What are Brain-Computer Interfaces?
Have you watched Black Mirror’s Playtest episode? A man volunteers to test a game company’s BCI horror game. The game is special because the BCI can read the player’s mind and create the scariest game for the person. Also, the game is just an illusion, created by the signals sent by the BCI. In the end, the game becomes so realistic that the man confuses about what is reality. And now if we come back to our reality: What are BCIs and what are they capable of doing?
Let's jump back to the 1920s, Hans Berger discovered the electrical activity of the brain and developed electroencephalography(EEG). EEG collected the user’s electrical signals of brain activity and analyzed the brain waves. Berger first did an experiment on a 17-year-old boy. He used a silver wire and taped it under the scalp of his patient. This was the first non-invasive BCI in history. He could only analyze the repetitive patterns of the brain activity -in other words, the oscillatory activity-. He could understand the alpha and beta waves of the brain (showing mental activity) and observed that the brain waves were different around tumors.
In 1973, Jacques J. Vidal (a researcher at UCLA) published a research paper called Toward Direct Brain-Computer Communication -which is the first peer-reviewed paper about BCIs-. The paper demonstrated using brain signal detectors we can read brainwaves and by making some equations we can understand them(like emotions and movement thoughts). This technology would be really helpful for neurophysiology. Vidal has also included that identification of the brain states can benefit the improvement of computer-assisted learning programs. Vidal’s experiments were deeper and included more technology.
In 1988, Stevo Bozinovskia and Liljana Bozinovska have made the first EEG-controlled robot. They used an EEG to get brainwaves, observed the alpha waves, and created a CαV pattern recognition. And connected the brain data to the robot.
Still, what are BCIs?
Brain-computer interfaces mainly collect data from the brain, encode the data, and deploy it. We can do this process because the brain sends signals by electricity so we can collect that electric data(as Berger explained in his research). The structure is the same for every BCI model but there are some different types of BCIs. We can categorize BCIs by 3 factors: dependability, type of recording data, and mode of operation.
Categorizing BCIs by dependability:
There are dependent and independent BCIs. Dependent BCIs are connected with the neuromuscular system. When the BCI is tracking your brainwaves (ideas and emotions) it also tracks the signals you are sending to your muscles (central nervous system (CNS) output). For example, when you are looking at something it is also tracking which way your eye is looking. On the other hand, the independent BCIs do not track the natural CNS outputs (the neuromuscular signals), so it is more focused on what you are thinking or feeling, rather than the physical components.
Categorizing BCIs by the types of recording data:
There are two main types of recording data: Invasive and Non-invasive.
Invasive BCI utilizes physical implants placed in the brain to capture and relay neural signals. They are typically used to control external devices, such as robotic arms, prosthetic legs, and other assistive technology.
An example of an invasive BCI is BrainGate’s BCIs. BrainGate works on providing solutions to people with neurologic diseases, disorders, and injuries. They use a tiny array of electrodes implanted in the brain to detect electrical activity and translate it into commands. Their systems can allow people with severe neurological conditions to control devices with their thoughts. For example, someone with a spinal cord injury can use the BrainGate system to control a wheelchair or a robotic arm.
Invasive BCIs have the potential to revolutionize the way people with disabilities interact with technology and their environment. They allow for more precise control and higher-fidelity communication than non-invasive alternatives, making them more useful for applications such as controlling assistive devices or restoring lost motor functions.
The advantage of invasive BCIs is that they allow for more direct control and feedback than non-invasive BCIs.
There are two types of invasive BCI: Intracortical and Electrocorticography(ECoG).
Intracortical BCI, plant intracortical microelectrodes into the patient's brain.So that the collected data is much more accurate. It is usually used when the BCI is for treating a neural disease.
ECoG is placed on top of the cortex. ECoG can detect the electrical activity of neurons in the cortex and is often used to understand the underlying mechanisms of motor control, speech, and other cognitive functions.
While Intracortical BCI uses implanted electrodes, ECoG uses electrodes placed on the surface of the brain. Additionally, Intracortical BCI can measure signals from a much smaller area of the brain, while ECoG BCI can measure signals from a larger area.
On the other hand, non-invasive BCIs are external systems that use electroencephalography (EEG) to measure electrical activity in the brain and translate these signals into commands for the computer. These devices can be used to control a variety of applications, from playing video games to controlling robotic devices. By interpreting brain signals, non-invasive BCIs allow users to control computers and other devices without any physical movement or contact.
As the main purpose of BCIs is to analyze brain data we can fMRI and MEG as kinds of non-invasive BCI. In fMRI, the brain is put in a multi-magnetic field and small changes in blood flow are observed. In MEG there is again a magnetic field and small electrical changes in the brain are observed. Even though these machines are used to scan brain data, whether to consider these as BCIs or not is open to discussion.
The most common non-invasive BCI is EEG(electroencephalogram). Apart from others, EEG is not a big device. You place small electrodes on top of your head and it scans the electrical changes in your brain. It is not as accurate as invasive BCI but is much more basic and easier to implement than others.
We also divide EEG into two: evoked potential and spontaneous. Evoked potential EEG focuses on a certain circuit of the brain or the spinal cord, while, spontaneous EEG looks at the general oscillations.
Evoked potential EEG involves stimulating the brain with auditory, visual, or somatosensory stimuli, such as clicking sounds or images, and then recording the brain’s response with the EEG. This technique can be used to assess the electrical activity of the brain in response to various stimuli and can provide valuable insight into the function of the brain. It can be used to test the brain to mark cognitive dysfunction(P300).
Spontaneous EEG is a technique used to measure electrical activity in the brain. It is used to monitor brain activity in patients, diagnose neurological disorders, and identify areas of the brain that are associated with cognitive functions. Spontaneous EEG can detect subtle changes in brain activity that occur in response to various stimuli, including changes in environment, emotions, and behaviors. It can also be used to measure the effectiveness of treatments or medications. Spontaneous EEG is a non-invasive technique that is relatively easy to perform and can be used in both research and clinical settings. ADHD treatment (SCP neurofeedback), creating a path through which the brain interacts with the external environment (motor imaging EEG), and targeting depression, anxiety, apathy, and sleep disorders (non-motor EEG) are some use cases of Spontaneous EEG.
Categorizing BCIs by mode of operation:
We categorize BCIs as asynchronous and synchronous. Brain activity that occurs naturally is characterized by rhythmic patterns called oscillations. In contrast, a cue-based brain-computer interface (BCI) system, known as synchronous BCI, operates within a pre-determined time frame, while an asynchronous BCI, operates independently and at the user’s own pace, not requiring a cue stimulus.
Synchronous BCIs typically use a specific cue or task to elicit a certain brain activity, such as flashing light or playing a sound. The user is required to respond to the cue in a specific way, such as by focusing on the light or sound, or by performing a specific mental task. These systems are often used in research and require a high degree of user training. An example of a synchronous BCI is a P300 speller, which uses a matrix of letters that flash rapidly. The user is asked to focus on the letter they want to spell, and the system detects a specific brain activity called the P300 response, which occurs when the user recognizes the target letter.
On the other hand, asynchronous BCIs operate independently of any specific cue or task. They allow the user to initiate the interaction with the system at any time, without requiring any specific response or mental task. These systems are often used in real-world applications, such as communication devices for people with severe motor impairments. An example of an asynchronous BCI is an SSVEP-based system, which uses flickering lights to elicit specific brain activity. The user simply looks at the light that corresponds to the desired action, such as typing a letter or clicking a mouse button.
How do we understand brainwaves?
We can understand brainwaves and categorize the oscillations by analyzing the patterns. Here you can see the flowchart I’ve created to show the ML algorithm which is used to analyze the oscillations:
There are 5 brainwave data we end up with at the end of the process: gamma(when there is an active thought), beta(when the patient is working), alpha(when the patient is relaxed), theta(when the patient is drowsy), and delta(when the patient is sleepy). We can also target muscle movements using dependent BCIs.
Spinal and Temporal Resolution:
We can compare invasive and non-invasive BCI according to their spinal and temporal resolution as well.
Spatial resolution refers to the level of detail at which a BCI can identify and locate neural activity within the brain. The higher the spatial resolution, the more specific the location of neural activity that can be detected.
Temporal resolution refers to the level of detail at which a BCI can identify and track changes in neural activity over time. The higher the temporal resolution, the more precise the BCI can be in detecting changes in neural activity.
Invasive BCIs have higher spatial and temporal resolutions than non-invasive BCIs because they use electrodes that are implanted directly into the brain. These electrodes can measure neural activity from specific regions of the brain with high accuracy. For example, Intracortical BCIs have a very high spatial resolution because they can measure signals from a small area of the brain. ECoG also has a high spatial resolution as it is placed on top of the cortex and can detect the electrical activity of neurons in the cortex and is often used to understand the underlying mechanisms of motor control, speech, and other cognitive functions.
On the other hand, non-invasive BCIs have lower spatial and temporal resolutions than invasive BCIs because they use electrodes that are placed on the scalp rather than within the brain. These electrodes can measure neural activity from larger regions of the brain, but with less accuracy than invasive BCIs. For example, EEG, which is a non-invasive BCI, has a lower spatial resolution than invasive BCIs, as it can only measure neural activity from the cerebral cortex, rather than specific regions of the brain.
Use of AI and ML:
Artificial intelligence and machine learning are playing an increasingly important role in the development of BCIs. AI and ML algorithms are used to analyze and interpret the electrical signals that are generated by the brain and translate them into useful information or commands.
One of the key ways in which AI and ML are used in BCIs is in the development of decoding algorithms. These algorithms are used to analyze the electrical signals generated by the brain and determine what the individual is thinking or trying to communicate. This can be done by training machine learning models on large sets of data, such as EEG recordings, to identify patterns in brain activity that are associated with specific actions or thoughts. The models can then be used to interpret new data and predict what the individual is trying to communicate or control.
Another way in which AI and ML are used in BCIs is in the development of control algorithms. These algorithms are used to translate the individual’s brain activity into commands that can be used to control external devices, such as prosthetic limbs or assistive technology. For example, a person with paralysis could use a BCI to control a robotic arm by sending neural signals associated with the intention to move the arm. These signals are then analyzed by the AI and ML algorithms, and the robotic arm will move accordingly.
AI and ML are also used in the development of adaptive BCIs, which can change and adapt to the individual’s needs over time. For example, an adaptive BCI system can learn from the user’s brain activity and adjust the sensitivity of the system to better match the individual’s neural signals. This allows for more accurate and efficient communication and control.
Overall, AI and ML play a crucial role in the development of BCIs, enabling these devices to be more accurate, responsive, and adaptable to the individual’s needs. With continued advancements in AI and ML, it is likely that BCIs will become an increasingly important tool for individuals with disabilities, as well as for a wide range of other applications.
What potential does BCIs hold?
BCIs have the potential to revolutionize the way we interact with technology by allowing people to directly control devices with their thoughts. This technology has the potential to greatly benefit people with disabilities, such as those with paralysis or ALS, by providing them with a means of communication and control over their environment. BCIs can also be used in a variety of other fields, such as gaming, virtual reality, and robotics. BCIs have the potential to be used as a diagnostic and research tool in the fields of neuroscience and psychology, providing new insights into brain function and helping to develop new treatments for neurological conditions. BCIs can also help with concentration, by analyzing when the person is working and helping the patient concentrate when they are distracted. BCIs can also be used to analyze if the person is lying or not.
One example of a potential application for brain-computer interfaces is in the field of neuroprosthetics. A neuroprosthetic is a device that can be implanted in the brain and used to control a prosthetic limb or other assistive technology. For example, a person with paralysis could use a neuroprosthetic device to control a robotic arm, allowing them to perform tasks such as eating or writing that would otherwise be impossible. The BCI would interpret the neural signals associated with the intention to move the arm and translate them into commands for the robotic limb.
Current Companies:
BrainCo: BrainCo is working on BCIs to help people be focused. FocusCalm is one of their products that help with depression and anxiety. Their BCI design is user-friendly and connected with a mobile application that you can download to your phone.
OpenBCI: OpenBCI provides open-source hardware and software tools for brain-computer interface research and development. Their goal is the make BCIs more accessible and ethically safe (by protecting users’ mental health and personal data).
Neuralink: Neuralink works on using BCIs to communicate with the brain, understand it better, and heal neural disorders.
Limitations:
Currently, BCIs have several limitations and challenges that need to be addressed. One of the main limitations is the accuracy and reliability of the technology. BCIs rely on interpreting brain signals, which can be difficult to accurately and consistently decode the intended actions or thoughts of the user. Invasive BCIs can have serious risks and complications and require a high level of technical expertise. Additionally, the cost of developing, manufacturing, and implementing BCIs can be high, making them less accessible to many individuals and organizations. BCIs also raises a number of ethical concerns such as privacy, security, and autonomy. For example, the use of BCIs in surveillance or mind-reading applications raises serious privacy concerns. BCIs are currently used by a limited number of individuals, mainly those with severe physical disabilities, and are not yet widely available to the general population. Furthermore, BCIs require a significant amount of training for the user to be able to control the device effectively which can be time-consuming and difficult for some users. While BCIs have a lot of potential for different applications, current technology is limited in terms of the number of applications that it can be used for. Additionally, external factors such as noise, movement, and electromagnetic interference can affect the accuracy of BCIs. This technology holds great potential but currently, it is still improving and getting ready to be integrated into our daily lives.
