News

Behind Elon Musk’s Neuralink, the Company Developing Brain-Machine Interfaces

July 20, 2020 by Adrian Gibbons

What's going on at the silicon level of Elon Musk's human brain-machine interface?

In the past, the notion of human brain-machine interfaces (BMI) was confined to science fiction, depicted most recently in film and television shows like Anon and Altered Carbon. These two productions and others like them foretell of an age where human interactivity with machines will be seamless, providing enhanced data capabilities and beyond.

The technology to make such BMIs a reality may not be as far-fetched as you might think. Neuralink, a company co-founded by Elon Musk, is developing next-generation brain-machine interfaces with scalable neural channel density and real-time data processing.

 

Evolution of Neuralink prototypes

Evolution of Neuralink prototypes. Screenshot used courtesy of Neuralink
 

Some of the stated use cases for this technology include a full articulation of the digits in a prosthetic hand via a wireless Neuralink implant or the ability to see with electrodes replacing the optic nerve. 

What's going on at the circuit level of such complex biomedical devices?

 

Neuralink’s N1 System on a Chip

Featured in the Neuralink Launch Event in 2019, the N1 is a hermetically-sealed SoC featuring 1,024 channels in a 4 mm x 5 mm package, processing 200 Mbps of data per channel using only 6.6 microwatts of power.

It is capable of compressing neurological data up to 200 times making it suitable for Bluetooth transmission. The chipset is embedded in the skull and connected to ultra-thin filaments called "threads," containing 32 gold electrodes per thread. These threads are embedded in the flesh of the brain, with electrodes less than 60 microns from their target neuron by a custom-built, surgeon-assisted robot. 

 

Basic structure of Neuralink chips, though designs vary significantly in complexity from chip to chip

Basic structure of Neuralink chips, though designs vary significantly in complexity from chip to chip. Screenshot used courtesy of Neuralink

The N1 is subdermally wired to an inductive loop, which is surgically embedded in the skin behind the ear. A wearable device called the "Link" provides power and Bluetooth connectivity to the SoC. It receives signal-conditioned neuron impulses called "action potentials" as well as direct stimulation of neurons.

There are three key challenges with this implanted technology from an electronics perspective: 1) analog processing of neuron spikes (action potentials), 2) automatic spike detection, and 3) every channel stimulation (generation of action potentials).

 

A Neuron's Journey from Analog to Digital

The first problem with a neuron spike is the amplitude, typically less than 10 microvolts. Amplifying signals that are only 10 microvolts requires a system gain of 43 dB to 60 dB in order to place the signals within the 10-bit resolution of the onboard ADC (~ 1 mV).

Impedance concerns compound the narrow ADC resolution because decreasing electrode geometries mean greater resistivity and noise in the system. In the company's research journal published in August 2019, Neuralink investigated two surface treatments (polystyrene sulfonate and iridium oxide), which have promising impedance characteristics (as seen below).

 

A single thread of Polyimide dielectric with gold filaments & electrodes (left) with associated impedance profiles (right)

A single thread of Polyimide dielectric with gold filaments & electrodes (left) with associated impedance profiles (right). Image used courtesy of Neuralink

 

Spike detection is the second major hurdle Neuralink has to overcome since certain spikes are critical to specific BMI tasks. The N1 has on-chip spike detection technology to directly characterize the shape prior to ADC conversion, providing the designers with more information about different neurons based upon their shapes. 

 

Neuralink characterizes the N1 chip by three key characteristics

Neuralink characterizes the N1 chip by three key characteristics: analog pixels, on-chip spike detection, and stimulation on every channel. Screenshot used courtesy of Neuralink

 

The final challenge with the N1 SoC was providing electrode stimulation, not just reception. The N1 can stimulate any of its 1,024 electrodes in groups of 64—simultaneously. The ability to stimulate a greater number of neurons allows for more complex tasks to be performed.

 

FDA Approval, Clinical Trials, and Charge Limitations

Neuralink, as of this writing, is still undergoing the lengthy process of FDA approvals. This process includes extensive classifications for neurology, toxicology, and immunology, among others. The company hopes to begin clinical trials no later than 2021. 

The "Link," which provides inductively coupled power and connectivity to the threads and SoC, is a limitation to the overall system because of its charge cycle, according to Neuralink researchers. Neuralink developers hope that in the future, a person can regain complete motor control function without being limited by the device’s charge cycle. 

 

Direct Brain Stimulation

According to Musk, BMIs are an interface that intersects neurology and technology—and the potential for treatment is endless. In a May 2020 interview with Joe Rogen, Musk spoke of the future capabilities of the Neuralink implant: “In principle, it [a Neuralink device] could fix anything that’s wrong with the brain.”

 

An early generation prototype featuring 3,072 channels and a USB-C connector for real-time data logging of laboratory tests on Long-Evans rats

An early generation prototype featuring 3,072 channels and a USB-C connector for real-time data logging of laboratory tests on Long-Evans rats. Image used courtesy of Neuralink
 

This includes repairing the optic nerve, providing for improved or complete restoration of motor control, or the ability to synthesize speech. Musk makes it very clear that there is still a great deal of research to be completed before this technology is adopted. 

 

Featured image used courtesy of Neuralink