Modeling the BOLD correlates of competitive neural dynamics

Area of Science:

• Computational Neuroscience
• Cognitive Neuroscience
• Neuroimaging

Background:

• Winner-take-all (WTA) models are standard for decision-making tasks involving selection from multiple options.
• Existing random walk and diffusion models explain decision processes using psychometric and neurophysiological data.
• Model-based functional magnetic resonance imaging (fMRI) studies aim to identify neural correlates of decision-making, but the BOLD signature of WTA circuits remains unclear due to hemodynamic responses reflecting synaptic rather than spiking activity.

Purpose of the Study:

• To develop a biologically realistic spiking WTA neural network model.
• To integrate neurophysiological and brain imaging data through Synthetic Brain Imaging.
• To predict and analyze the blood-oxygen-level-dependent (BOLD) response of WTA circuits during decision-making tasks.

Main Methods:

• A biologically realistic spiking WTA model with coupled excitatory and inhibitory neural populations was created.
• Decision task difficulty was manipulated by adjusting input contrast (relative strength of competing options).
• Synthetic Brain Imaging transformed model simulation outputs to estimate BOLD responses and analyze their relationship with input contrast.

Main Results:

• The model's performance accuracy was analyzed across a parameter space.
• The relationship between input contrast and the peak BOLD response was determined for accurate task performance.
• The study identified conditions under which the model accurately performs the decision task.

Conclusions:

• Biologically plausible neural network models are crucial for interpreting brain imaging data.
• The developed spiking WTA model provides a framework for understanding the BOLD signature of decision-making.
• This approach aids in localizing neural correlates of cognition by grounding fMRI analyses in neurophysiological data.