Neural spike compression through salient sample extraction and curve fitting dedicated to high-density brain implants

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Increasing channel density in brain implants necessitates efficient on-implant signal processing.
• Constraints in power, area, and data transmission are critical challenges for current neural recording microsystems.

Purpose of the Study:

• To develop a data reduction framework for extracellular neuronal action potentials.
• To enable efficient on-implant processing for high-density brain implants.

Main Methods:

• Proposed a framework that selects salient spike samples for interpolation.
• Developed a method to transmit attributes of salient samples for external reconstruction.
• Implemented a 128-channel neural signal compressor using 130-nm CMOS technology.

Main Results:

• Achieved a high data compression capability with hardware efficiency.
• The 128-channel compressor occupied 1.05 × 0.35 mm².
• Demonstrated an average neural data compression rate of ~2176 at 8 spikes/s.
• The compressor consumed 0.164 µW/channel at 1V and 32MHz.

Conclusions:

• The proposed framework substantially reduces data representing spike waveforms.
• This technique is suitable for high-channel-count brain-implantable neural recording microsystems.
• Enables next-generation high-density neural implants to effectively transmit acquired neuronal activity.