PyMVPD

PyMVPD models multivariate interactions between brain regions using functional magnetic resonance imaging (fMRI) data to characterize multivariate dependencies in neural activity.

Key Features:

• Multivariate Pattern Dependence (MVPD): Implements MVPD to model multivariate dependencies between brain regions using linear regression and artificial neural network models and supports training and testing with independent datasets.
• Pre-implemented models: Includes pre-implemented linear regression and artificial neural network models for application to fMRI data analysis.
• Customizable framework: Supports customization of model architectures to adapt analyses to specific research questions.
• Region-specific model performance: Evaluates model performance across brain regions and reports that artificial neural networks often outperform linear regression while optimal architecture is region-dependent.

Scientific Applications:

• Cognitive neuroscience: Investigating dynamic interplay between cortical regions during cognitive tasks using multivariate models of functional connectivity.
• Seed-region analyses: Exploring neural bases of cognition with seed regions such as the fusiform face area (FFA) and the parahippocampal place area (PPA).
• Model comparison: Comparing linear versus nonlinear multivariate models to assess differences in captured dependencies among brain regions.

Methodology:

Uses a data-driven MVPD approach employing linear regression and artificial neural network models, with training and testing of multivariate models on independent datasets to capture multivariate dependence in fMRI data.