On Brain-Computer Interfaces

30 min readOct 11, 2023

A brain dump of what I learn.

Imagine a world where we can communicate with just our thoughts.

A world where the music we listen to, matches how we feel.
A world where we can track our brain performance to improve our cognition.
A world where brain disease is finite and brain health is infinite.

Our most potent ideas don’t arise from the world around us.
They arise from the world within us.
From the most powerful tool at our disposal.
A tool more powerful than any technology we can ever harness.

The Human Mind

Imagine a world where we can unlock the full potential of the Human Mind.

This is the world of Brain-Computer Interfaces.

BCI Basics

Simply put, a Brain-Computer Interface (BCI) is a computer-based system that allows for communication between the human brain and an external device.

Like a computer.
Or a phone.

But how?

Well, BCIs have the ability to read our minds.

At least, sort of.

They have the special ability to read the brain’s electrical activity.

The billions of neurons in our brain constantly fire generating electrical signal at the axon.

Electrical signals generate at the axon and propagate to the axon terminals releasing neurotransmitters

These electrical signals are called action potentials.

Action potentials are constantly generated in our brain as we live our daily lives.

While we walk, talk, and even think.

It’s as automatic as our blood flow.

In fact, without these action potentials, we literally wouldn’t be able to function. They’re what allow for all movement, sensing, and biological functions. They’re vital to our existence.

Invasive BCIs, meaning BCIs that are surgically implanted in the human brain, have the ability to capture specific action potentials from specific brain regions or individual groups of neurons.

Deep Brain Stimulation (DBS), an invasive BCI

Now, the billions of neurons that fire at once amount to somewhere between 86 billion to 17.2 trillion action potentials per second! [1]

These billions to trillions of action potentials release an immense amount of neurotransmitters which can then trigger the generation of post-synaptic potentials.

Post-synaptic potentials refer to the electrical activity in our brain that’s produced after a neuron receives a neurotransmitter.

The high amount of post synaptic potentials that are generated amount to brainwaves [2].

The different brain waves that represent different physiological states.

Non-invasive BCIs, meaning BCIs that function without entering the human body, are able to read these brain waves through detecting the extracellular field potential.

The extracellular field potential refers to the sum of propagated action potentials detected at a distance from the brain region they were produced at.

They can be heavily useful for determining the mental state of a person. Detecting whether someone is drowsy, alert, or calm.

Emotiv’s Epoc, a commercial non-invasive BCI

Semi-invasive BCIs, are BCIs that enter the human anatomy, but aren’t surgically implanted as deep into the brain as invasive BCIs. They typically sit right below the skull yet above the brain. Some types include electrocorticography (ECoG) and stereoelectroencephalography (sEEG).

BCIs intend to accurately capture electrical brain signals but the gathered data tends to get messy. Each neuron can release electrical signals with a voltage of only a mere ~60 millivolts (mV) [3].

Therefore, a BCI must be finely tuned and highly sensitive to detect these tiny electrical potentials.

Unfortunately, the high sensitivity of a BCI can inadvertently lead to unwanted artifacts.

Artifacts are undesirable electrical signals that contaminate brain signals and create unwanted noise.

Functions of our physiology such as eye blinking / movements, muscle activity, and organ function can interfere with signal quality in a BCI system as they release electrical signals as well [4].

External electrical factors such as phones, lightbulbs, laptops, and wiring can also produce unwanted artifacts.

To address this challenge, we introduce signal processing in a BCI system.

Signal Processing

Signal processing is the removal of any unwanted and inhibiting electrical noise. It enables the extraction of any meaningful and useful information from a BCI system and translates it into usable data for external devices.

This is typically completed in 5 stages [5].

Signal Acquisition
Pre-processing
Feature Extraction
Classification
Application Interface

**Deciphering brain signals through 5 stages |** LINK

Signal acquisition involves the capture of the electrical signals.
Through the selected BCI device, electrical signals are captured which are then amplified through a Low Noise Amplifier (LNA).

This is a low noise amplifier (LNA). They can be small. Or big.

This is done in order to clear any unwanted electrical noise and improve the Signal to Noise Ratio (SNR).

SNR represents the balance between a clear signal and background noise.
A higher SNR indicates a quality signal
A lower SNR indicates poor signal quality and a high level of unwanted noise.

Then, the signal is digitized from analog signals, meaning continuous physiological signals that constantly vary, to digital signals in bits (0s and 1s) through a Analog to Digital Converter (ADC).

This is an analog to digital converter (ADC).

The next step, pre-processing, involves filtering out any redundant noise or artifacts that happened to inadvertently be captured by the BCI.

It’s done through a process called spatial filtering.

Unwanted electrical signals and electrical frequencies are removed from the dataset in order to provide a clearer picture of necessary electrical signals.

Once our dataset is properly cleaned and filtered, we extract the valuable information we need. This is referred to as feature extraction. The electrical signals that are relevant for our intended use are identified and extracted from the rest of the dataset.

Classification is the process of classifying certain electrical signals to their corresponding actions. During classification the extracted data is fed into classifier algorithms that aid with the identification of an electrical signal to the related mental or physical actions.

There are a multitude of classifier algorithms that can help aid the process of classification. Some of them include neural networks, support vector machines, linear discriminant analysis, and gaussian classifiers.

This process aims to determine the event related potentials (ERP).

Event related potentials (ERP) are the action potentials that our brain releases with direct relation to specific extrinsic or intrinsic stimuli.

Finally, after classification, the data is sent to an application interface where we can use the provided neurofeedback.

Neurofeedback refers to the digitized biosignals / feedback we receive from our brain and its electrical activity.

The electrical signals are then translated into specific commands that can be used for the intended application. The specific commands and application interface can vary depending on the intended use of the electrical signals.

For example, the specific commands can be movement control for the T-Rex in the Dinosaur Game

Credit: Uri Shaked on YT

Or the specific commands can be the backend functions for a program that tracks your focus / stress levels via neurofeedback.

Neurable Enten

Or, the specific commands can be used for brain stimulation to help patients with Parkinson’s disease

**Closed-loop deep brain stimulation. We’ll get into this later!**

Great! Now you have a very basic understanding of BCIs.

Let’s dive in a little deeper.

[talk about invasive bcis as a whole first, then refer to examples??]

Non-Invasive BCIs

Non-Invasive BCIs are BCIs that function without entering the human body. They work through the introduction of electrodes onto the skull of a human.

These electrodes detect the electrical signals of the brain in order to determine the brain waves they’re producing which then can determine the brain state they’re in.

Delta Waves (.5–4 Hz) indicate deep sleep.
Theta Waves (4–8 Hz) indicate dreaming, creativity, or meditation.
Alpha Waves (8–13 Hz) indicate relaxation and passive attention
Beta Waves (13–35 Hz) indicate alert mental engagement.
Gamma Waves (>35Hz) indicate state of intense focus and concentration.

Non-Invasive methods come with multiple benefits and downfalls.

And with cool real world applications as well.

Let’s explore an example, Electroencephalography.

Electroencephalogram (EEG)

The word electroencephalogram might seem like a mouthful, but don’t worry, it isn’t too complicated to understand.

The # of electrodes on an EEG system can highly vary depending on its intended use. An EEG system may have as few as four or as many as 256 electrodes [7].

A cool advantage of an EEG system is it’s high temporal resolution.

Temporal resolution refers to the level of accuracy when measuring the electrical signal at the precise time it occurred.

This high temporal resolution is insanely useful when accurately determining the event related potentials.

They’re also relatively cheaper when you compare them to invasive BCIs. And pretty simple to use!
At least when you compare them to using invasive BCI systems…

Commercial EEG systems are pretty straightforward.

They come in the form of wearable headsets or wearable headbands.

Some examples of these wearable EEG systems are

The process of setting up these wearable EEGs, analyzing neurofeedback, and personalizing it, is very intuitive and user friendly.

These commercial EEG products typically come with a pre-designed application for putting the wearable to good use. They’re typically designed for a specific use-case such as Neurosity’s app which tracks focus.

Here’s an example:

To keep things simple and easy to use, these wearables use dry electrodes.

Unlike wet electrodes (we’ll go into them later), dry electrodes don’t require extensive preparation of a subject’s scalp for use. Also, dry electrodes don’t need to come into direct contact with the scalp. They can either be contact electrodes or non-contact electrodes.

Non-contact electrodes have the ability to track electrical signals through clothing, thin mesh, plastics, insulating mediums, and other materials.

Commercial EEGs typically use non-contact electrodes as they make it easier for the creation of easy-to-use devices such as Neurable’s Enten.

In the case of the Enten, the electrodes go under the headphone mesh pads which gives it a need to be non-contact.

**Neurable Enten — Tracking focus and stress**

Though, some systems may make use of contact electrodes in addition to non-contact electrodes in order to increase signal quality.

Applying non-contact or contact dry electrodes to the scalp via commercial EEGs is pretty straightforward.

All you need to do is put on the headset / headband. Simple.

Applying dry electrodes to the scalp in non-commercial EEG systems is similar in the context of simplicity and straightforwardness.

You can very easily just stick them onto the scalp.

But won’t they fall off?

They don’t. At least, if you have the right system for mitigating that risk.

Electrodes can have the mechanical support of a headband, electrode caps, or other mechanical structures that allow them to stay in place.

They can also be supported using non-irritating adhesives to help them stay in place before any structural support from the electrode itself or an additional mechanical structure comes into play.

The design of the electrode allows it to remain on the scalp without easily falling off.

Each contact dry electrode typically consists of an array of Silver/Silver-Chloride (Ag/AgCl) coated spikes at it’s surface.

OpenBCI’s Electrodes coated with **AgCl** at the tips| LINK

These spikes are what can allow for the electrode to stay put.

The material of the electrode spikes, Ag/AgCl, can allow for improved electrical conductivity allowing for higher data clarity, improved signal processing, and accurate interpretation of neurofeedback [6].

Interestingly, dry electrodes appear to have a better Power Spectral Density (PSD) than wet electrodes [8].

PSD measures the power and intensity level of an electrical signal at a specific electrical frequency.

This is because of a saline solution that’s often used in wet electrodes.

It’s use affects the electrical contact between the electrode and the scalp ultimately changing electrical impedance levels [8].

Impedance levels appear to be higher in dry electrodes which happens to increase the PSD [8].

This can be an advantage when attempting to gauge very subtle changes in a large dataset of electrical signals.

Yet, the higher impedance levels of dry electrodes can come with a big disadvantage.

The lack of a conductive layer, such as hydrogels, aerogels, or other solutions, can lead to increased signal impedance and the increased signal impedance can lead to a decrease in spatial resolution and poor signal quality.

Spatial resolution refers to the level of precision when attempting to capture electrical signals from singular or small sets of neurons.

The Ag/Cl material just isn’t enough to improve the signal quality

Just more unnecessary noise… 😔

But wet electrodes are a little different.
Rather than having a spiked surface, it’s surface is metallic.

They’re typically coated in gold, silver, or copper.

This is due to their impressive conductivity properties. This becomes crucial when attempting to capture the electrical signals released by our brain.

The risk of degrading of the metallic materials is little to non-existent which ensures optimal signal quality.

The other factor that allows wet electrodes to achieve optimal signal quality is the use of conductive mediums.

They make effective use of

Aerogels
Hydrogels
Saline Solutions

These conductive mediums can provide many benefits to an EEG system.

First, they allow for the electrode to stay on the skin without additional support. The aerogels, hydrogels, and saline solutions applied onto the electrode are designed to be sticky which allows it to stay on the scalp.

They allow for improved conductivity between the electrode and the scalp which can reduce electrical impedance.

The reduced electrical impedance improves the spatial resolution which then allows for the capture of increasingly precise electrical signals from smaller groups of neurons.

This decreased impedance also allows for a reduction in artifacts that arise from motion that may disrupt signal quality.

The reduction in motion artifacts is heavily useful when conducting studies that require movement from the subject.

This means tests can be conducted while a person goes about their daily lifestyle as long as nothing comes in direct interference with the electrode itself.

Wet electrodes also exemplify a higher SNR due to their conductive medium. This decrease in noise with a higher signal quality provides benefits to spatial resolution.

So as you can see, wet electrodes can have an improved signal quality over dry electrodes.

But they do come with some limitations.

Setting up wet electrodes can take up waaaayyy more time than dry electrodes.

They’re a tad bit more complex to setup than dry electrodes, yet definitely not too difficult if you know what you’re doing. If you don’t know what you’re doing, you may need the help of a professional [10].

Here’s why:

First off, in order to achieve optimal signal quality, the scalp of the subject needs to be properly prepared to make effective use of each electrode.

The scalp must first be thoroughly cleaned by removing any amount of scalp buildup or hair products.

From there, the electrode can then applied with a conductive gel or solution and then placed on the scalp.

Yet, we still have to wait a bit longer.

Once on the scalp, the electrical impedance levels may not be at acceptable values.

We’d need to spend time waiting for the gel to dissipate onto an even layer. Only then would the electrical impedance levels be deemed acceptable for good quality [8] [10]. Acceptable impedance levels are generally around 10 ohms but going to 5 ohms and below is definitely recommended and could be even better depending on the context [9] [10].

Yet, ironically, waiting for the gel to dissipate into an even layer to solve a problem inadvertently creates another problem.

As we wait for the gel to dissipate into an even layer, the gel can begin to dry up [8] [10].

When the gel dries up, it can result in a decreased signal quality [8].

Electrical impedance increases which can tarnish the SNR and then further lead to a decreased spatial resolution. This increased impedance can also lead to more artifacts, both physiological and environmental, which degrades our signal quality even more…

We also can be limited to using a lower count of total electrodes.
At least when relying on a wet electrode system.

A higher electrode count implies that there’s a smaller distance between each electrode. Through the gels or solutions used, this can inadvertently generate conductive bridges between each electrode.

The conductive bridges lead to increased signal noise which lowers the SNR which then decreases the spatial resolution.

Remember, a high spatial resolution is required for higher accuracy of neurofeedback. Therefore, conductive bridges are detrimental.

All of these factors you may need to consider can be complicated, at least for the average layperson which is why one may need professional help.

Some other problems, this time regarding EEGs as a whole is the volume conduction problem.

Volume conduction describes the process by which electrical signals travel and spread through a conductive medium.

This problem centers around the fact that EEGs tend to be sensitive to surrounding electrical noise. They can inadvertently pick up on signal from deeper regions of the brain that lower the SNR and degrade the spatial resolution.

Some solutions to the volume conduction problem do exist. They involve the use of spatial filtering techniques, such as surface laplacian, in order to mitigate this. It involves estimating the electrical signals that arise as a result of deeper brain signals and filtering those deep signals out. Then each electrode isn’t sensitive to the deeper signals, rather only to the surface signals. [11]

Either way, signal noise from neural activity can be inevitable at times.
If someone wants to capture the activity of individual neurons, this isn’t possible with EEGs. Instead of capturing the action potential of specific neurons, they gather the extracellular field potential

When a neuron generates an action potential, the electrical activity propogates throughout other neurons resulting in each one generating their own action potential. The sum of all these electrical potentials is referred to as the extracellular field potentials. It represents the synchronized activity of a group of neurons.

A singular electrode records the vector sum of all neural activity within the proximity of that specific electrode resulting in the capture of the field potential rather than individual neural action potentials.

Now, the science behind EEGs can make them seem intimidating.
Though, overall, EEGs aren’t really too complicated to use.

The complex scientific functions behind them can make them seem complex but their use can be straightforward.

And they’re definitely simpler to use when you compare them to invasive or semi-invasive BCIs.

Let’s find out why.

Invasive BCIs

As mentioned earlier, invasive BCIs are BCIs that enter the human physiology. Meaning they go inside us. Inside our skull.

Doesn’t that sound cool?
Well, maybe it’s a little creepy…

This is done so through neurosurgery.
The BCIs electrodes are implanted into deeper areas of the brain in order to detect the electrical signals from the intended brain region.

Some types of invasive BCIs include, Microelectrode Arrays (MEA), Deep Brain Stimulation (DBS), and Endovascular BCIs.

Real world use cases for these invasive BCIs include:

Innovative Research
Aiding Paralysis
Mitigating Epilepsy
Alleviating Neuromuscular Disorders

These invasive BCIs can be divided into two categories.

Single Unit & Multi Unit [12]

Single unit BCIs are very useful for detecting the electrical signals of specific groups or individual neurons. They’re able to track the precise signal and firing power of very tight regions [12]. These type of BCIs are great for applications that need to record very specified brain signals such as paralysis recovery or communication via thoughts (for patients with locked-in syndrome)!

Multi unit BCIs are used when there’s a need to record the collective activity of a group of neurons. They still provide a very good signal quality given that they’re invasive. These types of BCIs are useful for applications that don’t need fine grained control but rather need to record a greater breadth of neurological activity. They can be used for robotic limbs or neurological / behavioral research.

These invasive BCIs have some distinct and unique advantages over non-invasive BCIs.

Invasive BCIs happen to harness a very high signal quality with a very high spatial resolution. Given that they’re invasive and go into deeper brain regions, they’re able to capture increasingly precise brain signals from those deeper brain regions.

Their invasive characteristic means they also have a higher SNR which further enhances the spatial resolution. There’s a reduction of crosstalk and noise from neighboring groups of neurons as the invasive electrode is already on site of the intended brain region.

In addition, invasive BCIs don’t have to worry about the volume conduction problem. Typically, the electrodes of invasive BCIs are already placed at the intended site of where the intended brain signals originate from. The electrical brain signals don’t need to travel any further distance, they’re immediately picked up by the invasive electrodes.

Even though the temporal resolution was already excellent in non-invasive BCI systems, amazingly, there’s an even greater level of temporal resolution in invasive BCIs.

So by now, it may seem that invasive BCIs are like heaven on earth.

But,

Invasive BCIs do come with some limitations.

First off, they can expensive.

The combined price of the neurosurgery and the device can can be pretty big.

The procedure itself can also be risky.

Risks of brain scarring are present and can occur due to an incorrect placement of the BCI. The formed scar tissue can heavily degrade signal quality as the scar tissue surrounds the BCI and directly interferes with its electrodes.

There’s also a higher risk of neural damage which can lead to severe complications.

For example, if for any reason, the motor regions of the brain get damaged, this can lead to forms of paralysis.

Also, there are security risks. Invasive BCIs, just like any other electronic device, can be prone to cyberattacks. Cyberattacks can compromise confidentiality, integrity, and your brain activity [13].

Imagine a cybercriminal gaining access to how you think.
They’d be able to read your mind.

If, in a theoretical (and cool) future, where your neurological activity is used to authenticate you just like a fingerprint does, cybercriminals could potentially infringe upon your secure data.

Anyways, enough theory.

Let’s take a pleasant walk into real examples of invasive BCIs

Microelectrode Array (MEA)

Microelectrode Arrays (MEAs) are BCIs in the form of a tiny grid.

A very tiny grid.

MEAs have been typically made with the materials of silicon, glass, and a gray-white metal called tungsten [17].

They can vary in size but they’re typically on the very small order of micrometers (µm) to millimeters (mm).

Common MEAs range from 2x2mm to 6x6mm.

Some real world MEA technologies include:

Blackrock’s Utah Array
Microprobes’ FMA

On these tiny grids, are tightly spaced electrodes well embedded onto a multi-well MEA plate.

A multi-well MEA plate is the plate consisting of multiple tiny wells or chambers inside of which reside the electrodes.

These electrodes are only a couple hundred of µm apart.

For comparison, 300 µm equates to 1/3rd of a millimeter.
This is only about 3–4x the diameter of a strand of human hair…
MEA electrodes can only about 3 human hair strands apart. So tiny…
This isn’t the only astonishing fact.

Given that the electrodes on an MEA are only a couple of hundred µm apart, the thickness of the electrode itself must be insanely small.

The diameter of the electrodes can be as tiny as 5 µm! [15]

5 µm is about 1/14th of a human hair strand! (I did the math)
Isn’t crazy that we can build technology on such a nanoscale level?

Electrode tip (Top) compared to a human hair strand (Bottom) | LINK

Or, they can scale up to 120 µm. [15]
It depends on the context of its use-case scenario.

The height of these electrodes can heavily vary on the order of centimeters. In some contexts, the height of the electrode can be as small at 1.5 mm or go up to 10 mm [16]. Typically, if you want deeper brain stimulation or signal detection, you’d go for the taller electrodes as they have more of a reach.

MEAs definitely do have some advantages.
Especially when considering their signal quality and flexibility in use-case.

Given that MEAs are a form of invasive neurotechnology, they have a very high spatial resolution and temporal resolution. Definitely higher than non-invasive BCI systems. MEAs are able to get in very close proximity of the target brain regions. It’s the very close proximity that decreases the SNR which then increases the spatial resolution.

This completely eliminates the volume conduction problem which we talked about earlier.

With this close proximity they have the invaluable ability to precisely track the action potentials of singular neurons! [18]

This ability is insanely useful for finding the event related potentials (ERPs)!

Harnessing the power of action potentials from individual neurons can lead to life changing innovation.

It can potentially allow for:

Advanced robotic prosthetics
Greater understanding of neurological disorders
Advancement of neural networks
Mind communication (theoretical but check this out)

In addition to this, they also have the ability to be setup to track neurotransmitters. This very useful and exciting ability can make way for understanding specific brain states and functions on a deeper level!

For example, if effectively used, in theory, MEAs could detect the specific neurochemical response to states of high stress. This could lead to a deeper understanding of how an individual’s brain reacts to stressful stimuli and how one can begin to change their response remain composed under stress.

Yet, MEAs aren’t only about tracking individual neuronal activity.
They can also, very easily, be applied to larger groups of neurons.

Rather than recording the action potentials of singular neurons, they can track the field potentials of wider groups of neurons which can be useful when attempting to gain a more holistic understanding of the brain.

Now,

Would you like your brain zapped ⚡️?

Well, you might not.

But zapping the brain can be very helpful in curing neurological disorders and diseases.

Luckily, MEAs can zap the brain.

Electrical currents can be sent through the MEA electrodes which then can generate action potentials within the targeted neurons.

This can potentially allow for

The treatment of paralyzed patients
Epilepsy prevention, management, and treatment [19]
The study of neural circuits

Yet, like most technologies, there can be some limitations.

Remember how MEAs are typically made of silicon, glass, or tungsten?
Well when fully optimizing MEAs to be useful, safe, and long-lasting, the high stiffness of those materials can be a limiting factor.

Those materials possess a very high Young’s Modulus (GPa) when compared to the tissues of the human body.

Young’s Modulus is a measure of the ability of a material to remain changeless in terms of length when put under heavy load. Basically, it just measures how stiff a specific material is.

Silicon has approximately ~190 GPa [21]
Glass is typically between 60–100 GPa [21]
Tungsten is generally at ~410 GPa

On the contrast, human tissue is on the order of kilopascals (kPa) which is nothing when you compare to GPa. [21]

If we take gray matter for example (a tissue in our brain essential to our central nervous system), it’s approximately at a stiffness of 3 kPa.

MEA materials are therefore about 3,000,000 to 136,666,667 times stiffer than gray matter.

The disproportional stiffness levels between MEA materials and brain tissue is detrimental to the former and the latter.

The incompatible materials, if used improperly, can lead to a lack of proper cell adhesion. This means the MEA isn’t correctly situated in the proper location of the brain. The MEA can inadvertently shift around inside the brain causing stress on the surrounding tissue.

Unfortunately, this can lead to risks of a foreign body response.
A response in which our physiology naturally reacts to the insertion of a foreign substance into our body.

While a foreign body response (FBR) is meant to protect our inner physiology, it can potentially degrade MEA quality.

Some FBRs in the brain can include

Increases in microglia and astrocytes
Brain scarring
Neuroinflammation

While initial brain scarring inherently isn’t bad for brain tissue itself as it can aid healing, it can be detrimental for MEAs. Increases in microgalia and astrocytes can begin just 30 minutes after initial MEA insertion.

Microglia and astrocytes are types of brain cells that play a crucial response in brain function and regulation They can be considered brain police.

It’s the heavy increases in microglia and astrocytes that lead to brain scarring.

After 24 hours, they can form a fully encapsulated sheet around the MEA [21].

Over a longer period of time, due to increases in brain scarring there can be heavy decreases in MEA signal quality and stability as well as increases in signal impedance which can ultimately then lead to signal failures [21] [22].

Also, microglia can become neurotoxic if they remain active for very long periods of time. This can potentially cause irreversible damage to surrounding neurons and tissues.

All of this is something to keep in mind.

Insertion of MEAs need to be precise.
MEA materials need to be well suited for the job.

A potential solution can be to use biomaterials that are properly suited for neuro-invasive technology. Special biomaterials can decrease the likelihood of FBRs which then eliminate a possibility of a degraded MEA.

Luckily, we’re constantly developing new technology that can not only solve these problems but also enhance the performance of MEAs

Deep Brain Stimulation (DBS)

Deep Brain Stimulation (DBS), is another way to zap your brain⚡.
With a very different methodical approach.

DBS is has been typically considered an invasive form of neuromodulation rather than an actual BCI. This is primarily because DBS usually puts more focus on brain stimulation rather than sending recordings of neural activity to a computer.

But recent advancements in the development of closed-loop DBS systems integrate a computer aspect converting DBS into a type of BCI.

Closed-loop systems operate continuously and automatically by detecting a stimulus, generating a response based on that stimulus, and then detecting new information from the response to generate another response. This continuous feedback loop allows the system to adapt and respond to changing conditions.

More specifically, DBS can be considered a bidirectional deep brain-computer interface [23].

Bidirectional refers to the ability to both capture neural signals and stimulate the brain.

DBS can have life changing applications in the field of medicine.
They can be useful when treating:

Parkinson’s
Depression
Epilepsy
OCD
Other neurological disorders and disabilities

DBS systems are implanted through a surgical procedure.

But before neurosurgery, a patient typically has a brain scan done via CT or MRI in order to determine the malfunctioning brain region which will be the target of DBS.

Once the intended brain region is determined, electrodes are surgically implanted into the intended brain region through a coated wire called a lead.

Electrode count within a DBS system is typically only 1 but there have been proposed systems that make use of 2 electrodes [24]

The lead is then connected to an extension wire at a small hole in the skull which then runs down inside the skin, through the neck and shoulders, to the chest or abdomen.

It’s connected to a pulse generator which sends out electrical signals to stimulate the brain.

The electrical signals that are sent out from the pulse generator can be altered in intensity or frequency depending in individual patient need.

In a closed-loop DBS system, electrical signals are captured by the electrode(s) implanted within the brain. Those electrical signals are then read by a neural sensor that interprets the brain signals for the pulse generator [25]. Then, using those electrical signals, the pulse generator automatically adjusts the intensity and frequency of the brain stimulation to suit the needs of a patient [26].

Let’s use Parkinson’s disease as an example.

It’s a disease in which areas of the brain experience degradation in neural circuitry and a lack of dopamine production.

Dopamine production and neural circuity (how your neurons are wired), are essential for movement.

Degradation within those systems impair movement resulting in tremors, slowness in movement, or impaired balance and coordination.

Closed loop DBS can detect when the brain misfires or fails to fire which happens to be the case in Parkinson’s. It detects this change and then stimulates the specific region of the brain with this fallacy in order to re-enable proper movement.

This is MIND BLOWING.

Just imagine the benefits for those who really need this type of technology…

DBS systems can be a replacement for medication, when it fails to work.

Parkinson patients are prone to adapting to their medications by developing a high tolerance towards it.

Medication can be finite.
Closed loop DBS can be infinite.

DBS can be life changing and even life saving.

If correctly applied, those with

Depression
Parkinson’s
Epilepsy
OCD
Other neurological disorders and disabilities

will be able to function more efficiently on a day to day basis.

Each and every one of them with their own personalized treatment.
With the higher ability to live out their lives they way they were meant to live.

Amazing huh?

If you’d like a more detailed look at how DBS can aid neurological disease and disorders, click here or here.

Yet again, like any technology, there are some limitations and disadvantages.

And some risks involved as well.

During neurosurgery, there runs a risk of complications arising.

3% to 4% chance of brain hemorrhage [27]
Infection / Sepsis/ Neuroinflammation
Coma
Stroke

Device malfunction can also occur.

Incorrect placement of leads
Leads coming loose
Failure of electrodes or pulse generator

Some side effects from brain stimulation such as

Speech or vision problems
Loss of balance
Memory problems

Yet, these risks aren’t very high, DBS is considered and see as a low risk procedure with most complications having about a 2% chance of occurring [28]

Innovation in DBS can provide useful treatment for many neurological disorders and diseases as well as create exciting futures.

I think it’s important to consider where we fix issues and make improvements within a DBS BCI system.

Luckily, there are studies currently exploring the feasibility of closed-loop DBS systems. You can check one out here.

So far, we’ve covered invasive and non-invasive BCI systems.

Now it’s time to go in between.

And quite literally as well.
You’ll see.

Let’s dive into semi-invasive BCIs!

Semi-Invasive BCI

Semi-Invasive BCIs are BCIs that enter the skull but don’t go as deep into the brain as invasive BCIs.

Rather than going deep inside the brain, BCIs are actually inserted onto the surface of the brain.

Right underneath the skull and on top of the cortex.

**Placement of ECoG, a type of semi-invasive BCI**

Like invasive procedures, semi-invasive BCIs require insertion through surgery.

Some types of semi-invasive BCIs include:

Electrocorticography (ECoG)
Stereoelectroencephalography (sEEG)

Can you pronounce them?

Just like invasive BCI systems, these semi-invasive systems are typically put into use for aiding neurological disease and disorder.

Semi-Invasive systems do have their advantages, especially over non-invasive systems.

They have a lowered SNR which then leads to a higher spatial resolution when compared to EEG and other non-invasive BCIs.

They aren’t as impacted by artifacts from surrounding devices or irrelevant physiological activity.

It’s improved signal fidelity is also heavily important.

Signal fidelity refers to how accurate a recorded signal is compared to the original signal itself. A lowered SNR would indicate improved signal fidelity which then leads to an improved spatial resolution.

The broader bandwidth of semi-invasive methods when compared to non-invasive methods is also heavily useful. This means they have the ability to capture data from a higher frequency range. This exemplifies an improved temporal resolution.

They also posit a high amplitude when compared to non-invasive systems.
This higher amplitude would mean non-invasive techniques have the ability to capture more electrical activity or more intense electrical activity at one time.

Amplitude refers to the total intensity of electrical activity in a given moment.

These increases in signal quality indicate that there’s a high potential for improved and precise control when compared to non-invasive techniques.

Did I mention that this type of BCI is bidirectional?

Meaning it can both, stimulate the brain and record electrical signals.

I didn’t?

Well now you know.

There’s also a lowered surgical risk when compared to invasive methods.
Semi-invasive methods don’t involve a need to penetrate the cortex which then reduces the risk of brain scarring, hemorrhage, and other neural dysfunctions.

Yet, a disadvantage of semi-invasive BCIs is it’s price. It can prove to be pretty expensive. The added cost of surgery and the device itself may not be affordable for those who need it.

Also, semi-invasive BCIs aren’t as effective for brain stimulation. You just can’t access the deep areas of the brain that you need with a device that’s planted above the cortex…

This also means that semi-invasive methods can suffer from the volume conduction…

This further degrades the SNR.

In addition, although there is a lowered surgical risk when compared to invasive methods, the risk isn’t yet obsolete. It’s just that the negative implications of these risks aren’t to the same degree. Just tad bit lower. Just like invasive methods, there is a slight risk of brain scarring and neuroinflammation is the device is inserted improperly.

So,

I think we should go into some quick examples of semi-invasive BCIs.

Let’s begin with electrocorticography (ECoG)

Electrocorticography

Electrocorticography (ECoG) is a semi-invasive BCI system that’s composed of a series of flat electrode contacts.

These electrode contacts are placed on a thin silicone sheet which sits right on the brain surface right under the skull [29].

**Just like this. Neat isn’t it? [link**]

The specification of ECoGs can heavily vary. They can go from being a couple millimeters in length and width to a couple of centimeters. Electrode count may also heavily vary. This all depends on its specific application.

Prior to it’s placement via invasive surgery, MRI or CT scans are typically used to identify the affected brain regions that need the aid of ECoG.

Once that’s done, ECoG can be inserted.

ECoGs can be placed on two different surface areas of the brain.

They can either be placed epidurally or subdurally.

Epidurally means above the dura
Subdurally means below the dura
The dura refers to a tough and thick membrane surrounding the brain protecting it from any potential damage.

If an ECoG is placed epidurally, it’s being placed right above the dura.
If an ECoG is placed subdurally, it’s being placed below the dura and right above the cortex (below the Pia Mater; refer to figure above).

It’s implementation can serve to be very useful for specific medical applications.

One huge example is it’s use in cases of epilepsy.
ECoG can allow for the identification of epileptogenic zones.

Epileptogenic zones are the brain regions that are responsible for seizures.
Remove them and the seizures will cease to occur.

Identifying the epileptogenic zones can enable the identification of appropriate medical procedures for treating epilepsy.

Given that semi-invasive BCIs exhibit a high spatial resolution (thought not as much as invasive BCIs) they can be very useful for neuroscientific research as well.

Let’s move on to another type of semi-invasive BCIs.

This one’s tricky to pronounce.

Stereoelectroencephalography (sEEG)

Stereoelectroencephalography, also known as sEEG, is another semi-invasive BCI system.

sEEG makes use of depth electrodes. Depth electrodes are specially designed to go deep into your brain. Hence the name, depth. These electrodes are thin and floppy wires about the thickness of a spaghetti noodle [30].

Similar to ECoG, sEEG is geared towards epileptic treatment.
They aim to detect epileptogenic zones when ordinary EEG devices aren’t able to.

And just like other semi-invasive and invasive methods, sEEG requires surgery.

To insert electrodes into the patients brain, thin incisions are made in the skin and skull, typically no bigger than a grain of rice.

Each individual electrode is then passed through each individual incision in order to reach the intended brain region. A sEEG system typically involves the use of 8–15 depth electrodes

Afterward, the electrodes are secured by a bolt at the skull.

Post-surgery, a patient is given a CT scan to ensure proper electrode placement.

Finally, their brain is closely monitored by the sEEG. The goal is to track and record seizures in order to identify the epileptogenic zones and possible treatments [30] [31].

Monitoring the brain for epileptogenic zones.

sEEG may seem like an invasive technique rather than semi-invasive.
But the primary reason it’s considered semi-invasive is due to the fact that electrodes can be placed without removing parts of the school.

There are several advantages to this.

First and foremost, there’s a decreased risk of post surgery complications.
Brain scarring may not be as prevalent as other invasive techniques.
Expected post-surgery pain from other forms of invasive BCI systems are mitigated.

The signal quality is definitely an improvement over regular EEG systems.
sEEG allows for the electrodes to be at the direct site of the target brain region enabling for an ultra high spatial and temporal resolution. The volume conduction problem is mitigated and the SNR is very low.

As a result patients have the opportunity to receive targeted and individualized treatment once the sEEG identifies the brain region responsible for seizures.

One disadvantage of sEEG systems is the inability to record contiguous cortical regions, meaning they can’t record surface brain regions that are located right next to each other.

This is due to the fact that there usually are a limited number of depth electrodes in use and those electrodes intend to record deep areas of the brain. Not its surface.

Despite this, just like DBS, sEEG can be life changing for patients suffering from epilepsy allowing them to live out their lives the way they were meant to.