Before diving into more profound knowledge about BCI, we should first understand what exactly are we decoding. In other words, what are the brain signals that we are decoding? Having this question, the first thing we should ask is “what is a brain? Is there anything inside?” The answer might have already been printed in the textbook you read during elementary school. The human brain was consist of billions of neurons. To be exact, the human brain contains about 86 billion neurons, moreover, there are glial cells, endothelial cells, and other chemicals that allow neurons to communicate.
However, to be honest, we know very little about the brain. First of all, for neurons, there is much different morphology inside the brain, some may even use different neural transmitters. Scientists hadn’t really categorized each of them. Furthermore, the connections between the neurons are very complicated, they are tangles and are in 3 dimensions; the amount of neurons is huge, and mapping out the whole brain connection may cost infinite time, by the way, the connections are always changing, at least that’s what we think how people learn so far. Even worse, we did not even have clear definitions of thought, free will, and many other psychological concepts. You might feel despair about brain science now, but the good news is we do know something about the brain. That’s why we start the discussion here. I don’t want to make this article rigid, so I will suggest a book called “Principle of Neural Science”. If you want to know more serious knowledge, you should definitely read it. ( Although I haven’t finished it either.
I have been thinking about how to start this conversation, but the brain is really complicated. It is really hard to start without making things too complicated. Maybe we will just start with some useful Cortex to BCI.
Cortex
The cortex is the surface of the brain which contains a lot of neurons, most people believe that is also where most neural signals are coming from
Motor Cortex
It is the cortex that controls voluntary motor behaviors. It locates around the top of the brain between the frontal lobe and the parietal lobe. Noted that it is “voluntary” motor behavior, for example, holding a water cup or tilting your head. Many BCI systems use the signal in this area to decode the movement and translate them to machine control. The good thing is it is at the surface of the brain and it contains large pyramidal neuron cells which are both good for signal detection. Also, since it controls motor behavior, we are able to link that behavior to machine behavior, it’s easier this way for people to learn and understand how to control the machine according to their will.
the orange part is the motor cortex, and the blue part is the sensory cortex
Visual Cortex
It is the cortex that receives and preprocesses visual signals. It locates around the back of your head called the occipital lobe. The good thing about the signal here is that the signal is strong and could be detected by other machines easily. The bad thing is that it is not voluntary, you only receive the signal whenever a designed visual signal is performed, which makes sense, because you don’t see things just because you want to see it
There are of course other cortexes like the auditory cortex, which uses voice as a stimulating signal, which is pretty similar to the visual cortex, and of course, it is also a passive signal
Neural Signals
Let’s talk about the most precise way to acquire neural signals. Neural signals are very weak, which is around 100 mV. If you want a very precise way to measure it, you will have to use a technique called “patch clamp“. The patch clamp allows people to record as precisely as only one ion channel across the membrane of the cell. There are 6 different ways of patch clamps. To avoid a long paragraph, I will just put its Wikipedia page here. Take a look if you have time.
However, this method only applies to one single neuron, and it has to be done in the lab. After the patch clamp, the cell is basically dead, so it cannot be applied to BCI technology. Other than this method, most recording methods can only record the signal from the outside of the cell. There is one problem with that, which is the high resistance between the cell and the probe. Because remember, there are cerebral spinal fluids inside the brain. The probe has to be as close as 10 μm (the diameter of a hair is around 60~80μm) to be able to get the signal. Of course, different materials would have different distances, but it is going to be around the range.
So the next step is to decode the neural signal. Neurons are amazing cells, you may think, well, it’s not too hard to decode, you just have to show the pictures again and again. Based on the excitation of the neurons, we know that neurons 1, 2, 3, 4, and 5 are the neurons related to this picture. Maybe it is going to take a long time, but at the end of the day, we will be able to link each thing with different neurons, there are “only” 86 billion neurons. We will finish that someday in the future. When we finish mapping them, when we trigger 1, 2, 3, 4, and 5 neurons, we can see the same picture again, and maybe 6, 7, 8, and 9, we can see another picture. Life is easy.
So, why can’t we? First of all, 86 billion is not a small number, there are only 8 billion people in the world, and it is this complicated already. Not to mention the connections between them. A neuron can have multiple connections with other neurons, they are tangled and tiny. Plus their keep-changing characteristic and different neural transmitters. It doesn’t matter how much time you spend, you are never going to map out the whole brain if you do it one by one.
The worst reality is that the excitation of a neuron is not a constant, it is by chance. This means every time you show the same picture, not all the same neurons would be triggered. The probability of triggering is close to a distribution called Poisson Distribution. The same picture only increases the probability of triggering but is not guaranteed.
Well, I guess we don’t have to do research about it then, the amount of neurons is huge and the excitation is random, there’s no way we are able to understand the human brain. Not exactly, we can stack multiple trials and see which neurons got triggered the most, and decode the signal based on that. Furthermore, is it important to know if that one neuron is triggered or not when seeing the picture? Not really, most of the time, it is more important to know if a group of neurons was triggered when stimulation is presented, which is also easier to be detected. So there we go, we survive in this narrow reality, one neuron is not able to represent anything; too many neurons are too complicated to decode. Research is still finding the optimal amount of neurons to represent an event, let’s wait until we figure it out.
Signal Processing
Signal processing may be the most important part of BCI. It is related to probability, statistics, mathematics, information theory, and sometimes machine learning. It is somewhat difficult, and it is hard to include them in this post, I’ll wait until someone asks. In short, signal processing usually requires filtering, stacking the signal to get rid of the noise, and then putting the processed signal into a system to judge what is the status of the brain. The machine was controlled based on the decoded information
We’ll stop the discussion here. The information may be complicated and disordered, but my main purpose is to get your attention about BCI. If you have any questions or suggestions, you are welcome to leave a comment, and I will reply in a timely manner. Thank you!
Comments