Neuralink, a neurotechnology company founded by Elon Musk, has developed a novel brain-machine interface system with unprecedented electrode density and bandwidth.
In a recent white paper, Neuralink scientists describe their progress towards a scalable, high-performance implantable device for interfacing with the brain.
The key facts:
- Neuralink’s device has over 3,000 electrodes per array, an order of magnitude more than previous clinical devices. This enables recording from thousands of neurons simultaneously.
- They use flexible polymer threads instead of rigid metal electrodes. The threads cause less damage to brain tissue and can access neurons in deeper brain regions.
- A robotic neurosurgical system precisely inserts the threads quickly avoiding blood vessels. This automation enables scaling up to high electrode counts.
- Miniaturized custom chips amplify and digitize signals from each electrode, enabling simultaneous streaming of data from thousands of channels.
- In rats, they have achieved up to 70% yield of spiking activity on electrodes and demonstrated simultaneous broadband recordings from over 3,000 inserted electrodes.
Source: J Med Internet Res.
This technology could enable high-fidelity brain-machine interfaces for restoring sensory, motor and cognitive function in neurology patients.
Neuralink aims to first create a research platform in rats to refine the technology, then ultimately create a wireless fully-implantable clinical human device.
Flexible Polymer Electrodes Minimize Brain Tissue Damage
A major challenge for brain-machine interfaces is biocompatibility – being able to record from neurons without damaging tissue over long periods.
Neuralink’s innovation is using flexible polymer-based electrodes rather than rigid metal electrodes.
Previous clinical brain implants have used arrays of rigid metal microwires.
However, stiffness mismatches between these electrodes and brain tissue can trigger neuroinflammation, leading to recording degradation over time.
Instead, Neuralink fabricates micron-scale electrodes on thin flexible polymer threads.
With widths from 5-50 μm, these threads are orders of magnitude narrower than metal electrode arrays.
This mimics the mechanical properties of the brain better.
The thin threads have up to 32 gold recording sites, allowing each one to interface with dozens of neurons.
By using a robotic insertion technique, they can implant up to 96 threads per minute.
This combination maximizes channel count while minimizing displaced brain volume.
Neuralink currently offers probes with up to 3,072 electrodes per array based on this approach.
Robotic Neurosurgery Enables Rapid, Precise Electrode Insertion
Delicately inserting flexible polymer threads into the brain is challenging. Neuralink built a neurosurgical robot to automate this process.
The robot uses computer vision and AI to precisely insert each thread along an ideal trajectory.
This avoids penetrating blood vessels on the brain’s surface, minimizing vascular damage.
A needle gripper picks up each thread and guides it into position.
The needle then inserts the thread by penetrating the brain tissue.
When the needle retracts, the thread stays embedded in the brain.
This allows rapid, massively parallel insertion of 96 threads per minute.
Each thread can be independently targeted based on pre-surgical planning.
Automating insertion not only speeds up surgeries, but improves precision and consistency.
This will be key for safely and effectively scaling up electrode counts even further.
Miniaturized Electronics Provide Full-Bandwidth Neural Data Streaming
Recording useful signals from thousands of electrodes poses serious data bandwidth challenges.
Neuralink designed custom chips to amplify and digitize signals right next to each electrode.
Their application-specific integrated circuit contains amplification and digitization circuits for 256 electrodes.
This limits wiring to just a USB-C cable per device.
The chips filter and amplify neural signals, discriminate spiking activity, and transmit full-bandwidth data simultaneously from all channels.
For 3068 electrode systems, 12 of these chips are integrated using advanced flip-chip manufacturing.
The whole system fits in a compact 23×18.5x2mm package.
This tight integration of custom electronics and electrodes in a single device enables scaling up channel counts dramatically compared to conventional wired electrodes.
Promising Electrophysiology Results in Rodents
Neuralink has tested their brain-machine interface systems by implanting them in rats.
The flexible electrodes integrate well and record spiking activity.
In one experiment with 1536 electrodes, they achieved 43% yield of spikes on electrodes.
On average across 19 surgeries, 45% of electrodes recorded distinguishable spikes.
They have demonstrated over 3000 simultaneously recorded channels in free-moving rats.
The signals have excellent signal-to-noise ratios.
This proves the technology can scale far beyond previous clinical intracortical brain-machine interfaces.
Recording more neurons should enable finer motor control and more intuitive prosthetics.
While further development is needed before human use, Neuralink has validated core technologies for a practical, high-performance clinical brain-computer interface.
Moving Towards Wireless Human Implants: Brain Connects to Computer?
Neuralink aims to first create a research platform for rapidly iterating on brain-machine interface designs in animals.
However, the long-term goal is developing clinical human devices.
While current implementations use wired connections, future versions will be fully implantable.
This requires integrating wireless power and telemetry into the devices.
Human versions of the device would be implanted through a single surgical procedure.
After healing, they could provide seamless high-fidelity connections with external computing systems.
This could enable paralysis patients to control computer cursors and mobile devices using just their thoughts.
When combined with emerging spinal stimulation techniques, it may even restore motor function.
Neuralink’s technology provides a promising foundation for creating practical, scalable clinical brain-machine interfaces.
There are still challenges to address before human use but the capabilities could be transformative.
Brain-machine interfaces have long seemed like science fiction but rapid recent progress makes clinical applications a realistic possibility.
Devices like Neuralink’s have potential to help patients by seamlessly integrating computing systems with the brain’s intrinsic neural circuitry.
This could restore sensory, motor and cognitive functions lost due to injury and neurological disorders.
Neuralink’s white paper demonstrates key technical strides towards making this a practical reality.
References
- Study: An integrated brain-machine interface platform with thousands of channels
- Authors: Elon Musk & Neuralink (2019)
