Categories and Functional Units: An Infinite Hierarchical Model for Brain Activations

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Understanding brain responses to stimuli is crucial for neuroscience.
• Functional magnetic resonance imaging (fMRI) provides insights into brain activity.
• Existing models may not fully capture inter-subject variability in brain responses.

Purpose of the Study:

• To develop a novel model for describing brain response structures.
• To identify stimulus categories and functional brain units from fMRI data.
• To account for inter-subject variability in neural responses.

Main Methods:

• A nonparametric hierarchical model was designed to analyze fMRI data.
• The model links brain activations to fMRI time courses.
• A variational inference algorithm was used to learn categories, units, and activation probabilities.

Main Results:

• The model successfully identified meaningful clusters of stimuli into categories.
• Consistent clusterings of voxels into functional units were found.
• The method demonstrated effectiveness in an fMRI object recognition study.

Conclusions:

• The proposed model effectively describes brain response structures.
• It provides a robust method for analyzing fMRI data by identifying stimulus categories and functional units.
• The approach accounts for individual differences in brain organization.