Scikit-NeuroMSI: A Generalized Framework for Modeling Multisensory Integration.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Multisensory integration is vital for cognition, with various computational models proposed.
• A unified framework for examining these models across contexts is lacking.
• This gap hinders systematic evaluation and comparison of multisensory integration theories.

Purpose of the Study:

• To introduce Scikit-NeuroMSI, an open-source Python framework for implementing and evaluating computational models of multisensory integration.
• To provide a tool for researchers to systematically explore model behavior and generate computationally informed hypotheses.
• To facilitate a comparative analysis of different modeling approaches, such as Bayesian and network models.

Main Methods:

• Development of Scikit-NeuroMSI, a Python-based open-source software framework.
• Implementation of diverse computational models of multisensory integration within the framework.
• Utilizing parameter sweeps in simulations for systematic exploration of model behavior in spatiotemporal causal inference tasks.
• Conducting a comparative analysis of Bayesian and network models.

Main Results:

• Scikit-NeuroMSI successfully implemented multiple multisensory integration models at various analytical levels.
• The framework enabled systematic exploration of model behavior in simulated spatiotemporal causal inference tasks.
• A comparative analysis identified commonalities between Bayesian and network models, bridging different descriptive levels.

Conclusions:

• Scikit-NeuroMSI offers an accessible platform for advancing computational research in multisensory integration.
• The framework facilitates the generation of computationally informed hypotheses for future studies.
• The comparative analysis highlights potential avenues for unifying different theoretical approaches in the field.