A Bidirectional Brain-Machine Interface Featuring a Neuromorphic Hardware Decoder.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Neuromorphic Computing

Background:

• Bidirectional brain-machine interfaces (BMIs) enable two-way communication between the brain and external devices, crucial for patients with motor and sensory deficits.
• Current BMI systems are often bulky external devices, limiting their clinical applicability.
• There is a need for compact, low-power, fully implantable BMIs for effective patient support.

Purpose of the Study:

• To develop a modular, compact, and low-power bidirectional BMI system.
• To implement a neuromorphic processor as a decoder for neural signals.
• To demonstrate the feasibility of using neuromorphic technology for adaptive BMI modules.

Main Methods:

• Developed a modular bidirectional BMI setup featuring a compact neuromorphic processor.
• Implemented a spiking neural network on the neuromorphic chip using ultra-low-power mixed-signal circuits.
• Utilized on-chip spike-timing-dependent plasticity for online learning and decoding of neural signals in an anesthetized rat model.

Main Results:

• The neuromorphic chip successfully learned to decode neural signals, controlling external device movements.
• The modular BMI design allowed for independent tuning of components.
• The system demonstrated robust control of object trajectories, interfacing the rat's brain bidirectionally with an external object.

Conclusions:

• Neuromorphic technology is sufficiently advanced for creating compact, low-power, and computationally powerful BMI modules.
• The developed modular BMI system represents a significant step towards fully implantable bidirectional BMI devices.
• This approach holds promise for enhancing clinical support for individuals with severe neurological impairments.