Unveiling Dynamic System Strategies for Multisensory Processing: From Neuronal Fixed-Criterion Integration to...

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Sensory Processing

Background:

• Multisensory processing is crucial for survival, enabling the brain to infer self-motion and object movement.
• The mechanisms by which multisensory brain regions process information and adhere to Bayesian strategies remain debated.
• The dorsal medial superior temporal area (MST-d) is implicated in multisensory integration.

Purpose of the Study:

• To investigate the neuronal circuit mechanisms underlying multisensory processing in the MST-d.
• To determine if MST-d neurons follow a Bayesian strategy in integrating or separating visual and vestibular information.
• To model multisensory processing using synaptically coupled multilayer continuous attractor neural networks (CANNs).

Main Methods:

• Combined physiological recordings from macaque MST-d neurons.
• Developed and analyzed synaptically coupled multilayer continuous attractor neural networks (CANNs).
• Investigated the role of synaptic coupling strength in neuronal responses and causal inference.

Main Results:

• Synaptic coupling in MST-d circuits induces cooperation and competition, enabling neurons to switch between sensory integration and separation modes.
• Neuronal switching is governed by a fixed-criterion causal strategy dependent on synaptic coupling strength.
• Population-level pooling of criteria represents sensory reliability priors, achieving Bayesian strategies in downstream neurons.

Conclusions:

• Synaptic input balance dynamically shapes neuronal direction preference, explaining observed misalignments between preference and inference.
• The study provides a computational framework for brain-inspired algorithms in multisensory computation.
• This work elucidates how MST-d circuits flexibly adapt to integrate or separate sensory inputs for robust self-motion perception.