The Future is Wild

An introduction to Brain-Computer Interfaces

Manroop Kalsi
11 min readOct 16, 2021
Originally posted on Dreamstime

Humans have always been sociable creatures. We thrive in environments where we’re able to collaborate, and the only way to collaborate is to communicate.

A study shows that humans spend approximately 75% of their waking hours in some form of communication: 9% writing, 16% reading, 30% speaking, and 45% listening. The majority of these communications occur through the use of a digital device. Every single day we spend hours formulating texts, emails, and on calls.

But, what if there was a way to directly teleport our thoughts to another person’s brain?

In 2018, an experiment done by Rajesh Rao was conducted where 3 individuals collaborated to play Tetris using brain to brain interfaces. 2 players (who were able to see the full board) were able to send their recommendation of whether to rotate the block or not to a third player (who was only able to see the block and not the state of the play), who would then initiate the action despite not having the full information about the block.

Originally posted on Newscientist

“Our experiment can be regarded as the first proof-of-concept demonstration that multiple human brains can consciously work together to solve a task that none of the brains individually could,”

~Rajesh Rao at the University of Washington in Seattle

This experiment was conducted with the use of Electroencephalography caps (EEG) — used to read + process signals from the senders — and Transcranial Magnetic Stimulation (TMS) -used to deliver information to the brain- which will dive deeper into later.

Brain to brain communication, controlling devices with your brain, being able to read someone’s mind.

All these things that we’ve seen for so many years in sci-fi movies now have the possibility of becoming a reality with brain-computer interfaces.

What is a Brain-Computer Interface? — BCIs as a Rock Concert

*Originally BCI Guys’ symphony analogy

A brain-computer interface is a device that is able to create a direct communication pathway between a brain and an external device (this can be one way, or more recently worked on both ways).

There are 3 main types of BCIs: non-invasive, semi-invasive, and invasive — which will be explained in the context of going to see a concert.

*spatial resolution: capacity to tell you exactly which specific area of the brain is active

*temporal resolution: how closely the measured activity corresponds to the timing of the actual neuronal activity

Non-Invasive BCIs

A non-invasive BCI is like you’re standing in the lobby of a rock concert with the doors shut. The sound is a little muffled, but you can hear the major movements in the concert, like when the music gets louder/ softer or faster/ slower. You don’t know exactly what songs are being played, but you can understand the overall genre and vibe, and hear the music from different areas to get different perspectives, but you still can’t get inside the room.

A non-invasive BCI sits on the outside of your skull, on top of the skin, because of all the layers in between the neurons in your brain and the actual device the signals are muffled.

Common non-invasive BCIs:

  • Electroencephalography or EEGs are the most common types of BCIs. These sensors are placed on the scalp to measure electrical potentials produced by the brain. EEGs allow for real-time data, excellent temporal resolution (milliseconds), yet, have a significantly low spatial resolution (6–9 cm²)
  • Magnetoencephalography or MEGs measure the magnetic fields produced by your brain’s electrical current, its temporal resolution is in milliseconds, whereas its spatial resolution is currently approaching a few mm².
  • Magnetic resonance imaging or MRI uses a magnetic field and computer-generated radio waves to create detailed 3-dimensional images of the body. (Temporal Resolution = 20–50 millisecs)
  • Functional Magnetic Resonance Imaging or fMRI measures brain activity by detecting changes associated with blood flow. (Temporal Resolution = 20–50 millisecs)
  • Positron Emission Tomography or PET is used to observe deferent processes such as blood, flow, metabolism, and neurotransmitters in the body (Temporal Resolution = 10–60 secs, Spatial Resolution = 4- 10 mm).
  • Functional Near-Infrared Spectroscopy or fNIRS measures brain activity through hemodynamic responses associated with neuron behavior. fNIRS combines the portability of EEG + spatial resolution of other neurotech like fMRI making it portable, cheaper, and a lot less susceptible to electrical noise than EEG.

Semi-invasive BCIs

A semi-invasive BCI is like you are a member of the audience, you can hear and engage with the music clearly, and may even be able to discern between sections, but not of any individual performers. You may be able to influence the music if you get enough people to join in.

A semi-invasive BCI is placed underneath the skull but doesn’t penetrate the brain.

The most common type of semi-invasive BCI is an EcOg, which is similar to an EEG but placed on the brain surface. It can be useful for longer-term monitoring or for brain control where higher accuracy and more permanency are needed.

Invasive BCIs

An invasive BCI is like you are a drummer in the rock bank, you can hear overall but the sound is dwarfed by your own drum. You are able to exert greater influence over a section by changing the notes to go with or against the orchestra.

Invasive BCIs are able to use micro-electrodes that are placed directly into the cortex and are able to measure the activity of single neurons. These electrodes allow for specific targeting but may only be useful for a few years before brain scarring starts to degrade the signal quality.

Each type of BCI has its own advantages and disadvantages. In order to significantly exponentiate the possibilities of BCIs, a higher spatial resolution is needed. Invasive BCIs have the highest resolution yet, are the most difficult to scale resulting from its risk. Non-invasive BCIs although scalable, and has a higher possibility of exception for mass adoption, have lower spatial resolution decreasing its effectiveness.

The Brain + Neuron

The nervous system is made of 2 main components: the central and peripheral nervous system. The central nervous system includes the brain and spinal cord whereas, the peripheral system includes the nerves that branch out. The main organ of the central nervous system is the brain.

There are 3 main divisions of the brain:

  • Brain Stem — connects the brain to the rest of the body
  • Cerebellum (Latin for little brain) — holds about half of all neurons, and is most involved in coordinating voluntary motion (e.g. posture, balance, speech, etc.)
  • Cerebrum — largest, responsible for most cognitive processing including sensory processing, generation of motor commands, emotional regulation, etc.

With the wide breadth of functions in the cerebrum, it has the most applicability when it comes to BCI technology.

The surface of the cerebrum has an abundance of hills (gyrus), and valleys (sulcus), the deep sulci are what splits the cerebrum into 4 distinct lobes in each hemisphere:

  • Frontal lobe — which includes the pre-frontal cortex (mainly complex cognitive functions), motor areas (executing motor actions), and Broca's area (associated with speech production)
  • Temporal lobe — which houses the limbic system (e.g. motivation, emotion, learning, etc.), primary auditory cortex (auditory processing, and Wernicke's area (helps understand language)
  • Parietal lobe — includes the somatosensory cortex (responsible for processing sensory information, such as touch and pain — sensory stimuli)
  • Occipital lobe — primarily responsible for visual processing

The brain itself has approximately 86 million neurons (the building blocks of the brain), and billions of glial cells (caretakers — support the neurons within the brain to keep them healthy, happy, and functioning).

Neurons share several characteristics with other types of cells in the body. The neuron contains several main components:

  • Cell Body aka Soma (houses important structural parts of the neuron-like the nucleus) — responsible for chemical processes that provide energy to the neuron from the dendrites, and where signal processing and firing decisions are made
  • Dendrites — a structure that extends from the neuron’s body and receives information from neighboring neurons usually in the form of molecules called neurotransmitters
  • Axon — structure that again extends away from neuron’s body — carries electrochemical signals from the neuron’s body and communicates it to other neurons by terminating on or close to other neurons’ dendrites
  • Synapses — allows signals to be transmitted between two neurons, or between neuron and muscle cells

The process of neurons communicating with each other is referred to as their action potential which is the term used to describe when neurons fire. These action potentials are mediated by ion flow, they trigger a chain reaction (which cannot be stopped in a healthy neuron) where eventually the signal reaches the axon terminal where it triggers the release of neurotransmitters into the space between the 2 connected neurons (the synaptic cleft), which then are received by the next neuron’s dendrites. If enough neurotransmitters are received by the next neuron’s dendrites, the process starts again.

When using EEGs, there are 3 main types of information that are collected:

  1. Action potentials along the axons connecting neurons
  2. Currents through the synaptic clefts that connect with axons
  3. Currents along dendrites from the synapses to soma

All this information above comes together to create effective BCI technology.

Components to a BCI using an EEG

When creating a BCI project using an EEG, there are 4 main components: Collect, Preprocessing, Feature Extraction, and Deployment.


EEG is able to measure and record the voltage differences between 2 electrodes, or more. There are an abundance of different types of electrodes available to use, including wet electrodes, dry electrodes, caps, needle electrodes, etc.

The typical configuration for EEG is the 10–20 system:

As a result that the distance and the sheer amount of layers that the signals need to travel through, they can become quite muffled. Therefore, after collecting signals from electrodes, they are sent to an amplifier that is able to amplify them. From there, an A/D converter which converts the amplified signal from an analog form to a digital form.

When collecting brain waves there are a variety of different frequencies of signals that may be received that indicate different things.

  • Gamma Waves— high degree of concentration
  • Beta Waves — normal cognitive processing
  • Alpha Waves — calm wakefulness
  • Theta Waves — sleep or deep meditative state
  • Delta Waves — high degree of synchronicity indicative of sleep


One of the main limitations with BCIs, especially with non-invasive interfaces is that the signals are prone to interference. There are 4 main sources of this noise: the EEG equipment itself, electrical interference that is external to the subject + recording system, the leads, and electrodes, and the subject itself (e.g. electrical activity from the heart, blinking, etc.)

Preprocessing helps to clean this data from noise and artifacts. There are different filters applied to filter out noise when processing neurological signals.

One example of these filters is the band-stop filter. This filter is used to remove specific frequency bands from the data, allowing the others to pass with minimal loss. The notch filter is one of the most commonly used bandstop filters. This filter typically removes frequencies between 50–60 Hz (usually noise from powerline systems).

Feature Extraction, Classification, Translation

The next step is to analyze the signals and extract the relevant features. EEG data can be very complex, therefore it is quite difficult if not impossible to find meaningful information just by looking at it.

Classification — With signals mostly clean at this point, another step can be applied to classify them. Using various machine learning techniques, it is possible to classify signals to recognize various features, that belong to various classes.

Translation — Following classification, the results go through a feature translation algorithm. This allows features to be translated to the action required (e.g. a specific signal could have come from looking at a specific object, which the translation would recognize and help evoke any following processes).


Using the translated signals, the algorithm is then able to send a command to the feedback device to execute. For example, it can be a tablet in which the signal will be used to pause a video or a vehicle in which the signal will be used to indicate go forward.

Challenges and the Future

With the BCI industry steadily growing, the future looks bright but there are still challenges that are hindering progress. These include:

  • Hardware — non-invasive applications are restricted as a result of the amount of noise, and their low spatial resolution, but invasive applications have a very high risk and electrodes placed in the skull have the possibility of creating scar tissue in the brain
  • Lack of understanding — present-day BCI technology being crude, and our understanding of the brain being relatively novel, restricting the speed of innovation
  • Ethical and Legal Challenges — with the possibilities of advancements in BCIs there are an abundance of ethical and legal concerns regarding user safety, privacy rights, risk of hacks and manipulation, liability, among many others

The field of brain-computer interfaces has an immense amount of excitement and buzz and a lot of potential for innovation. The industry has only scratched the surface of possibilities that this technology could make a reality in the future.


  • A brain-computer interface is a device that is able to create a direct communication pathway between a brain and an external device (this can be one way, or more recently worked on both ways).
  • 3 main types of BCIs: non-invasive, semi-invasive, and invasive. Each one has its own disadvantages and advantages in terms of signal quality and overall risk.
  • Electroencephalography or EEGs are the most common types of BCIs where electrodes are placed on the exterior of the skull to measure electrical potentials produced by the brain
  • The brain has 3 main divisions: brain stem, cerebellum, and the cerebrum (which is divided into 4 main lobes)
  • The brain has 68 million neurons, through the use of neurotransmitters, neurons are able to connect with one another to fire, this action potential is able to be measured by EEGs
  • When creating a BCI project using an EEG, there are 4 main components: Collect (different frequencies of brain waves can be measured), Preprocessing (reducing noise), Feature Extraction (identifying signal meanings), and Deployment (command execution based on signal translation)
  • There are many hurdles within BCIs including hardware, lack of understanding, ethical and legal challenges, among others, but the industry is steadily growing and the possibilities are vast

Let’s Connect!

If you enjoyed reading this article or learned something new, I’d love to connect on LinkedIn. If you’d like to stay updated on my new articles or projects you can subscribe to my newsletter here!