Prior knowledge improves decoding of finger flexion from electrocorticographic signals

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) translate brain signals into user intent.
• Electrocorticographic (ECoG) recordings offer insights into movement kinematics.
• Existing BCI decoding methods often lack incorporation of prior movement knowledge.

Purpose of the Study:

• To develop and evaluate a Bayesian decoding method for decoding finger flexion from ECoG signals.
• To improve BCI performance by integrating prior knowledge about finger movement constraints.
• To enhance the potential for fine-grained neural control of prosthetic devices.

Main Methods:

• Utilized electrocorticographic (ECoG) signals from human subjects.
• Developed a Bayesian decoding approach incorporating prior knowledge via a switched non-parametric dynamic system (SNDS).
• Combined a prior model with a measurement model derived from linear regression for posterior estimation.

Main Results:

• The Bayesian decoding model demonstrated improved finger flexion decoding performance compared to linear regression.
• Incorporation of prior knowledge about finger flexion constraints enhanced decoding accuracy.
• The proposed method shows promise for more precise neural control.

Conclusions:

• Bayesian decoding with prior knowledge integration offers superior performance for ECoG-based BCI applications.
• This approach advances the development of sophisticated neurally controlled prostheses.
• The findings pave the way for improved functional outcomes in individuals with motor impairments.