Elena Pitsik— Junior Research Fellow, Laboratory of Neuroscience and Cognitive Technologies, University
Communication between the computer and the brain
A brain-computer interface is a system that enables interaction between the brain and the computer.It opens a channel of communication between the brain andcomputer and allows them to exchange information with each other. The simplest example is the generation of commands for an external device using brain activity. An external device can be a computer, an application, a robot, a drone, a prosthesis, an exoskeleton, or whatever. The scope of such interfaces is very wide.
The most important part of the interface is the technology for recording biological signals from the brain.There are many technologies that can be applied to interfaces.Signals of electrical activity are recorded using an amplifier called an electroencephalograph.They are transmitted to the server that processes these signals.very small, compact interfaces that use wireless technology.But if you're planning to use the interface in a clinical setting (for post-stroke rehabilitation, for example), you probably want to have a lot of channels and big data.Big Data processing requires a lot of computing resources, so most often a separate computer is used, on which there is aAnd after this software is processed byA brain signal, it translates into commands for an external device, providing feedback.In place of a classic monitor, it can be anything: from the application of monitoring your mood to a robot control system.
Brain-computer interfaces can be divided into three main types.First of all, it is an active neural interface, which implies that the user directly and consciously controlsA neural interface is the brain activity that the user generates, reads, and processesThe second type is a reactive neural interface.In it, brain activity is not consciously produced by the brain, but is stimulated by external influences such as light, sound, tactile sensations, and commands.For example, images or sound stimulation that trigger a certain response in our heads.And we can't generate this response consciously, without an external stimulus, so we need some kind of feedback.The brain-computer interface searches our brains for this response to an external stimulus and usesto generate commands and control an external device or to profile feedback.The third type of neural interface is the passive neural interface, which is the most complex and incomprehensible, because it does not imply this at allno commands for the external device and no control.The interface passively sits on you and monitors your condition in the course of daily activities.That is, you read, work, drive, and the neural interface monitorsIt can be used to track your level of attention, focus, fatigue, cognitive load, and also give you some feedback about your condition.
Neurorehabilitation and other uses of interfaces
The neural interface has a high social significance for the development of various rehabilitation systems and improving the quality of life of people with additional needs.Such as people with locked-in person syndrome – they have lost any possibility of contact with the outside world, but at the same time they have retained their consciousness; They're completely paralyzed people who can'tspeak, move, and even move their eyes.There are no other options for communication than to use neural interfaces in order to read their brain activity and convert it into specific commands.These commands can be speech generation, controlling various applications in order to maintain day-to-day functionality, as well as providing basic communication with other people.A less severe condition is partial paralysis, such as people in a wheelchair who are unable towalking, or various paresis after a stroke or as a result of an injury.In the case of paralysis, if it is incurable, then it is possible to improve the quality of life of such a person if you provide him with some kind of exoskeletonThis is one of the most popular areas in the development of neural interfaces.
The most popular application of neural interfaces in medicine is neurorehabilitation.This is an opportunity to speed up the process of rehabilitation of a person after an injury or stroke with the help of training with a brain-computer interface, which impliesThere are many studies that have shown that such interfaces actually significantly accelerateand improve the rehabilitation process of people, returning their lost motor functions.
Recently, brain-to-computer interfaces have broken through to the mainstream market and gained a wider audience.Today, you can buy a neural interface on eBay for a small amount of money, it will beFirst of all, they are usedNeural interfaces are also very often used in the gaming industry.to monitor various states, such as mood or sleep phases.A person relaxes and is guided according to the results of monitoring their brain activity.Similar devices are also used to develop various systems for training cognitive skills, such as attention, concentration, or cognitive load.This is an important area because it can be used, for example, in the development of cognitive load systems for pilots in airplanes and for truck drivers.
What is the brain and how you can measure its activity
You and I are our brain.The brain is made up of cells called neurons.from the nucleus, the body, and the various processes that ensure its communication with other neurons – these are dendrites, axons, and a terminal (an extension of the axon for communication with other neurons – "High-tech").A single neuron is a fairly simple structure, but there are billions of neurons in the head.In this way, all together they create a huge network that canto process complex amounts of data and provide us as individuals.
The brain is a supercomputer, that is, a system that is capable of processing enormous amounts of information.Neurons communicate with each other mainly in two ways: electricity and chemical bonds.The electrical signal is produced by a neuron and transmitted through the axon to the synapses (connections between neurons) of another neuron.When an electrical signal arrives at the output of a single neuron, it is converted into certain chemical messages called neurotransmitters.And they are transmitted from the output of one neuron to the input of another, which, in turn, receives them, processes them, converts themA neuron can enter an electrical signal and transmit it further down the chain.As to receive and pass through oneself an electrical signal, as well as to extinguish it, not to receive it, and to interrupt the chain.Any of our activities forms a network of neurons that fire continuously, formingA pattern that allows you to process a variety of active types of human activity.
There are several methods to measure the electrical activity of the brain.Firstly, this is magnetic encephalography,which involves measuring electromagnetic fields that arise as a result of electrical activity in the brain. A MEG machine is a large device in which a person sits motionless because the electrodes that read the magnetic field are not placed on the head, but are located at some distance from it. MEG is very sensitive to various external interferences - including the earth’s magnetic field and other magnetic fields with which our entire reality is literally filled. Therefore, for MEG to work well, it must be placed in a shielded room. Another method is near-infrared spectroscopy. This is a neuroimaging technology that allows us to visualize the processes occurring in the brain. It uses infrared light to measure the concentration of oxy- and deoxyhemoglobin. The device consists of a cap to which cylinders are attached, tightly adjacent to the surface of the scalp and divided into two types - source and detector. The source emits a beam of light in the infrared range. This light passes through the skin, skull and the upper layer of the cortex, that is, the brain. It passes through, returns back to the surface of the scalp and is caught by the detector. Moreover, everything that can be seen carries very important information about the level of oxygen in the blood. When we perform an action, such as tapping, our blood in the motor zone is saturated with oxygen. We can measure this and define it as a pattern. The next technology is electroencephalography, the most effective of these. It can be used in a neural interface because it is relatively inexpensive and quite simple. I put on a cap, twirled the electrodes, lubricated them with paste - and they worked. An EEG installation - an electroencephalograph - is also a hat that is placed on a person’s head. It is needed to press the electrodes more tightly to the person’s head and position them correctly.
Why would neural interfaces be embedded in the brain
Electroencephalographs are different.They write signals with different resolutions and havedifferent number of channels. EEGs also come in two types. Basically, there are invasive and non-invasive EEG. An invasive EEG involves surgery to open your skull and place electrodes directly on the surface of your brain. It is clear that this technology is not used in humans. It is mainly used in animal science to study their activity, which is very similar to human activity. The EEG signal, which is read directly from the surface of the cortex and a very small group of neurons, has better performance than non-invasive technology. Because it is free from various artifacts that are characteristic of EEG.
There are areas in which invasive interfaces are very important and in fact the only technology that can alleviate a person's condition.For example, this is epilepsy.Our laboratory studies a type of epilepsy called absence epilepsy. This is epilepsy that is not associated with seizures. A person does not experience convulsions during an attack, he simply disconnects from the world for a while. And he doesn't know about it, that he's disconnected. This condition is also very dangerous, despite the fact that there are no external manifestations other than the fact that the person has disconnected from reality. Because he may, for example, not know that he has absence epilepsy and pass out while driving. Therefore, preventing all attacks is also very important, and the interface can prevent it. When a person has an attack, firstly, different brain rhythms are synchronized - this is very easy to catch. In addition, an attack is preceded by a certain type of activity - some time before the attack we can determine what will happen next. When the neural interface detects this, it can send impulses to the brain that help suppress this attack. At the same time, a person who wears this interface on a regular basis will not feel anything and will not know that he was just about to have an attack.
Neuralink Ilona Mask is different from othersinvasive neurointerfaces in that the electrodes that write the EEG are located on very thin filaments (thinner than a human hair), and a lot of electrodes are located on one filament.It is possible to record up to 3 thousand electrodes.This is a huge amount of data. Given that they are located on the cortex, they are very small and write a signal from a small group of neurons. This way we can catch very subtle activity.
Mask's technology partially solves the problem that is common when using non-invasive EEG - field spread (from the English "electric field strength" - "Hi-tech").When we write a non-invasive EEG, we superimposeelectrode on the head, and it records from a large group of neurons. It generates a general energy field. Electric fields from different neurons overlap and interfere with each other. So when we write a certain type of activity, we actually have many different activities, which interferes with processing. And a very big problem with all invasive interfaces is that our body does not want foreign objects shoved into it.
A non-invasive EEG is an EEG that we use on an ongoing basis.Essentially, these are electrodes placed onsurface of the head with a cap. Each electrode records the activity of a specific group of neurons. The location of the electrodes is very important because it greatly influences the processing result.
An electroencephalogram is an electrical signal.Processing this signal is a task of radiophysics.We can use conventional analysis techniques to isolate features of this signal. Our task is to highlight certain patterns that correspond to this movement.
The main difficulties in creating neural interfaces
Why is it so difficult to create a neural interface?Because these are very difficult tasks:process EEG and identify patterns. They require an integrated and careful approach. Firstly, because each person is individual, just like his brain. Even homozygous twins have very different brains, like fingerprints. We cannot create one universal system that works the same for all people. It must be adjusted, calibrated and constantly changed for everyone. The next reason is the specifics of the task. Each person can perform the same task differently and accumulate different activities. The placement of sensors is also a human factor. In different experiments, the sensors may be positioned slightly differently, but this is also very important, because they will write a different activity. Another reason is the non-stationary dynamics of the brain on all time scales. This means that it is constantly changing on all time scales. We can take a large piece of EEG signal, break it down into seconds, and each second will be different. Brains are constantly changing, evolving and adapting to their external environment. An important problem is the high level of noise in EEG recordings and artifacts. Because we place electrodes on the surface of the scalp, and underneath, between the skin and the skull, there are muscle fibers that work, move and create some kind of motor artifacts on the EEG signal. In addition, we constantly move our eyes, blink, and all this also gives guidance to the EEG, which is visible not only on the channels on the forehead, but also on the back of the head. Besides, our heart beats. Cardiac artifacts are present on the EEG in almost all people. They arise from the pulsation of blood in the vessels under the skin.
There are cool methods that help us get rid of all the artifacts.You can remove artifacts or filter them.There are methods that allow you to remove components corresponding to heartbeat, oculomotor artifacts, and filter the EEG signal in the range we need. Next, time-frequency analysis is used as a signal analysis method. We evaluate the energy of each rhythm and see which one contains some dynamics associated with the effect we are looking for. Time-frequency analysis includes wavelets, Fourier analysis, etc.
The next step is to establish functional connections between different areas of the brain.When we do something (we move our hands,imagine hand movements), our brain regions interact either synchronously or desynchronously. They influence each other during this interaction. Based on these influences, it is possible to build a network of functional connections that describes which connections are strengthened and which are weakened, and based on this, we can draw a conclusion about how our activity is built and processed by neurons on a scale of the entire brain. Artificial neural networks are the most promising here because they work quickly and are based on a biological model of neurons.
The EEG signal is a complex structure with many components, but its complexity changes over time and depending on what we do.From the dynamics of this complexity we can understand thathappens, evaluate it, classify it and make some kind of classifier on this basis. A specific example is the removal of artifacts. Let's say we have a multichannel EEG signal, and it shows artifacts that give interference on almost all channels. And they look like oculomotor artifacts. They are removed using the independent components method. Each signal is decomposed into several components that contain their own information. One of them contains something we don’t need in the signal—that very oculomotor artifact. We take it and throw it away. We get a clean, more informative signal containing what we need. Filtration is a basic thing.
One signal can be filtered in different frequency ranges.Alpha rhythm (8–14 Hz) and beta rhythm (15–30 Hz) -these are the two brain rhythms, the two frequency components, that contain the most information about how the brain processes motor activity. If we want to classify imaginary or real movements, we turn to the alpha and beta rhythm and look for a pattern on them.
Finding a pattern is a classification task.To look at the pattern we canuse the time-frequency analysis method. The pattern is motor activity on the EEG, and the most commonly cited is event-related desynchronization. We take the activity of the alpha rhythm. It contains a lot of information about how we move. This information is as follows: when we make a movement, the number of neurons that are involved in processing information decreases. Thus, the alpha rhythm is suppressed. Its amplitude decreases and it generates less energy. This suppression of the alpha rhythm continues throughout the entire movement. If we see that the alpha rhythm has dropped, we can say that movement has occurred at that moment. For example, a person clenches his hand into a fist. The alpha rhythm falls, its energy decreases. He's desynchronizing and we can see it. Then the person unclenches his hand, and the alpha rhythm is restored and returns to its normal state. The whole problem is that desynchronization is clearly visible in averaged data. We take several movements, calculate the wavelet surface for each, average them, and get a bright picture. We take one movement, count its wavelet surface and see nothing. We don't understand anything because this movement is a very subtle effect.
See also:
See what summer looks like on Saturn
There was a photo of how desert dust "flows" into the Atlantic Ocean
See what Hubble's successor telescope can see in space