Long
story short, it’s about pilot study on 41 healthy men and 30 men
with some mental disorders. They did Iowa Gambling Test which is a
psychological task thought to simulate real-life process of decision
making.
We
found some difference in brain activity during this exam between
healthy people and ill ones.
For
example our experiment showed that the people with psychiatric
disorders have the amygdala more frequently hyperactive compared to
healthy participants from the control group.
The
amygdala is an almond-shape set of neurons located deep in the
brain's medial temporal lobe.
Unfortunately, we did not have enough research sample to build a classifier for recognizing specific mental illnesses. Here are some questions for You: 1.How does Iowa Gambling Test look? Have you heard about this test before? 2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
3.What
do you think about using AI in diagnosis of mental ilnesses or other
sickness?
4.
Will it replace psychiatrists or other specialists in the future?
1. How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteIn Iowa Gambling Test on the screen we see four cards. Each card in one corner. We need to choose one of them. After choosing one card we get points. We need to decide which card we choose. Our goal is to find rule to get maximum amount of points. This is a short description of this task. I have heard about this test. Your presented article is a proof.
2. Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
I think this is possible. In this article we have a small group of participants and the result are amazing. With much more participants having specific disease we could build better classifier to recognize specific diseases. I belive we can do it.
3. What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think AI could be a great help for specialists diagnosing ilneses. AI could support them to take a right decision. We need to be careful and don’t trust in 100% AI decision. A specialist must have a last word to say.
4. Will it replace psychiatrists or other specialists in the future?
I don't think so. I think AI will support specialists to take a right decision. AI is a powerful and great tool. In right hands it could be a good help.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteThe Iowa Gambling Task (IGT) test was developed to test the propensity to take risks in people with CNS damage. Currently, the test is carried out mainly in people with nervous system diseases, affective disorders, patients with schizophrenia and addicted to psychoactive agents. IGT is a simulation of situations occurring in everyday life that require making decisions under strong emotional influence connected with reward and punishment. The test enables prediction of risk in conditions of uncertainty. When receiving a reward and punishment, a characteristic ERP stimulus is created in the brain of the person tested. From this signal neural activity is calculated. The signal can be divided into specific bands of neuronal activity. Using source localization algorithms it is possible to calculate which areas of the cerebral cortex were most active. Of course, the signal must be converted accordingly in advance. Do the preprocessing and post processing.
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
The results indicate that neuronal activity can indicate whether a cerebral has received a reward or a punishment. However, there is no information on the classification of disease units. The most active Brodmann's fields have been identified. To try to build such a classifier a lot of people need to be tested. The authors of the article indicate certain possibilities. However, the analysis of variable latents of brain function is very difficult. And it's expensive. Additionally, many research groups should be involved in this process to confirm a given phenomenon. So far nothing is certain and I am writing this as a co-author of the above article.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
There are trained neural networks that can partially recognize a disease entity from how it behaves and what it writes in social media. The same can be seen in the difference in brain behaviour for different stimuli in different disease units. By combining these phenomena it is, in my opinion, possible, although binary, to indicate whether there is a deviation in a person's neural activity. Like I said, this is very difficult.
4. Will it replace psychiatrists or other specialists in the future?
It certainly can't replace it. But it will detect the symptoms earlier than the obvious symptoms of the disease. That's the case with dementia. In the EEG and the fMRI, we see structural changes in the brain three years before the disease occurred. I think you, too, will be able to try and anticipate and respond to these symptoms. Many other mental illnesses show physiological preclinical phases of the disease. It will be worthwhile to examine those people who are genetically charged and respond to it.
1. How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteI've never heard of this test before. From wikipedia "Participants are presented with four virtual decks of cards on a computer screen. They are told that each deck holds cards that will either reward or penalize them, using game money. The goal of the game is to win as much money as possible. The decks differ from each other in the balance of reward versus penalty cards. Thus, some decks are "bad decks", and other decks are "good decks", because some decks will tend to reward the player more often than other decks."
2. Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
If we look at a small group of people we are already able to draw some conclusions, I think that increasing this group as much as possible may enable better and more accurate diagnosis of the disease.
3. What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think the analysis of this type of data by AI is much more objective than by psychologists or psychiatrists. I think we should definitely develop this technology in this area, because as people we don't really understand many things about our brains yet.
4. Will it replace psychiatrists or other specialists in the future?
AI will probably not replace them, but it will be the main working tool for both psychologists and psychiatrists.
1. How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteNo, I never heard of that. I needed to read the wiki article. and the comments of the commentators above to understand.
2. Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
I read this article, I think that if we increase the number of participants, we would be able to more accurately classify the disease. But the results, even with so many participants, are amazing
3. What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think that AI should already be introduced into medicine for a long time, as one of the most successful areas, it’s obvious that we still cannot trust 100% of AI, but we must already train
4. Will it replace psychiatrists or other specialists in the future?
Once I read about the experiment. where people pretended to be mentally ill, and psychiatrists could not recognize it, and when people said they were healthy, then psychiatrists did not agree with this, because they thought that the patients were still mentally ill. Therefore, I think that psychiatry is a very imperfect science, and help from AI is fatum
Hello,
ReplyDeleteCongratulations on publication in Frontiers. I am a little bit skeptical about the usage of EEG for subcortical structures activity monitoring, especially as small as the amygdala. However, I'll take a look at the GeoSource software.
1.How does Iowa Gambling Test look? Have you heard about this test before?
It is a neuropsychological test of real-life decision making. I found a nice presentation: https://www.youtube.com/watch?v=fRTghzKxRPM
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
Yes, in my opinion, the global, medical data-sharing platform is what would allow creating better results in the field of disease recognition and prediction.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think that it is a good idea because these methods are more sensitive to slight signal changes, and they are not becoming exhausted during a long shift at work.
4. Will it replace psychiatrists or other specialists in the future?
In my opinion, AI would be a great supporting tool.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteI did not hear about Iowa Gambling Test before. From what I read it is based on making decisions to get as much benefit as it is possible (e.g earn virtual money)
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
The common problem in ML an AI soultions is not sufficient amount of data, so maybe with the greater number of participant, you'll be able to to build a classifier.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
In my opinion it is a great idea. I have heard about research on discovering symptoms of depression from social media content by AI methods.
4. Will it replace psychiatrists or other specialists in the future?
I think AI-based solutions could be good supporting tools, but for sure should not replace doctors.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteI've never heard of it before, and reading the comments above, some very well informed, the test (or task) is about choosing symbols trying to discover rules of gratification, in order to maximize the gratification.
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific
diseases?
If the input to the classifiers contains the right features, then it might be possible. Problem might be that some diseases manifestations might overlap with others, and adding more participants might cause bluring of those.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
It can be a helping hand, but I think especially in mental ilnesses, human diagnostician searches a very large tree of possible diagnostic outcomes using very little data, and when asking questions, the questions very much depend on previous answers. I'm not sure we can implement such process using AI.
4. Will it replace psychiatrists or other specialists in the future?
I think that in most areas of life we care about, duo of AI and human will outperform AI alone, or human alone. So I'm not expecting those being replaced.
1. I have never heard of this test before. From what I read the Iowa test is that participants receive four virtual decks on the computer screen. Each deck contains reward and punishment cards that affect the money from the game. The goal of this game is to get as much money as possible. In the decks, the number of good and bad cards varies, so you can decide which deck will punish us more often, and which will reward us more often.
ReplyDelete2. When using artificial intelligence for your research, the most important thing is always to have a large and diverse data set, which is why I think that a larger number of participants can help develop the classifier. However, I also think that you should also invite people with various diseases to your research so that you can create a good classifier that differentiates not one but several diseases.
3. It is definitely very helpful. First of all, he can catch some nuances that a person sometimes does not notice and, above all, does not get tired. It can work 24/7. Thanks to this, doctors can focus on contact with patients and their treatment. However, even the best AI can make mistakes, so somebody should control the verdicts issued by AI. The doctor's experience is often better than the result spitted out by machine.
4. I think not. Some patients will certainly find it easier to open in front of a computer than another person, but the machine cannot replace empathy and understanding of a specialist doctor. That's why I think AI would be a great tool to help, not replace, doctors.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteThe test person is seated in front of four sets of cards (A, B, C, D). A loan of $ 2,000 is granted. The task of the examined person is to select cards from individual sets until he is informed about it. The selection of cards should be done in such a way as to earn as much as possible. This is a general presentation of the test without providing details. I have not heard of this test before.
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
In my opinion, a larger group of respondents selected according to strictly set criteria results in obtaining more accurate data on a given topic. It can also help to build a more accurate classifier.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
Artificial intelligence is a rapidly growing field of computer science and quite promising. Therefore, I believe that, when properly applied, it will help in diagnosing the age of diseases (not only mental). I recently read an article in which the Chinese robot 'Xiaoyi' passed the state medical examination. Of course, for now the robot will support doctors, not replace them - but I think it is a matter of time.
4. Will it replace psychiatrists or other specialists in the future?
As I wrote earlier, a Chinese robot based on artificial intelligence passed the state medical examination. A lot of information can be found on the internet about artificial intelligence supporting medicine. Certainly, the use of artificial intelligence in medicine will grow. This is particularly noticeable during the diagnosis, where the time in which the correct diagnosis will be made is quite often what counts.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteFrankly, I have never heard of IGT. As a brief Google search combined with reading your article revealed, the research is focused at investigating the process of decision making. The participant is presented four virtual card deck. The participant is asked to open the cards one by one. The cards can have two effects: prize or penalty. The goal of the participant is to gain as much money as possible. The thing is that the decks contain various number of 'good' cards, i.e. the cards with prize, not penalty. As per the reports in various papers, healthy participants usually find 'good' decks comparatively easily, while people with psychological disorders typically have hard time finding the profitable deck. Correct me if I am wrong :)
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
Well, as Andrew Ng said, Artificial Intelligence is like a space ship: it needs powerful engine and tons of fuel. The engine is a deep learning model, while the fuel is data. Both engine and data cost really much. I think, that theoretically it is possible to improve the performance of the classifiers recognizing specific diseases, but it requires a whole lot of work from both engine designers (mathematicians developing the models) and fuel producers (gathering data). Also we should take into account the fact, that EEG data are pretty crude fuel: they are noised and require a great deal of preprocessing: removal of artifacts and outliers, feature extraction and selection. So, to me, accurate classification of EEG signals is like flying to Mars: theoretically possible, but quite hard to perform.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think, that it is a good idea to use AI for diagnosis of mental diseases, moreover. As a brief Google scholar search shows: it is used for that purpose: for long time, by many researchers and in many fields.
4. Will it replace psychiatrists or other specialists in the future?
I think, even if it is going to happen one day, the day is very, very distant. This picture illustrates my thoughts quite well:
https://www.facebook.com/Slidenerd/photos/a.109363625910483/1327260040787496/?type=3&theater
To my opinion, for now all the artificial intelligence models are merely more or less sophisticated approximating functions with more or less lack of interpretability. So, to sum up: for very long time the experience and skills of psychiatrists and other qualified professionals will be irreplaceable.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteI didn’t heard about this test yet, from what I learned, it’s a psychological test though to simulate real-life decision making. This decision-making required that the participants simply select a card from one of four decks. Two of the decks will reliably provide a profit in the long-term. They don’t give big wins, but the losses are smaller, and you’ll come out on top if you stick to them. The other two decks however provide huge profits, but also huge losses, and over enough trials will leave you bankrupt. While the participants complete the test, they have some information available to them – how their earnings are looking. The decks never change in their likelihoods of delivering a profit or loss, so eventually the average person figures this out, and sticks to a strategy. They consistently select the decks that deliver profit over the long-term.
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
In theory, based on the way how machine learnings models works – I say yes – It’s only a matter of training your algorithm on big enough data set to have a good classifier. On the other hand I think that variety of psychological disorders might be a big obstacle and cause extending the data set to very, very big scale to obtain a medically useful classifier.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think that AI is a great tool for diagnostics – it’s objective, thorough, and in many cases better than human in catching out outliers and finding patterns, but I think it’s a long way ahead of us, to create a useable tool that will be approved to be used in medical diagnosis process on a daily bases.
4. Will it replace psychiatrists or other specialists in the future?
Not in a near future, but I think that some day when technology will be mature enough, and medical use cases thoroughly tested we will be treatment by algorithms, not psychiatrists.
1.How does Iowa Gambling Test look? Have you heard about this test before?
ReplyDeleteIt's simulation of decision making process...I would reproduce information which were written berofe. That's useless :)
2.Do You think that with much more participants we will be able to build classifiers for recognizing specific diseases?
Maybe a few, but I can't image how. The diseases are very similar to each other and symptoms are other for everyone.
3.What do you think about using AI in diagnosis of mental ilnesses or other sickness?
I think that AI could be only tool which will recognize only a few sickness but the last opinios should came from doctor which will analysis the information about patient and verify this diagnosis. I couldn't imagine to get diagnosis from pure algorithm.
4. Will it replace psychiatrists or other specialists in the future?
Urfortunately I think that every doctor is the nearest of patient and know the best how to treat all of us (in ideal world...but still had a intuition which machine doesn't). That tool could be very helpful but I want get the last information from true doctor which will listen me and have at least some empathy.