A Bayesian Decoder Representing Single-Directional Connectivity between Neurons in Brain-Machine Interface.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Understanding neural information transmission requires modeling directional connectivity between neurons.
• Current brain-machine interface (BMI) decoders often use simplified models that overlook synaptic directionality.
• Accurate neural encoding models are crucial for effective BMI performance.

Purpose of the Study:

• To develop a Bayesian decoder that incorporates directional neural connectivity for improved BMI performance.
• To enhance the precision of neural encoding models by representing functional neural circuits.
• To enable the evaluation of causality between neurons at the behavioral level.

Main Methods:

• Derived a chain-likelihood using Bayes' rule to model single-directional influence between neurons.
• Integrated directional neural connectivity into a Bayesian decoder framework for BMIs.
• Validated the method using synthetic data from a rat's two-lever discrimination task.

Main Results:

• The proposed method accurately represents directional neural connectivity, outperforming existing approaches.
• The new decoder model is more computationally efficient due to fewer parameters.
• Demonstrated the potential to evaluate neural causality at the behavioral level.

Conclusions:

• The developed Bayesian decoder effectively models directional neural connectivity, enhancing BMI performance.
• This approach offers a more precise representation of neural circuits and their function.
• The method provides a novel way to assess neural causality in relation to behavior.