Visiting Naked Science, Mikhail Lebedev, a graduate of the Moscow Institute of Physics and Technology, senior researcher at the Center for Neuroengineering at Duke University and one of the main "creators of chimeras", who told us how these strange works can be useful - and how to create an army of rats obedient to the will of monkey commanders …
Perhaps one of the most important recent scientific developments was the publication of a group of American scientists from Duke University who "connected" the brains of monkeys. The laboratory under the guidance of a native of Brazil, Professor Miguel Nicolesis, arranges such sensations with enviable regularity. By introducing microelectrodes in close proximity to the brain neurons of experimental animals, the experimenters connect them with each other, with additional sensors and computers, creating a kind of "brain chimeras". Rats seeing in infrared light and exchanging thoughts, a team of monkeys controlling a virtual hand.
- All such work originates from Galvani, who for the first time caused the muscles of a frog to contract with weak discharges of electricity. Well, those studies that immediately preceded ours began much later, around the 1960s, when Edward Evarts, using a microelectrode, first began to record the activity of individual neurons in the brain of a monkey.
For its time, this was a huge breakthrough: an animal with a microelectrode embedded in the brain remained completely normal and was awake while scientists were recording neural impulses. Time after time, the monkey could be given tasks for various types of activity, to study the responses of certain neurons, and therefore to establish their function. So cells were found whose work is associated with movement, with the processing of visual information, etc.
This, apparently, was largely random work? After all, then it was impossible to place a microelectrode in a strictly defined, predetermined neuron …
- Of course. The introduction of a microelectrode is like fishing: it is difficult to predict what kind of prey will be caught this time. Although, in general terms, of course, we can make certain predictions by working with neurons in a particular area of the brain.
This single-cell recording experiment uses a microelectrode implanted in the monkey's visual system to track the electrical activity of an individual neuron.
However, one should not forget that there are many billions of neurons in the brain, and in order to truly understand its work, it would be worthwhile to cover all its cells with these observations at once. Of course, this task is completely unattainable even now. But today we can track hundreds, and sometimes even thousands of neurons in parallel.
And what about the uniqueness of each neuron? After all, the brain is an extremely plastic, flexible system; it is formed in the process of individual development. If in one animal you place an electrode in a precisely defined area, this does not mean that another will have the same neuron in it with exactly the same functions and connections …
- This is certainly true, and in this sense, each record is unique. But since we observe several animals, we get some average picture and find important general trends. In addition, with rare exceptions, such studies do not concern higher nervous functions, but are aimed at relatively "simple" tasks, such as the presentation of sensory signals or movement control.For the structures involved in these tasks, the structure is more or less the same even for different animals.
The task of multichannel registration of a large number of neurons rests on some technical issues: it is relatively easy to introduce a hundred microelectrodes into the brain, but it requires sufficiently developed electronics and computer technology to record, transmit and process signals in real time. Therefore, it is in the last decade, when progress in this area has been especially rapid, that a real heyday has come in the field of multi-electrode leads.
The paralyzed patient uses her mind to control a robotic arm created by a team led by Andrew Schwartz.
Parallel to this, a new and promising direction has emerged as the brain-machine interface. The idea here is that if we can record simultaneously a sufficiently large number of neurons and process this data in real time, then we can turn them into control signals, for example, for an artificial limb - a prosthesis, for a speech generator or an exoskeleton. This work also began about ten years ago, and today new results are published almost every week.
For some time now, such experiments have been conducted on human volunteers. Sometimes these prostheses are connected to the spinal motor neurons of patients by recording an electromyogram, and sometimes directly to the brain, it all depends on the specific case. For example, to control the exoskeleton, the best option may be to register myoelectric signals from the muscles of the arms, trunk and neck - there is no need to “dig” into the signals of the cerebral cortex. In clinical practice, a case is known when a person lost both arms as a result of burns from electric shock. Then neurosurgeons sewed his torn nerves in his arms into the pectoralis major muscles and sent an electromyogram of these muscles to the prosthetic arms - and the patient moved both prostheses well with them.
But the work of the groups of John Donahue and Andrew Schwartz is in many ways close to ours: they managed to remove signals directly from the human brain and transmit them to robotic prostheses. So far, the quality of these signals is not very ideal, therefore, for example, Donahue and his colleagues used an extremely complex robotic system capable of independently coordinating their movements, and a person only gave general, "strategic" commands.
Such invasive registration of brain activity has not yet entered widespread medical practice, but I think this is a question for the next ten years. In any case, it is absolutely clear that the method works, that it can be improved quickly and qualitatively. After all, the cortex is the largest part of the human brain, and almost any signal can be removed from it. Moreover, she is surprisingly flexible, quickly adapts to the performance of a variety of tasks, even initially unfamiliar to her.
And if we talk about receiving signals going in the opposite direction? In what areas and how can sensory information be "introduced" instead of signals from the senses, lost or damaged?
- First, sensory nerves can be stimulated if they are preserved. The most successful example in this area is cochlear hearing restoration. The number of such devices already implanted in patients today reaches about 200 thousand. If the auditory nerve is preserved to one degree or another, electrodes are simply connected to it, coming from microphones responsible for different tones of sound, from low to high. As a result, people can hear words, recognize others by their voice, and so on.
Another option, which, in my opinion, will become widespread in the near future, are interfaces for vision restoration.For example, a degenerated retina can be replaced with an artificial one and try to transmit signals from it to perceiving neurons, or directly stimulate the optic nerve. It will be more difficult in the event of the death of the entire optic nerve: then it will already be necessary to stimulate the cerebral cortex directly. But such work is already underway. The areas of the visual cortex receiving signals from different parts of the retina have been well mapped, and it has been shown that they can be stimulated with electrodes. So far, only illusory flashes can be created in this way, but they can also be used to convey letters and even simple visual images. I believe that in the future the quality of transmission of such information will increase.
How useful are the works of your team for these practical tasks?
- I must say that the head of our laboratory, Professor Miguel Nicolesis, began his work in this area with rats. Back in the 1990s, he set the task of recording information from different areas of the somatosensory cortex, covering it all: both the cerebral cortex and the thalamus (a small area in which the primary processing of all sensory information takes place, except for signals from the olfactory organs, after which it distributed to the rest of the brain - NS), etc. At the same time, information was removed from peripheral sensory neurons - so as a result, Miguel received very interesting results that seriously pushed this whole area of research.
Then it was decided to move on to work on another model, on monkeys. These animals are incomparably smarter, allowing to obtain more interesting results and more adequately develop research in the direction of brain-machine interfaces. From about that moment, somewhere in 2002, I also joined this work. We studied the monkey brain signals that control the movements of the arms and legs when walking … Soon we developed the topic towards feedback interfaces, in which the artificial hand not only obeyed commands from the brain, but also transmitted certain signals back to the brain. For example, she felt various objects, communicating certain tactile characteristics.
Almost every one of our articles was cutting edge, all of these experiments were carried out for the first time, and after the publication, colleagues from laboratories around the world took up the work, finding different ways of practical development and application of these results. After all, we still work on monkeys, so our research has no direct access to clinical practice, it is necessary to take several more important steps. Nevertheless, all subsequent experiments on humans, including those with sensory feedback, relied on our results.
“It seems like one of the keywords you can use to describe your research is“extravagant”. What experiments of your group would you call the most extravagant, the most unexpected and loud?
- Our field is developing extremely quickly, so even those experiments that today look quite obvious and no longer attract wide attention, for their time were both loud and extravagant. For example, an employee of our laboratory, Eric Thompson, carried out interesting experiments on the direct connection of neurons in the sensory cortex of the rat's brain to an infrared sensor. As a result, the animal acquired "superpower" - the ability to see light in the infrared part of the spectrum that is new to it.
A lot of noise was made by the work, during which we implanted microelectrodes into the monkey's brain, which allowed it to control the robot literally with the power of thought. The animal grabbed an object (for example, a banana) with its manipulator hand, although it could be in another state … And two years ago, Miguel Pais-Viera, who was in the United States, and Carolina Kuniki, who was in Brazil, connected the brains of two rats at a distance of many thousands kilometers from each other.The American rat received sensory signals from the Brazilian brain and reacted as if it were "reading" her thoughts.
I must say that in nature, of course, no one ever stimulates a rat with electrodes, and for her this is a very unusual experience. For the brain, this is not at all the same as the signals coming from its own nervous system. He needs to learn to perceive and interpret these “voices in the head”. Therefore, rodents are previously trained to learn to recognize such stimulation.
The setting of such experiments looks really extravagant. But when you plan them, what do you focus on more - looking for serious scientific results or trying to do something like that, something amazing that no one else has ever done?
- There is both. We are trying to both get new data and try something amazing. Although a variety of factors always play a role. For example, when we connected a monkey with a robot in Japan, we needed the most perfect walking robot that our Japanese partners could find. We had the most perfect brain-machine interface at that time, they had the best walking robot … That's why they connected so successfully.
The example of this work shows the practical orientation of our experiments. After all, paralysis of the lower extremities is a fairly common medical problem that can develop as a result of spinal cord injuries. Raising such people from wheelchairs is a very urgent task, the solution of which will be helped, among other things, by our "extravagant" experiment with a walking robot and a monkey's brain.
So, if you look closely, in all our works, despite all their sensationalism, these two components are always present: firstly, the study of the fundamental mechanisms of the brain and the nervous system as a whole, and secondly, a completely practical, medical orientation, associated with the creation of interfaces "brain-machine" and "brain-brain".
What fundamental findings did you manage to make?
- In this sense, it is important to highlight one more point. Neurophysiological studies of walking (as a rule, experiments are carried out on four-legged animals) have been going on for more than a hundred years, and traditionally the main efforts in this area are focused precisely on the spinal cord, on the pacemaker that coordinates the corresponding movements of both legs and the whole body. For professionals who are working on this problem, the very formulation of the question - control of walking at the expense of only the brain, bypassing the spinal cord - at that time seemed almost heretical. But we managed to prove that this is not only possible, but also fully realizable. This was an important demonstration in support of the new approach.
Before that experiment, it was not at all clear to what extent the cerebral cortex was associated with the movements of the legs when walking. We found out what happens in this case both in the motor and in the somatosensory cortex - no one has ever done anything like this before.
Of course, it was clear to us that, on the whole, everything corresponded to the classical concepts of sensory and motor “homunculi”. If you implant the electrode closer to the top of your head, you will be trapped in the neurons associated with the lower extremities. If you move to the side, the torso will be presented there, then the arms, face, mouth, and so on. To some extent, these views are valid - but only to some extent. Careful studies - including ours - have shown that any movement activates neurons in almost all parts and regions of the brain. You can move your arms, and at the same time, a certain excitement can be registered even in the areas associated with the movements of the legs … This is quite logical: after all, our body is a single whole, and even if you move your arm, the rest of the parts must compensate for this movement.
Moreover, one should not think that the brain is arranged in the manner of computer blocks: there is a motor cortex that controls movement, and there is a sensory cortex, which is occupied only with receiving and processing information from sensitive neurons. This is a simplified book description. In fact, when we record the activity of the brain, we do not see a big difference between the activity of the motor and sensory cortex. It is even difficult to distinguish them in the recording - the brain works as a single and dynamic whole. It cannot be said that, they say, “such and such a separate area of the brain felt such and such”: any activity embraces it entirely.
That is, if I now, sitting and not looking at my hand, slightly raise it, then activity occurs in the areas of the motor cortex associated with the legs, and even in the sensory cortex?.
- Of course. Even if your movement is automatic, it involves large areas of the brain and cerebral cortex. Your subconscious mind constantly creates a three-dimensional model of your body, its position in space and the position of its parts relative to each other (the so-called "body map"). The relevance of this model is constantly maintained, and this work involves many neurons from a wide variety of fields.
The brain does this in addition to your conscious will, including due to the work of the cortex. And although it is generally accepted that the cerebral cortex is necessarily some kind of conscious activity, some complex cognitive processes, in fact, the cortex is involved in a variety of processes, including basic, unconscious …
The brain remains active during sleep, especially during periods of dreams. When we sleep and experience something in a dream, excitement engulfs the corresponding parts of our brain, although in general it spreads very widely. Theoretically, this activity can be registered - and find out exactly what movements the sleeper dreams of. Moreover, the subject can simply sit and imagine the movement, say, with his hand, and we recognize it. In reality, the hand will remain motionless, but thoughts about this movement will be "visible".
The imagining brain and the acting brain work in a similar way, just in the first case, a certain block is triggered in the nervous system, an inhibitory mechanism that prevents this activity from being transmitted to motor neurons and muscles. It is not entirely clear where exactly this block is located and how it is arranged; this remains one of the most interesting problems of science. The motor signal can be recorded in the brain and even in the spinal cord at the level of interneurons, but the body parts will remain motionless.
Perhaps braking, preventing unnecessary movements, occurs little by little at all stages of signal transmission, so that when we only need to think about an action, but do not perform it, the signal simply does not reach the “final recipients”. Such distributed inhibition is the most probable hypothesis: in the nervous system, almost all processes are organized in exactly this way, distributed, without any "main buttons" and "switches".
Let's get back to the work of your laboratory. Medical applications are quite understandable and worthy … but for what purpose did you start connecting monkey brains "directly"?
- I have already said that we always try to be half a step ahead, to be pioneers in our field. And therefore, when the situation with the exchange of information between brains is ripe, we decided to act: sooner or later, someone had to do this kind of work.
Remember two years ago we connected the brains of two rats? Then one animal, based on its sensory information, solved a simple problem - it was the simplest choice: turn left or right. His decision was transmitted to the second rat, which did not receive any other signals, that is, the command "from the outside" was used. Each such successful interaction was accompanied by a reward, so that the system of two animals quickly improved and they conveyed information better and better.
Well, then we began to think about how to do a similar work on monkeys, and I suggested using a slightly different system, which I called “brain plus brain”. Information from multiple brains can be added together, allowing us to more efficiently read the desired signal and get rid of noise. This is how recent work began.
A total of three experiments were performed. In the first, two monkeys jointly controlled an artificial hand, the signals for which were simply the sum of the signals we read from them. Indeed, this allowed us to better capture information, eliminate noise - and, as a result, give more accurate commands to the virtual hand.
In the second experiment, each monkey controlled only one coordinate: the first monkey with the X coordinate, and the second with the Y coordinate. Here, the improvement in control was achieved due to the fact that the task for the individual participant was simplified. I must say that the monkeys quickly adapted to this non-standard situation, over time the brain completely switched over to control "its" coordinates, and together they controlled the manipulator even more efficiently than when simultaneously solving the same problem.
After making sure that the brain-plus-brain system works, we began the third experiment. For each of the participants in this game, the task, as in the second experiment, was simplified: the control was parallelized, so that one monkey controlled the movements of the hand along the X and Y axes, the second - X and Z, the third - Y and Z, and only by joint efforts they could adequately control the movements of the virtual hand in all three dimensions. Thus, each of the monkeys controlled movements in two of the three dimensions. None of them had a clue about the problem as a whole - each was doing something different - but together they solved it, and they did it very effectively.
By the way, if we imagine that sometime in the future such a "parallelization" of tasks will be possible for people, delegating to each a small and simple part of the problem, together they will be able to do what each individual brain is simply incapable of. Anything that is inaccessible to a person as such, as an individual, in principle.
Despite all the fantastic, I think that sooner or later our work will find application in humans. For example, the use of the brain-plus-brain interface for medical purposes can be predicted. Imagine the job of a physical therapist helping a patient to regain movement after an injury or other damage to the nervous system. He "catches" a person's movements, the awkward beginning of these movements, helping him to finish what he started … And if it was possible to "connect" the nervous systems of the therapist and the patient, this work would certainly be more effective.
The idea may seem absolutely fantastic - but tell me, are you going to combine different species? Let's say the same rats with the same monkeys
- By the way, yes, I have already suggested doing that. The rat is good “in the arena”, “in the field” - it is such an impeccable and well-controlled performer, especially if you additionally stimulate its “pleasure zone”, achieving the desired behavior.
In comparison with a rat, a monkey is a much smarter creature, and it is more difficult to control it: receiving signals from nowhere, it constantly tries to think what is wrong with it. But as a "manager" a monkey is much more effective than rats - in fact, a rat for it can become something like a cursor. But instead of a rat, you can use someone simpler, for example, fish. Perhaps someday such work will be done: theoretically, this way you can create at least a whole army of controlled rats …
… and take over the world?
- Well, the scope for imagination is almost endless here, it is not necessary to immediately capture it. You can, for example, "connect" - to a certain extent, of course - the brain of a pet and its owner, and return the dog that has run away home, or simply communicate with your pet directly.
However, you are not engaged in decoding and interpreting these signals and are not interested? It turns out that you only read information from one brain, extract a signal and transmit it to another brain or to a computer. And everything that happens in the brain itself remains, as it were, in a black box?
- In many ways, this is so: we remove the signal, we extract and decode it, but only at the level of correlations. That is, if we see that such and such a signal correlates with the movement of the hand in such and such a direction, we can state that such and such a neuron is discharged during such and such a movement. Such a correlation does not yet guarantee dependency. Therefore, we cannot say exactly what exactly this or that neuron does in the general system. But little by little our work brings this understanding closer. The box, of course, is black, besides, we are crawling into it deeper and deeper …