Model-based approaches to neuroimaging: combining reinforcement learning theory with fMRI data

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Psychiatry

Background:

• Functional magnetic resonance imaging (fMRI) and computational models offer a robust framework for investigating neural computations in the brain.
• Reinforcement learning (RL) models can effectively explain human choice behavior in value-based decision-making tasks.

Purpose of the Study:

• To outline the methodology for integrating fMRI with computational models, specifically RL, for cognitive neuroscience research.
• To demonstrate how internal model variables can be used as predictors for fMRI data analysis.
• To review neuroimaging studies that have applied this approach to reward-related decision-making.

Main Methods:

• Utilizing reinforcement learning (RL) models to generate internal variables representing cognitive processes.
• Constructing fMRI predictor variables from these internal model variables.
• Regressing these predictors against individual subjects' fMRI data to correlate with blood oxygenation level dependent (BOLD) activity.

Main Results:

• The regression coefficients quantify the correlation between BOLD activity and the internal variables of the RL model.
• This approach successfully identifies brain regions involved in reward-related decision-making computations.
• The review highlights specific neuroimaging studies employing this integrated analysis strategy.

Conclusions:

• The combination of fMRI and computational modeling, particularly RL, provides a powerful tool for understanding the neural basis of decision-making.
• This framework enables the testing of hypotheses about specific computations underlying cognitive processes.
• The reviewed studies demonstrate the utility of this approach in mapping reward computations to brain regions.