Compressing Ultra-Dense Neural Recordings in Space and Time: A 3-D Modeling Approach

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Wireless implantable brain-computer interfaces (iBCIs) require more channels for better neural signal acquisition.
• Increasing channels in iBCIs escalates power consumption and data rates, straining device resources.
• Efficiently compressing neural data while preserving critical information is a key challenge for high-channel iBCIs.

Purpose of the Study:

• To propose a 3D spatiotemporal neural signal compression method for ultra-high-channel wireless iBCIs.
• To leverage spatiotemporal correlations in neural signals for enhanced compression efficiency.
• To maintain the integrity of action potentials (APs) and local field potentials (LFPs) during compression.

Main Methods:

• Developed a 3D spatiotemporal matrix representation of neural signals.
• Implemented three-dimensional compressed sensing (3D-CS) by synchronizing APs and LFPs.
• Validated the method on a 1024-channel wireless neural recording system.

Main Results:

• Achieved a total compression ratio (CR) of 156.
• Preserved all action potentials (APs) intact.
• Maintained a structural similarity index measure (SSIM) > 0.95 for local field potentials (LFPs).
• Reached an average signal-to-noise and distortion ratio (SDNR) of approximately 24.16 dB.

Conclusions:

• The proposed 3D spatiotemporal compression method offers an efficient, low-power solution for ultra-high-channel wireless iBCIs.
• This approach effectively exploits neural signal correlations for data compression.
• The method ensures the integrity of neural signals, crucial for reliable brain-computer interface operation.