An explanation of what an EEG Brain-Computer Interface is and how they work, and the neuroscience behind it
What is a BCI?
Imagine a reality where you could use your brain to control things around you, almost like Yoda. Although this all seems like something out of a sci-fi movie or like a fantasy book, this is actually (kinda) possible and is currently being done right now by thousands of people around the world, and no, this is not limited to people with ‘gifted superpowers’, and yes, you can do it too!
This is done by what we call a Brain-Computer Interface (BCI), a system that allows for direct communication between your brain and a machine. This is done in three simple steps outlined below:
- Collecting signals from the brain
- Interpreting these signals
- Outputting the corresponding commands
Different types of BCIs
Because there are many different ways in which we can collect and measure data from the brain, there are many different types of BCIs, with each one having a different application due to its various pros and cons.
However, all these BCIs can be divided into three main groups:
- Non-invasive — One of the most common non-invasive BCIs is an electroencephalography (EEG) BCI. Collecting an electroencephalogram is done by placing electrodes on the surface of the scalp to measure the electric potentials caused by the brain, in the neurons. This is done externally and does not penetrate the skin nor the brain.
- Semi-invasive — The electrodes for semi-invasive BCIs are placed on the surface of the brain. This form of BCI requires surgery to create an exposure so that the electrodes can be either placed in the dura or the arachnoid. Where non-invasive BCIs collect EEG signals, semi-invasive BCIs collect ECoG signals.
- Invasive — This type of BCI involves directly inserting micro-electrodes into the cortex, to collect intraparenchymal signals, which is the activity of a single neuron or a very specific local field.
How do neurons even work?
The membrane potential refers to the difference in electrical charge (voltage) between the inside and the outside of a neuron when a neuron is at rest. The difference in electrical charge is due to the grouping of ions inside and outside of the neurone.
So the main ions (charged particles) which come at play here are sodium ions (Na+), chloride ion (Cl-), potassium ions (K+) and a few organic anions (-). When a neuron is at rest, there is usually a higher concentration of K+ and organic anions (-) inside the neuron than outside, whereas on the other side, Na+ and Cl- are usually at higher concentrations outside the neuron than the inside.
At rest, the neuron is more negatively charged than the outside, thus it has a resting membrane potential of -70mV (millivolts).
This voltage is maintained by sodium-potassium pumps which are on the cell surface membrane. This is what we call a ‘transport protein’, and what it does is it uses energy to constantly pump out 3 Na+and at the same time pump 2 K+ into the cell.
Additionally, there is an ion channel, another transport protein, which allows ions to pass through the cell surface membrane. Potassium ions tend to move through these ion channels fairly easily. So the potassium ions diffuse out of the neuron via the ion channel until it reaches the point where it is at equilibrium (when there is the same amount of K+ on the inside as the outside). It is at this point where the membrane potential is around -65mV to -70mV.
As there are more positive ions being pumped and diffused out than in, it helps to keep the membrane potential negative.
Now that we know the basics of the membrane potential, let’s have a look at how the neuron sends signals. This is what we call the ‘Action Potential’. The action potential is the momentary reversal of the membrane potential, and this is the basis of for electrical signalling within neurons.
For a neuron to fire (send a signal), it must first be activated. The activation happens when a certain amount of neurotransmitters are sent from the axon terminal of one neuron to the dendrites (receptors) of the next neuron. (See Figure 2)
When neurotransmitters bind to the receptors of the dendrites, they can have an effect on the neuron known as depolarization. This means that the membrane potential is less polarized, and the voltage moves towards 0mV.
When neurotransmitters repeatedly bind and interact with the receptors, eventually there will come a point where the potential of the reaches the threshold membrane potential, which is generally -55mV.
When the threshold is reached, a large number of sodium ion channels open, which allows many sodium ions (Na+) to enter the neuron. Because these ions are positively charged, this causes massive depolarization as the membrane potential reaches 0mV and becomes positive.
Eventually, the membrane potential reaches its peak at around 40mV and an action potential will fire, sending the electrical signal down the axon. Then the sodium ion channels close, and the potassium ion (K+) channels open, allowing K+ to flow out of the cell. This loss of K+ is what is known as repolarization, and is known as the falling stage of the action potential.
During the falling stage, K+ is lost a bit too much, thus the membrane potential becomes hyperpolarized. This phase is called the refractory period. And during this phase, it is quite difficult to cause the neuron to fire again. And finally, the potassium ion channels close and the membrane returns to its resting membrane potential at -70mV, where it is ready to be activated and fired again.
Figure 5: An action potential going from a presynaptic cell to a postsynaptic cell
After an action potential, neurotransmitters diffuse across the synapse to carry the impulse to the postsynaptic cell. The chemical transmission of neurotransmitters at the synapse results in something called the postsynaptic potential (PSP), which is a temporary change in the polarization of the neuron membrane. The PSP leads can lead to the firing of a new impulse.
So how does EEG work?
EEG measures the electrical activity in the brain, more specifically the postsynaptic potentials, but it is only when clusters of neurons fire together do they provide enough signal to be detected from the scalp using an EEG.
But what is actually being recorded is the difference in voltage (in the brain) between the placements of a minimum of two electrodes. These differences in voltages should be recorded simultaneously so that we can better understand and interpret an Event-Related potential — which is the measured brain response that is the direct result of a specific sensory, cognitive, or motor event.
The EEG data collected is comprised of the rhythmic activity of the brain, which reflects the neural oscillations that take place within. These neural oscillations are driven by the interactions between neurons and occur at specific frequencies — These include delta, theta, alpha, beta, and gamma. Studies have found that there are associations between these rhythms and different brain states, as shown below. Commercial EEG headsets, which are often used for meditation purposes, usually measure the amount of brain activity occurring in the alpha frequency.
EEG Signal Acquisition — Equipment
- Electrodes — Different types of electrodes can be used to collect EEG signals. Wet or dry electrodes can be used, wet electrodes having higher conductivity because the electrical distance is minimized. Although the majority of electrodes are made of stainless steel, these electrodes also come in many materials, such as tin, silver, and gold. While dry electrodes are more convenient and easier to use, they can lose higher frequencies.
- Amplifiers — Because the signals picked up by the EEG headset are poor, due to the electrodes being quite far from the neurons and the signal having to travel through the skull, an amplifier is required to bring the microvolts to a range that can be digitized. However, the cables of the amplifiers can act as antennas and receive external signals which would interfere with the EEG data and result in the noise being amplified. Alternatively, you can use ‘active’ electrodes which have a pre-amplifier within, which can avoid this interference, but they are quite expensive and large so that may be deemed as inappropriate for certain situations.
- A/D Converters — These are needed to convert the EEG data collected from analogue to digital so that it can be understood and processed by a computer. The bandwidth for EEG signals is confined to roughly100Hz, making 200Hz sufficient for sampling EEG signals.
- Recording Device — This can be any device which can record, store, and display the converted signals — generally a computer.
The raw EEG data that is collected often has a lot of noise and thus isn’t ‘clean data’. There are three main sources of noise in raw EEG data:
- EEG equipment
- Electrical interference external to the subject and recording system
- Electrical activity from the heart, eye blinking, eyeball movements, muscles movements in general of the subject.
Preprocessing helps us in identifying the noise in the data and remove it so that we have clean data. There are many different ways to preprocess data, a common example is using filters. The DC (direct current) components of a signal often result in distortion of the signal, thus, high-pass filters are used where a frequency cut-off of 1Hz is usually enough. Low-pass filters can also be applied to remove the high frequencies of the signal as frequencies above 90Hz are generally not studied in EEG. Other techniques are employed to eliminate noise and artefacts caused other factors such as the movement of the eyeballs or blinking.
This step includes looking at the data, analyzing it and extracting meaningful information. Because the EEG data is so complicated, it is next to impossible to get any useful information just by looking at. Therefore, we use sophisticated mathematical algorithms that can extract information hidden within the EEG data.
Additionally, classification models can be applied to the data. These classification models are machine learning algorithm which can be trained to recognize and classify features of the data into groups. Classification helps us to find out which kind of mental task the subject is performing
After the signal has been classified, it is passed through a feature translation algorithm, which translates the features to the corresponding action that the subject wanted to carry out, and is sent to the feedback device.
The feedback device receives the command from the translation step. For example, if the subject wanted to move the cursor on the computer screen, as the subject would think of it, the algorithms would pick up the signals which correspond to wanting to move the cursor, and this is sent to the feedback device to carry out
Brain-computer interfaces is a growing field with more and more advancements being made over time. From moving a cursor on a computer screen to direct brain-to-brain communication by transmitting EEG signals over the internet. Consequently, the application of EEG BCIs will only expand further in the near future and will soon be a part of our daily life.
Hi, I’m Mir Ali, I’m a 17-year old innovator working with The Knowledge Society (TKS), leveraging exponential technologies to solve the world’s biggest problems. Join me on my journey as a writer as I develop my knowledge and skills by working on fascinating projects to impact billions in the future.