Computational parametric mapping of functional neuroimaging data.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Neuroimaging Analysis

Background:

• Understanding the neural basis of cognition necessitates integrating theoretical cognitive models with neural data.
• Current functional neuroimaging methods often regress cognitive model variables onto neural data, limiting mapping to single parameter values.
• A gap exists in directly fitting cognitive models to neuroimaging data and mapping parameters across the brain topographically.

Purpose of the Study:

• To introduce Computational Parametric Mapping (CPM) as a novel framework for cognitive computational modeling.
• To generalize the Bayesian population receptive field framework for broader application in cognitive neuroscience.
• To enable direct fitting of cognitive models to neuroimaging data and facilitate voxelwise or regionwise parameter mapping.

Main Methods:

• Developed Computational Parametric Mapping (CPM), a method building upon the Bayesian population receptive field framework.
• CPM allows direct fitting of cognitive models to neuroimaging data.
• Implemented voxelwise or regionwise mapping of cognitive model parameters to brain data, enhancing topographic analysis capabilities.

Main Results:

• CPM successfully fits cognitive models directly to neuroimaging data.
• Demonstrated the feasibility of voxelwise and regionwise mapping of cognitive model parameters onto brain regions.
• Showcased the efficiency of CPM for mapping over large regions of interest, illustrated with reinforcement learning models.

Conclusions:

• CPM offers significant advances for cognitive computational modeling by enabling direct model-to-data fitting.
• The method extends topographic mapping techniques from sensory sciences to cognitive neuroscience.
• CPM provides an efficient approach for detailed brain mapping of cognitive model parameters, applicable to various cognitive domains like reinforcement learning.