8 min read

Brain-Computer Interface: Neuronal Signals and Arrow of Progress

Abstract: One of the biggest scientific discoveries yet to be made is understanding how the human brain works. Humans have built great things from aeroplanes to democracies, pyramids to bitcoin yet we don’t understand how the brain behind these inventions works. Brain-Computer interfaces could help us both better understand the brain, radically transform the way we interact and model cognitive functions as well as diseases to develop new therapies.

Introduction

Human innovation continues to thrive based on our capacity to create information and our ability to convey information. Communicating (conveying information) quickly, accurately and efficiently has played a big role in shaping modern society.

From fire and smoke signals in the prehistoric era to pigeon mails in the 5th century, and from telegraph in the 19th century to smartphones in the 21st century, the rate of exchanging information across long distances has continued to grow faster. And ever so faster since the 20th century or what Josh Wolfe calls ‘The Directional Arrow of Progress’ i.e communication technology is becoming more and more intimate with us:

Credits: Josh Wolfe

But there is also another angle of approach towards building BCI devices. The fear that innovations in AI is starting to outpace human capacity and the only way to cope up with machines would be to integrate with them. Elon Musk is the biggest champion of this. His goal at Neuralink is to build a neural implant that could sync the human brain with AI enabling us to control devices with our thoughts.

David Deutch also takes the same approach but in an optimistic way as stated in his essay:

“The worry that AGIs are uniquely dangerous because they could run on ever better hardware is a fallacy since human thought will be accelerated by the same technology. We have been using tech-assisted thought since the invention of writing and tallying. Much the same holds for the worry that AGIs might get so good, qualitatively, at thinking, that humans would be to them as insects are to humans. All thinking is a form of computation, and any computer whose repertoire includes a universal set of elementary operations can emulate the computations of any other. Hence human brains can think anything that AGIs can, subject only to limitations of speed or memory capacity, both of which can be equalized by technology.”

Technical Landscape

But what is a Brain-Computer Interface? BCI: A device that either translates neuronal information into

a. insights that help us better understand the brain by recording/stimulating specific brain areas through external stimulation modalities

b. commands capable of controlling external devices using sensors such as computers, smartphones or robotic arms

To better understand how BCI works, we need to dive a bit into how a neuron works:

The human brain is composed of billions of neurons. Each neuron is composed of various parts but we need to understand only 4 of them: axon, dendrites, soma and synapse

Credit: opentextbc.ca

A single nerve in the human brain is connected to roughly 10000 other nerves. Communication between 2 nerves generates a current, an electro-chemical process. A top-level communication works something like the following:

  1. an electrical signal is transmitted along an axon.
  2. The electrical signal at the end of the axon is converted into a chemical signal and the axon releases neurotransmitters, a bunch of chemical messengers
  3. The chemical messengers travel from the synapse to the dendrite of the next neuron where it is converted back again to electrical signals.

Just like an anode and a cathode of a battery cell, when a current leaves a nerve cell (neuron), it is considered to be in positive potential and negative when a current enters. The difference in potential aka voltage can be picked up, for example, by electrodes. While the signal from one neuron is extremely weak to pick up, the brain is constantly generating electric signals across thousands of neurons simultaneously. This makes it possible to have a measurable signal despite the fact that the skull and the skin of the human head are strong electrical insulators.

BCI can be broadly classified into Invasive and Non-invasive:

Non- Invasive

Advancements in Electroencephalography(EEG), an electrophysiological monitoring method to record electrical activity on the brain, have lead to an increase in the adoption of Non-invasive BCI devices. The working principle behind an EEG is as follows:

Electrical signals in the brain can be either brain rhythmic waves(BRW) or event-related potentials(ERP). While BRWs are specific repetitive patterns of neural activity used for medical monitoring, ERPs are triggered by events that can be measured as the brain response to external stimuli.

EEG data contains rhythmic activity, which reflects neural oscillations that occur at specific frequencies. Commercial EEG headsets typically measure the amount of brain activity that occurs in a certain frequency that can be used for monitoring or meditation purposes.

Credit: ResearchGate

An EEG can measure:

  1. Action potentials (a temporary spike  in the neuron's membrane potential caused by ions flowing in and out of the neuron) along the axons connecting neurons
  2. Currents through the synapses connecting axons with neurons/dendrites
  3. Currents along dendrites from synapses to the soma of neurons

There are other non-invasive ways to measure the electrical activity of the brain such as Magnetoencephalography(MEG), Positron Emission Tomography(PET) and functional Magnetic Resonance Imaging(fMRI). However, EEG is the most advanced and best suitable to build a viable non-invasive BCI.

Invasive:

Involves surgical implantation of the device into the skill. There are two major approaches in invasive BCI:

  • Electrocorticography(ECOG) where an electrode plate is in direct contact with the surface of the brain. Sometimes, semi-considered semi-invasive, it works similar to that of EEG but the electrodes are embedded in a thin pad that is placed right above the cortex. ECOG offers higher spatial resolution, better signal-to-noise ratio and wider frequency range than EEG. Almost all companies in this space such as Neuralink leverage ECOG for building invasive BCI.
Credit: Wikipedia
  • Intracortical Microelectrodes are implantable devices that can be used in either sensory prosthetics such as cochlear implants as stimulating interfaces or as recording electrodes. While stimulating interfaces are effective in producing sufficiently high signal strength, current implantable microelectrodes are limited in capacity to record single- or multi-unit activity. ( Single unit BCIs detect the signal from a single area of brain cells, while multiunit BCIs detect from multiple areas.)

Invasive vs Non-invasive:

There are pros and cons to both approaches. While surgical ones clearly have both ethical and safety concerns associated with them, the signal to noise ratio for non-invasive BCIs is comparatively low.

Invasive BCIs will largely be limited to medical use-cases for years to come. Non-invasive BCIs have high adoption potential and sufficient accuracy can be attained even for low-quality signals through the implementation of Conventional Neural Networks.

Components of BCI:

Credit: IOS Press
  1. Signal Acquisition: The first component is sensing and measuring the signals of the brain based on neuronal activity. Once this signal is acquired by electrodes on invasive/non-invasive devices, the signal is amplified and passed through an analogue-to-digital converter.
  2. Signal Pre-Processing: The raw signal is usually affected by noise(from the equipment) and other artefacts (eg. potential difference due to eyeball movement), particularly in EEG, and must be processed. Elements such as high-pass filters are used to clean the data.
  3. Signal Processing:
  • Feature Extraction is the first step where different characteristics such as user intent are decoded from the signal.
  • Classification: Once the necessary features are extracted, data is passed through classifiers running in parallel and fed into a convolutional neural network(CNN) to detect features in real-time.
  • Translation: The extracted information is then transferred to an algorithm that translates the information to instructions that can be executed by an external device.
  • Execution: A Feedback/External Device such as a computer or robotic arm receives the command to execute the task intended by the user.

Market Landscape

While approaches for neural interface can be broken down into invasive and non-invasive, applications can also be broadly classified between data analytics and user interface.

Non-Invasive:

Kernel is the biggest player in the non-invasive landscape. A full-stack neurotech company that combines cutting-edge hardware with Neuroscience as a Service software platform to enable neural data acquisition and analytics. The data could be used in discovering cognitive biomarkers, training in computer vision tasks, and more importantly, for better understanding the human brain. Largely self-funded by the founder Bryan Johnson himself, the company recently raised a 54M $ Series C funding.

Psyber Health is an ATAI Life Sciences company tackling mental health through a combination of BCI and psychedelics. By offering digitally enhanced personalized information about brain activity using EEG-enabled BCI devices that complements therapies, Psyber aims to improve psychotherapy.

Neurable is a US-based startup that has raised 9M $ in funding with a mission to translate brain activity into simple, actionable insights. The company recently launched an IndieGoGo campaign for customers to pre-order BCI-enabled headphones that help the customer improve focus based on insights using brainwave sensors embedded in the headphones.

Ctrl-Labs, acquired by Facebook (ethical *cough* concerns?), is a neural interface company that seems like the company furthest in the market in bringing a new UX interface that changes the way we interact with machines. (Remember action potential, the little spike?) While the device inherently leverages action potential from the neuron, the signal is detected at the muscles once the current proliferates from the neurons to the muscle fibres. This is due to the reason that muscle fibres have higher amounts of electrical activity as compared to neurons.

NextMind is a French startup focusing on the gaming market that uses EEG-based to command machines by using thoughts. They launched a Software Development Kit (SDK) for virtual and augmented reality applications.

Invasive:

Neuralink is the most widely known BCI startup, thanks to the brain behind it: Elon Musk. The company has raised 154M $ in funding so far to develop ultra-high bandwidth brain-machine interfaces to connect humans and computers. They leverage ECOG technology by placing electrodes near neurons to detect action potential and decode neuronal signals. The company aims to build a 2-way communication channel and believes invasive is the best approach to both stimulate and record signals. The company initially aims to start with medical applications in treating diseases such as Parkinson’s and Alzheimer’s.

Paradromics, a US startup with 29M $ in funding that included 4.1M $ in equity crowdfunding, is a BCI company developing low-power, high data rate neural sensors to enable massively parallel neural recordings for therapeutic applications. The official vision statement of the company is to increase the data transmission rate between brains and machines.

Ceregate is a German BCI startup focused on treating Parkinson’s disease by focusing mostly on machine learning based approached to personalize the device that prevents tremors and other related issues.

InBrain Neuroelectronics is a Spanish startup with 15.5M $ in funding developing the next generation of neuroelectronic therapies powered by graphene-based devices and machine learning. The company claims graphene’s properties(thinnest material) perfectly adapts stimulation to targeted brain anatomy leaving no side effects while having the capability of reading signals at a resolution never seen before.

BIOS Health is a UK-based full-stack neural interface platform, that leverages AI to decode and encode the signals from the body, to treat chronic health conditions such as Parkinson’s. The company, having raised 5.8M $ in funding, will initially focus on software to better understand neural signals and eventually build implantable devices to replace deep brain stimulators and in certain cases replace pill-popping.

While the industry is certainly booming and has lots of potential in treating cognitive diseases, there are certainly ethical and safety concerns associated with BCIs.

Resources and Recommendations:

NeuroTechEDU provides a detailed explanation of BCI and an awesome list of BCI-related Resources related to the current developments and associated tools.

Comments powered by Talkyard.

Comments

Become a Infiniti Ventures member below to join the conversation (it's free!). As a member, you will also receive new posts by email (you can unsubscribe at any time).