Bayesian technique for investigating linearity in event-related BOLD fMRI

Area of Science:

• Neuroimaging
• Systems Neuroscience
• Statistical Modeling

Background:

• Blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) is a key tool in neuroscience.
• Understanding the linearity or nonlinearity of the BOLD response is crucial for accurate physiological parameter estimation.
• Previous models often assumed linear time-invariant systems for BOLD fMRI data.

Purpose of the Study:

• To develop a rigorous statistical test for detecting linearity and nonlinearity in event-related BOLD fMRI systems.
• To apply this test to empirical fMRI data to characterize the nature of the BOLD response.
• To investigate individual differences in BOLD response linearity.

Main Methods:

• Modeling event-related BOLD fMRI data as a linear time-invariant system.
• Employing Bayesian inference techniques to develop a statistical test for linearity/nonlinearity.
• Applying the test to fMRI data from eight subjects using a visual stimulus (switching checkerboard).

Main Results:

• Group analysis of the fMRI data demonstrated a clearly nonlinear BOLD response.
• Individual subject analyses predominantly showed nonlinear responses.
• Evidence suggests potential dynamic shifts between linear and nonlinear BOLD response regimes within individuals.

Conclusions:

• The BOLD response system exhibits predominantly nonlinear characteristics, even at the individual level.
• The observed potential for linear-to-nonlinear regime switching has implications for personalized physiological parameter estimation.
• Hemodynamic response function estimates did not consistently reveal a post-stimulus undershoot in the event-related BOLD response.