Brain computer interface (BCI) is one of the emerging technology fields. Elon Musk’s neural interface technology company, Neuralink, had demonstrated such technology using monkey as a tested subject. Here is the video:
Although there are some ethical issues raised recently about such kind of experiment, the novelty of the technology cannot be denied.
I have been following the development of BCI for a long time, and I also took a class about it during my master’s degree. I will bring you all to become more familiar with the field. Let’s dive right into it!
Hold on! Before we do so, let’s talk about the definition of BCI, otherwise the discussion will become a mess.
Table content
Definition
Brain computer interface, some people also called it brain machine interface (BMI), is a bridging system between biological brain and artificial devices. You may ask, is cellphones included in the definition? Since I use my hand to control cellphones, and my hand is controlled by my brain, so technically it is also a BCI technology. Broadly speaking, yes. However, when we talk about BCI, we are focusing on the system that directly interact with the brain and the device. In other words, the device is going to read the signal “directly” from central nervous system. And for central nervous system, we specifically said “brain” in brain computer interface, so the device interact with spinal cord is also excluded.
Purpose
Now we know the definition of BCI, the next step is to talk about its purpose. Every technology development has its purpose, and BCI is no exception. According to Neural Engineering, 3rd Edition, Springer, 2020, there are 5 main purposes.
Replace
Replace the function that ones already have. For example, use robotic arm to replace the original arm function.
Restore
Restore the lost functions of an individual. For example, stimulate muscles through the reading of brain signal to restore walking ability.
Enhance
Enhance body function by strengthen original ability. For example, make E-sport players focus longer on the game
Supplement
Gain additional body functions, such as having and additional arm
Improve
Stimulate the growth of neurons by learning how to control the device using brain signals. The lost neuron connection may be able to grow back and therefore improve the body conditions.
The primary point of listing the purposes is to make you understand what BCI can achieve. I don’t think it is important to actually distinguish between each functions
Other than these function, it is expected that the device can provide signal feedback to the brain. When you touch a thing, you feel the texture, the pressure and the temperature of it. That is because the neurons on your hand provide the feedback signal to your brain. We hope to also implement the function in the BCI system.
Structure
There is a certain structure for the development of BCI. I don’t know if it is going to change in the future, it most likely will, but knowing the structure is the first step to improve it.
Signal Acquisition
Acquiring neural signals from brain
Signal Processing
Since the acquired signal may be weak or be buried in the noise, the signal processing can be further divided into signal amplifying, signal filtering, feature extraction, and feature translation. There may be some machine learning algorithms applied in the process to improve the results. The purpose of this is to gain useful information from the recorded brain signals.
Control Interface
The interface would utilize the translated features to control the device.
Control Display/External Stimulation
Like what I said in the previous section, we want to provide a signal feedback to the user which allows them to adjust their behavior (or thoughts) based on the signal they received. Most commonly used method is a display screen as shown in the video above. Noted that, here, we specifically refer to the devices that only provides indirect feedback to the brain, meaning that the signal is received by other body parts and be passed down to the brain. There are other types of feedback mechanism to achieve the same goal, such as audio, or tastes, but all they want to do is the same, which is providing feedback to the brain for behavioral adjustment.
Device
This is just the device that you hope to control, such as drone, car, and robotic arm etc.
If all of the above have been implemented in the system, we call it an open loop system. The closed loop system would need an additional feature.
Feedback
The feedback function would provide “direct” feedback to the brain. Just like your hands, they provide feedback to your brain so you can feel the texture, pressure and temperature of an item. With the this, the close loop system can be establish and the so-called real brain computer interface can be realized.
Types
Brains are mostly protected by the skull in most of the mammals, and human is also the case. This results in the difficulties on acquiring clear neural signals from the brain. So far, there are many different types of BCI system, and this is because, my own opinion, people did not find a very good way on acquiring brain signals. There are always serious tradeoffs in the signal acquiring methods used nowadays. I will talk about them in detail in the future, for now, let me just simply divide them into two categories.
Invasive
Invasive technique acquires brain signals by electrodes implemented physically inside the brain. Surgeries are required to insert the electrodes. The advantage of invasive technique is that the signal is much cleaner than the non-invasive one. Also, the resolution is really high, it can even distinguish between different neurons nearby. However, as you probably thought about, the disadvantage is really obvious, it is invasive. The insertion of the electrodes damages the brain tissues and may cause serious infection after the surgery if it is not handled well. Furthermore, due to the invasiveness, the monitoring area is very limited, only electrodes inserted areas can be monitored. This is also the technique that Neuralink uses, and that’s the reason why it is very controversial.
Non-invasive
Non-invasive technique detects and translates some physiological responses created when brain is functioning. Most commonly seen ones utilize electrical phenomenon related signal created by the neurons in brain, such as EEG, MEG. On the other hand, the functioning part of the brain will need more nutritions and oxygen, which would be supplied by the blood flow. Therefore, the brain activities can also be decoded by monitoring the blood flow inside the brain. The advantage and disadvantage of noninvasive technique is completely the opposite of invasive techniques. Larger area of the brain can be explored and people are more likely to use it, which comes with larger data base. But, of course, noisy signal and low resolution are the main disadvantages of it. Also, only the neural signal on the brain surface can be recorded, deeper signals inside the brain will be filtered out by the tissues before reaching the electrodes placing on the outside of the skull. This may greatly decrease the value of decoded signals.
Challenges
Actually, there are so many challenges that it is difficult to list them all. I don’t want to go into too much detail in this article, but the greatest challenge is simply that we don’t know enough about the nerves, the brain, and how it generates and transmits signals. There is an interesting thing to think about. Originally, our neural signals were transmitted through a specific pathway by nerves, and received and translate by the targeted body part, but now, because we read the neural signal directly from brain, the signals must directly interact with the device. It may be a bit difficult to understand, let me give an example.
If we want to control our fingers, we will have a signal about moving our fingers created inside the brain, and then this signal is transmitted all the way to the fingers, at the end, the finger moves. We don’t know if the nerves in the middle are doing something else, or even a set of neural networks are translating the signals, so that the brain’s signal of moving the finger is converted into signals that can be understand by the muscle controlling the fingers. But today, if we want to use the brain-computer interface to move the fingers of a robotic arm, the neural signals created inside the brain will be the direct command to move the fingers on the robotic arm. We don’t know if there are differences between the two signals.
Last thoughts
Lastly, there is actually a very important point that you can think about. Currently, brain computer interface is actually not as good as people’s imagination. I personally think it is also not as good as what Elon Musk said in a brain chip presentation that it can access people’s memories in the near future. (I forgot what he said exactly, but that was most likely what he meant.)
What can be done now is actually implementing a set of actions or program-designed functions, and your brain is a middleman who decides which result the device will go to. In other words, your brain does not have a complete control over the device. Just like a remote control with only channel switching buttons, you can only go to the next channel no matter how you press it, you cannot increase the volume or jump to other channels. Of course, the remote control can also be designed to have volume keys, number keys or other keys, but that must be designed in advance. Interestingly, isn’t it the same with every part of the body? Your hands or feet can only move up, down, left and right. Perhaps the only thing that is truly free is your thoughts, but maybe even your thoughts are not free?
Thank you for reading, because it is just an introduction, so it may be a little boring. But I believe there are many aspects that are worth exploring and thinking. I will continue to write the related topics. If you have any ideas or opinions, please share and leave me a message.