Source of this article: AGI interface Author: Chen Guangjing

AI and brain-computer may be subverting human cognition of the brain and life.

This process is faster than we thought. We may see with our own eyes the picture depicted in the science fiction film "Robocop" in the 1980s: a policeman who was killed tragically by bandits with his arms broken on his first mission, was revived by a doctor, and obtained a steel and mechanical body, and jumped from a rookie in the police world to a super cop, not only invulnerable and powerful, but also able to use a precise positioning system to hit opponents from any tricky angle......

Nearly 40 years have passed, and a series of imaginations such as "cyber humans" and "brain resurrection" are accelerating into reality. The first to be deeply affected is neurosurgery, which is one of the world's top neurosurgeons who have applied AI tools to surgery and scientific research.

Recently, Tiger Sniff had an exclusive conversation with two of the world's top neurosurgeons and scholars, Peter Warnke and Sean P. Polster, from the University of Chicago School of Medicine.

Peter Warnke, MD, Professor of Surgery and Director of Stereotactic and Functional Neurosurgery at the University of Chicago School of Medicine.) He is a world-renowned neurosurgeon who has performed more than 2021 stereotactic surgeries and more than 0 brain tumor surgeries; He is one of the few neurosurgeons trained in pediatric neurosurgery, brain tumor surgery, and sports and epilepsy surgery. In 0, he became the second neurosurgeon in the world to perform laser cerebral hemispherectomy for the treatment of drug-resistant epilepsy in children.

Sean P. Polster, M.D., is associate director of the Neurovascular Surgery Program and director of the Skull Base and Neurovascular Laboratory at the University of Chicago School of Medicine. His main focus is on the treatment of neurovascular diseases of the head and neck, as well as brain and skull base tumors. As a physician and scientist, Dr. Polster is constantly researching new treatments and techniques to improve the care of neurosurgical patients. ）

"The bionic brain-computer interface we are developing has realized that the brain can sense what the hand is doing while using the robot." Peter Warnke told Tiger Sniff,"It's really exciting. It can be said that the prospects of AI in the field of neuroscience are limitless. ”

Peter Warnke and Sean P. Polster are also some of the first top doctors to try AI tools in surgery, and the surprises, problems, and concerns about the future of humanity encountered in the process of encountering AI are quite representative.

In the course of communicating with Peter Warnke and Sean P. Polster, Tiger Sniff learned that the human brain has hundreds of billions of neurons, most of which are between 15 and 0 microns, which makes the human brain the most complex and delicate structure, and also makes it extremely difficult to treat lesions. In today's extraneurologic surgery, not only is the surgical incision getting smaller and smaller, but the error range is also repeatedly compressed, and it can even reach less than 0.0 mm. As the ultimate solution to neurological disorders, electrodes implanted into the human brain can be as thin as one hundredth of a human hair in invasive brain-computer interface surgery.With the addition of AI, neurosurgery, including brain-computer interfaces, has been put on a "rocket".

AI decodes massive amounts of brain signals in an instant

Tiger Sniff:In recent years, many new technologies have been used in the field of neuroscience, and AI is one of them. What are your initial impressions of the use of this technology in neuroscience?

Dr. Warnke：The first time I used artificial intelligence was in '2019, when I was nearing retirement, and I worked with the University of Pittsburgh to work on an experimental brain-computer interface (BCI) project on the first patient. The experience of using AI to analyze brain signals in real time is impressive, as it requires the instantaneous decoding of massive amounts of brain neuronal signals.

AI-driven brain signal analysis is the perfect application for this technology.

Previously, we mainly used AI to analyze common electroencephalogram (EEG), and the signals mainly came from large areas of the brain. Now we use it to analyze stereo EEG with implanted electrodes – in this EEG, we implant electrodes into multiple brain regions, so that in addition to the classical EEG, a kind of three-dimensional data is generated. A single group of brain cells from epilepsy patients can be analyzed over a long period of time to generate terabytes of data.

It is impossible to analyze such a large amount of data by human beings, and AI can do so by identifying specific markers and developing response algorithms. Moreover, in this process, artificial intelligence still conducts multimodal analysis of circuit signals and image data.

It can be said that the application prospects of artificial intelligence in the field of neuroscience are limitless.

Tiger Sniff:Indeed, the speed at which artificial intelligence analyzes brain signals is indeed very fast, and there are reports that AI can record thousands of images of human brain activity in one second, and can recognize the pictures that people see in 5.0 seconds. The University of Chicago School of Medicine has also used AI in clinical practice, what are the main aspects of its use, and what is the effect of the application?

Dr. Warnke：A typical example of this is the application that started just a few weeks ago. In the past, we performed sustained deep brain stimulation in patients with Parkinson's disease or other tremors. Now, we record the depth signal from the basal ganglia online in real time, and use artificial intelligence algorithms to analyze and adjust the intensity of the stimulus based on the signals we recorded.

Dr. Polster:Angiography techniques, such as CT and MRI (magnetic resonance imaging), can quickly identify health problems if they are used in the screening of populations. But the pain point of these tests is that radiologists need to look closely at the images and combine them with symptoms to identify potential problems. If AI does the work, it can more efficiently identify and highlight potential areas that are easily overlooked and need to be monitored. I think this will amplify the scale of CT and MRI applications.

We have now put AI image analysis into practice. For example, if a patient is suspected of having cerebrovascular disease, the emergency doctor will perform perfusion CT angiography, and before the doctor checks it, the AI will process the relevant information, and if it is found that there may be lesions such as large vessel occlusion and cerebral hemorrhage, the system will notify the entire medical team, so that the doctor can triage the patient and determine what emergency treatment is needed, and timely intervention to improve the treatment effect.

Tiger Sniff:You mentioned the issue that radiologists may have overlooked some of the details. How much improvement has AI brought in this regard?

Dr. Polster:The data is still being collated, and what I can tell you is that our decision to apply AI to the clinic was based on data analysis. In the specific application of identifying pathology, we have evidence that AI technology is able to identify where human radiologists miss diagnoses. Especially when it comes to detecting aneurysms, AI detects them in every case and performs better than neuroradiologists.

For example, a patient was admitted to the hospital due to a ruptured middle cerebral artery (MCA) aneurysm that was bleeding and required emergency surgery. When the doctor focused on the aneurysm in the brain, the AI found and suggested that the patient also had an ophthalmic aneurysm that was at risk of rupturing in the future. In this way, the doctor treats two aneurysms at once, avoiding the hassle of finding new lesions and operating again.

It's hard to quantify how many lives this has saved, but I think it's a very useful tool because people can get tired and miss information, but the AI is performing steadily and it's working really well at the moment.

Patients are not a collection of variants

Tiger Sniff:What do you think is the biggest advantage of AI? What should be done if the AI advice conflicts with the intuition of a human doctor?

Dr. Warnke：I think the biggest advantage of AI is that it can go down to a very subtle level, especially when it comes to MRI diagnosis.

In MRI, completely different images are produced accordingly due to the different imaging sequences. AI is able to see more details that clinicians can't see based on its vast contextual analysis of millions of scans.

The best example is epilepsy surgery, where clinicians may think this is a perfectly normal scan, but AI analysis can detect small differences that can lead to new diagnoses, such as cortical dysplasia in the brain causing epilepsy that is invisible to the naked eye. Therefore, I don't think AI will contradict clinical experience, but will complement and improve it.

Tiger Sniff:With the proliferation of AI in clinical practice, there are concerns that doctors will become overly reliant on AI, leading to deterioration of capabilities, or reducing patients to a series of data. What's your take on this, and is there any way to avoid this?

Dr. Warnke:There are two aspects to this. On the one hand, AI is often criticized for reducing doctors' ability to perform their professional duties.A similar situation occurred with the first MRI.At the time, it was criticized that doctors would no longer be able to examine patients themselves if they relied directly on imaging, but this claim was not confirmed. Therefore, AI is just a tool.

On the other hand, your problem also involves treating the patient as just a set of numbers, and AI processes those numbers, which is the primary problem facing modern technology and medicine. We can't ignore the humanity behind the patient. In fact, the opposite should be true. Disease is a specific problem that needs to be addressed, but that's only on a holistic scale.For the patient himself, if the disease is not the main problem, but something else related to the patient's lifestyle and personal experience, it is pointless to just treat his specific disease perfectly.

Dr. Polster:When you're sick, you don't go to a scientist, you go to the doctor. Advances in tooling won't change that.

Tiger Sniff:The illusion of generative AI has always been a much-discussed issue, and it is also the key to its application in medical care.

Dr. Polster:Yes, it is also possible to have this kind of problem. The main benefit of using AI in the field of CT or MRI is that AI is a tool that can detect some unusual signs, some subtleties, and then prompt doctors to pay attention to these areas and determine if there is a problem. AI may misidentify lesions, but it's also not appropriate to treat patients as mere collections of lesions.

For example, the risk-benefit profile of a 85-year-old patient with a 0 to 0 mm proximal aneurysm is different from that of a 0-year-old patient with an aneurysm of the same size. At present, AI is not able to take these factors into account. That's why it's better to use AI as a tool, to give it a hint at the area to be examined, and then for the doctor to make a judgment based on the patient's condition.

The last line of defense for the human doctor

Tiger Sniff:What limitations do you see in AI as a tool used in the medical field?

Dr. Warnke:There are several typical limitations. One is the processing speed of the computers used by artificial intelligence, but this problem is constantly improving. The second is the ability to learn, and now the learning speed of general artificial intelligence (GenAI) products is close to that of humans, and they can adapt to different situations faster. But there's another limitation that will always be there, and that's that when confronted with emotionally unstable patients, AI can't quickly learn to respond to their unexpected reactions, as this is a unique human experience.

For example, when treating a mental illness such as obsessive-compulsive disorder, it is necessary to determine whether the concomitant depression is reactive or independent, which depends largely on the patient's social background.If information is collected in the form of questionnaires, reducing human life to a few numbers for artificial intelligence to handle, it will not be able to truly reflect the complexity of human beings.

Dr. Polster:As a tool, AI can perform image analysis and tell you if there is an abnormality, but it is up to the doctor to determine whether treatment is ultimately needed. That's the limitation of tools like AI.

Tiger Sniff:Does AI being too sensitive cause patient anxiety?

Dr. Warnke：This is indeed a real problem. In terms of tumor diagnosis, artificial intelligence may misjudge and misjudge some lesions as tumors. Similar situations exist in areas such as epilepsy, psychiatric disorders, neurodegenerative diseases, etc. AI can detect subtle changes in brain structure, such as those associated with Huntington's disease, but it's so sensitive that it can make patients anxious if they are diagnosed based on it.

We need long-term research to determine which is more predictive, AI or clinical judgment. AI can only process based on the input data, and although it is faster than humans and processes a larger amount of data, the clinical relevance can only be verified through long-term clinical studies, and the study period cannot be shortened, and the results of the study must be waited.

Dr. Polster:For example, each of us has moles on our bodies, and AI can identify all of them, but it can't tell if these moles are cancer and need to be removed. As a tool, AI can perform image analysis and tell you if there is an abnormality, but it is up to the doctor to determine whether treatment is ultimately needed.

The change of AI is unstoppable

Tiger Sniff:從2019年到現在也有五六年時間了，相比您第一次接觸AI，現在AI在腦機介面領域的應用，又有了哪些提升，您的職業軌跡又發生了哪些改變？

Dr. Warnke：AI has improved both signal decoding accuracy and real-time adaptive systems, but in order to achieve this, it will first take years to generate massive amounts of data.

Today, we specialize in nerve reconstructive surgery rather than mechanically removing the lesion or clipping the aneurysm.

With the help of AI, there have been really exciting situations in brain-computer interfaces, and now we are seeing at least one or two patients who, by decoding thoughts and intentions, can use robotic arms, speech generators, or other similar devices. Of course, the current speed is very slow and one-way. We're working on a very exciting new direction to develop one with other teamsBionic Brain Machine InterfaceThis interface not only allows you to use the robotic hand, but also allows you to give your brain a sense of what you are actually doing through feedback from your hands. This is indeed a very new technology, through faster decoding,You can do it almost as quickly as you would with your own hands.

All patients at the University of Chicago and the University of Pittsburgh lost most of their hand function, with cervical spinal cord injury being the most common cause. Brain-computer interfaces decode signals from the hand and specific finger areas, allowing them to use a robotic hand, which also provides sensory feedback to the sensory cortex.

Our goal is not to completely replace the function of the human hand, because the "sensors" of the human hand are far more than the robotic hand, but they will be infinitely closer.

Tiger Sniff:After AI enters the medical field, doctors can extend their experience to a wider field, which is also bringing changes to the medical model, including more and more sophisticated new technologies such as brain-computer interfaces, and also extending to the lives of ordinary people. We note that Dr. Polster has published a study in Nature on the relationship between gut microbiota and brain damage. It mentions that mice that were given emulsifiers (commonly used as preservatives for food) had more severe cerebral hemorrhages. In fact, this is also shifting from disease treatment to health management. You're a surgeon, why did you do this research, what inspired you, and how does it impact clinical care now?

Dr. Polster:We observed that people with similar diseases had very different symptoms, and it was this observation that led us to carry out this study. We studied cerebral cavernous malformations (CAs), which are tiny tangles of blood vessels in the brain. Most people who are diagnosed with this disease are able to live a normal life without any problems, but a small percentage of patients will be affected, such as: stroke, epilepsy, etc.

We don't have a good way to distinguish who has problems and who don't; We also don't know why some people get sick and others don't. In basic scientific research, we have found that a component of the gut microbiota is a sufficient and necessary condition to trigger bleeding from these lesions.

Our findings extend from animal experiments to humans, and we found that people with bleeding had intestinal dysbiosis.

The complication lies in the fact that it is very difficult to translate the results obtained in animal experiments, in a completely controlled rearing environment, or in the long-term use of powerful antibiotics into human medicine.There are many diseases that can be cured in animals, but it is extremely difficult to apply to humans, even with the help of artificial intelligence, high-throughput genomics, proteomics, and powerful computing power.

It's exciting to be able to understand the relationship between cerebrovascular abnormalities and the gut microbiota.However, beyond common-sense advice like eating healthy and not smoking, we are currently unable to give practical advice on how people should make lifestyle changes.I think digging into these specifics is something we need to do in the near future.

As for probiotics or similar supplements, our current research information even suggests that these may be harmful to you.But we're not sure yet if they're good or bad when it comes to interacting with the brain. This is an uncharted territory that we are currently actively working on.

Tiger Sniff:The University of Chicago School of Medicine has had a lot of exchanges with the Chinese medical community and has collaborated with Chinese doctors to perform many surgeries. I would like to ask the two experts, what are some of the most impressive events that have exceeded your expectations in working with Chinese doctors? And how does China's experience help your knowledge system and the global healthcare model?

Dr. Warnke：The most impressive thing is how quickly China is catching up with new technologies. The best example of this is when we first introduced laser ablation, which only started at 2014 years. However, just a year or two later, hospitals in China quickly picked up the technology. In fact, we also performed the first laser ablation procedure for epilepsy in collaboration with a Chinese hospital. As a result, the speed at which China is developing in the field of technology is amazing.

Dr. Polster:When it comes to image analysis, China has a huge advantage. Because of the large number of patients in China and the large number of CT and MRI examinations performed in a single hospital, this data can be combined with a lot of information. By processing this information, we are better able to triage patients, identify potential concerns early, and refer patients to the appropriate specialty, which in turn helps patients receive better treatment. That's what China's image analysis capabilities have helped us.