framework.
• To establish a generic method for evaluating the fidelity of fMRI simulation techniques.

Main Methods:

• Generated an fMRI simulation database using a parametric bootstrap technique, mirroring parameters from a real database (40 subjects, five tasks).
• Evaluated simulations by comparing statistical measure distributions between real and simulated datasets.
• Applied the bootstrap framework to assess an additive mixture model and a non-linear motion process model.

Main Results:

• Both simulation models successfully reproduced key brain dynamics (baseline, physiological noise, neural activation, random noise), relative variance, and temporal autocorrelation.
• The additive model exhibited significantly lower spatial autocorrelation compared to real data.
• The non-linear motion model intrinsically generated slow time drifts and increased spatial autocorrelation, improving realism.

Conclusions:

• The parametric bootstrap framework is a powerful tool for evaluating fMRI simulation models.
• The framework can identify specific strengths and weaknesses of different simulation approaches.
• This methodology advances the development of more realistic and reliable fMRI simulations for research.