A Fully-Integrated 1µW/Channel Dual-Mode Neural Data Acquisition System for Implantable Brain-Machine Interfaces

Area of Science:

• Biomedical Engineering
• Neuroscience
• Electrical Engineering

Background:

• Implantable brain-machine interfaces (BMIs) require efficient neural data acquisition systems.
• High channel count and low power consumption are critical for advanced BMIs.
• Existing systems often face limitations in power efficiency and channel density.

Purpose of the Study:

• To present an ultra-low power mixed-signal neural data acquisition (MSN-DAQ) system.
• To enable a novel low-power hybrid-domain neural decoding architecture for high-channel-count implantable BMIs.
• To demonstrate significant back-end power savings compared to prior art.

Main Methods:

• Designed and implemented a 32-channel custom chip in 180nm CMOS technology.
• Operated the system at a 1V supply voltage.
• Evaluated the fabricated prototype with in vivo human tests.

Main Results:

• Achieved ultra-low power consumption of 1.07µW/channel.
• Demonstrated excellent performance metrics: 2.37/5.62 NEF/PEF and 88dB common-mode rejection ratio (CMRR).
• In vivo human tests showed performance comparable to commercial recording systems.

Conclusions:

• The developed MSN-DAQ system offers a significant advancement in low-power neural data acquisition for BMIs.
• The hybrid-domain neural decoding architecture provides substantial power-saving advantages.
• The system's performance in human trials validates its potential for clinical applications.