Reducing power requirements for high-accuracy decoding in iBCIs.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Intracortical brain-computer interfaces (iBCIs) typically use neural spikes for decoding, requiring high-sampling-rate data.
• Local field potentials (LFPs) offer an alternative, lower-bandwidth signal but have historically shown lower decoding performance than spikes.
• Existing LFP-based decoding methods have not matched the accuracy of spike-based decoding for real-time control.

Purpose of the Study:

• To develop and validate a novel strategy for enhancing LFP-based decoding performance in iBCIs.
• To reconstruct neural firing rates from LFPs using a neural dynamics model.
• To enable high-accuracy decoding from LFPs, approaching spike-based performance while reducing system requirements.

Main Methods:

• Trained neural dynamics models using LFPs to reconstruct underlying neural firing rates.
• Tested the LFP-based reconstruction and decoding strategy on macaque reaching tasks and human attempted speech data.
• Compared decoding performance of LFP-based dynamics models against direct spike decoding and LFP-alone decoding.

Main Results:

• LFP-based neural dynamics models achieved firing rate reconstruction accuracy comparable to spike-based models.
• Decoding performance using LFP-based dynamics models surpassed that of LFPs alone and neared spike-based model performance.
• In most applications, LFP-based dynamics models outperformed direct spike decoding in accuracy.

Conclusions:

• The proposed LFP-based dynamics models significantly improve decoding performance for iBCIs.
• This approach allows for high-accuracy neural decoding using lower bandwidth and sampling rates, reducing iBCI power requirements.
• Findings suggest a pathway towards more power-efficient and practical iBCI systems without compromising control accuracy.