The TL;DR on Neuralink’s First Press Conference

In this post, I will try to cover the important information covered in the company’s first conference and research paper.

This was Neuralink’s first public meeting since it was established roughly 2 years ago. Up until now, little has been known about their progress in brain-machine interfaces (BMIs).

In the long term, it seems that Neuralink will function as a type of platform, think App Store, where more and more BMI-centered technology will be readily available.

In the short-term, they have made some impressive advances. If you haven’t already, I highly recommend you watch the Livestream or read the Abstract

Here are the key points:

First a recap of the major problems in current BMI technology:

1. More accuracy has required more invasive techniques.

There are BMIs capable of processing neuronal activity from outside of the brain, but they are not very accurate. And even the more accurate, invasive techniques thus far have been limited because they have only been able to take the average of thousands of action potentials and have not been able to record signals deeper inside the brain.

2. The materials used are not robust enough.

BMIs use microelectrodes as the gold standard. For example, here is an image of the ubiquitous Utah array:

Developed in the 1990s, this array has just 100 electrodes and is manufactured by Blackrock Microsystems. It was has been a critical device in neuroscience and clinical research, and is FDA-approved and has a best long-term case clocking in at 1000 days.

For long-term BMIs, the microelectrode arrays have been problematic because:

• They are composed of rigid metal or semi-conductors. The problem is that this type of alloy is too stiff for the different types of brain tissue, resulting in immune responses that limit their functionality and lifespan.
• Their size is also too big. This places constraints on which neurons they can access.

3. Full wireless circuitry is not possible.

The chips used for these devices are not operable for large scale wireless data transfer.

What Neuralink has made progress on regarding the above:

1.A specialized neurosurgical robot

This is a core part of Neuralink’s technology. As mentioned, while a more flexible material is beneficial for reaching different parts of the brain, it makes their implantation difficult.

The following 2 pictures show the robot and the insertion process on a gel-model of the brain. (A) is the needle head in both images.

The robotic electrode inserter; enlarged view of the inserter-head shown in the inset. A. Loaded needle pincher cartridge. B.Low-force contact brain position sensor. C. Light modules with multiple independent wavelengths. D. Needle motor. E. One of four cameras focused on the needle during insertion. F. Camera with wide angle view of surgical field. G. Stereoscopic cameras.

Needle pincher cartridge (NPC) compared with a penny for scale. A. Needle. B. Pincher. C. Cartridge.

Making use of computer vision, stereoscopic cameras, and software-based monocular extended depth of field calculations, the robot arm is incredibly accurate, allowing each electrode to be inserted individually, rather than in an array:

1. The inserter approaches the brain proxy with a thread. i. needle and cannula. ii. previously inserted thread. 2. Inserter touches down on the brain proxy surface. 3. Needle penetrates tissue proxy, advancing the thread to the desired depth. iii. inserting thread. 4. Inserter pulls away, leaving the thread behind in the tissue proxy. iv. inserted thread.

This allows for robust planning features and the ability to avoid vasculature during insertions, previously not possible.

2.A new type of microelectrode “thread”

The “threads” can target specific parts of the brain due to improvements in the polymer probe design and chemical construction.

Below are images of 2 types of these probes (A, B). They magnify individual electrodes more efficiently relative to their surface area (C), greatly improving the transmission rates (D)

Neuralink’s novel polymer probes. A. “Linear Edge” probes, with 32 electrode contacts spaced by 50 μm. B. “Tree” probes with 32 electrode contacts spaced by 75 μm. C. Increased magnification of individual electrodes for the thread design in panel A, emphasizing their small geometric surface area. D. Distribution of electrode impedances (measured at 1 kHz) for two surface treatments: PEDOT (n = 257) and IrOx (n = 588).

The current version uses over 1000 electrodes, which is an order of magnitude better than anything out there, but the human-ready version should have roughly 10,000. Each electrode produces approximately 200 Mbps of data.

3.Custom integrated circuit

Major advancements in the design of the chips have resulted in the Neuralink Application Specific Integrated Circuit (ASIC) It is capable of the following:

— 256 individually programmable amplifiers (analog pixels)

— On-chip analog-to-digital converters (ADCs)

This allows the data transfer to sample in at 19.3 kHz at a 10-bit resolution while an entire chip consumes ~6mW. Below is a photograph of the prototype:

A packaged sensor device. A. individual neural processing ASIC capable of processing 256 channels of data. This particular packaged device contains 12 of these chips for a total of 3,072 channels. B. Polymer threads on parylene-c substrate. C. Titanium enclosure (lid removed). D. Digital USB-C connector for power and data.

Though still in prototype stage, the wireless version will utilize an induction coil, where the chips attached to the end of the threads are powered by a combination of a battery/electronic device that (1) transmits power inductively and (2) computes the transmitted data.

Exciting new progress will be made, first by applying this to quadriplegics and those with crippling neurological disorders such as Parkinson’s and dementia, after which more general applications for synthesizing neuronal bandwidth with electronic devices can be made. We will see what happens in 2020 with FDA approval!