Increasing Robustness of Intracortical Brain-Computer Interfaces for Recording Condition Changes via Data...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Intracortical brain-computer interfaces (iBCIs) aim to restore motor function in paralyzed individuals by decoding neural activity.
• Shifting neural-to-kinematic mappings due to changing recording conditions degrade iBCI performance.
• Conventional methods require extensive training data or frequent recalibration, leading to user fatigue and computational load.

Purpose of the Study:

• To develop a novel approach for enhancing the robustness of iBCIs against changing recording conditions.
• To reduce the need for extensive training data and frequent recalibration in iBCI systems.

Main Methods:

• Utilized three neural augmentation operators to generate synthetic neural activity mimicking various recording conditions.
• Employed contrastive learning to learn latent factors from augmented neural data, maximizing similarity between augmented activities.
• Aimed to ensure learned factors remain stable and maintain consistent correlation with intended movement despite recording variations.

Main Results:

• The proposed iBCI demonstrated superior performance compared to state-of-the-art methods.
• Achieved robustness against changing recording conditions over multiple days in long-term use.
• Showcased satisfactory offline decoding performance even with limited training data.

Conclusions:

• The study presents a method to decrease the necessity for frequent iBCI calibration and large annotated training datasets.
• Future work includes improving offline decoding with minimal training data and enhancing robustness against electrode failure.