Slashdot Mirror
IBM's AI Can Predict Schizophrenia With 74 Percent Accuracy By Looking at the Brain's Blood Flow (engadget.com)
Andrew Tarantola reports via Engadget: Schizophrenia is not a particularly common mental health disorder in America, affecting just 1.2 percent of the population or around 3.2 million people, but its effects can be debilitating. However, pioneering research conducted by IBM and the University of Alberta could soon help doctors diagnose the onset of the disease and the severity of its symptoms using a simple MRI scan and a neural network built to look at blood flow within the brain. The research team first trained its neural network on a 95-member dataset of anonymized fMRI images from the Function Biomedical Informatics Research Network which included scans of both patients with schizophrenia and a healthy control group. These images illustrated the flow of blood through various parts of the brain as the patients completed a simple audio-based exercise. From this data, the neural network cobbled together a predictive model of the likelihood that a patient suffered from schizophrenia based on the blood flow. It was able to accurately discern between the control group and those with schizophrenia 74 percent of the time. What's more, the model managed to also predict the severity of symptoms once they set in. The study has been published in the journal Nature.
"In 2013, the American Psychiatric Association released the fifth edition of the DSM (DSM-5). To be diagnosed with schizophrenia, two diagnostic criteria have to be met .... The person had to be suffering from delusions, hallucinations, or disorganized speech. A second symptom could be negative symptoms, or severely disorganized or catatonic behaviour"
In other words, Watson finds symptoms that would result in those traits. Schizophrenia is a loose subjective diagnosis, and its not consistently diagnosed between experts, so Watson has simply made itself one of those subjective experts.
Lack of blood flow would cause hallucinations, and disorganization, but then that's what lack of blood flow does. The disease there is lack of blood flow.
How is 1.2% "not particularly common"?
At a guess, it would be blow flow to the place where dreams are generated and it would be a reflection of dream state interacting with conscious state.
Chaos - everything, everywhere, everywhen
discriminating between people who have sz, and those who don't isn't very difficult.
how does it go discriminating between people with sz, and say, bipolar? these can be genuinely difficult for clinicians to tell apart, and would be useful.
Consider fire. Consider fire fighting, fire detection, and fire prevention.
There are many well known ways of using either heat, or presence of smoke in the air, to indicate a high likelihood that there is a fire in a region of a building. But these detection methods do not tell you anything about how the fire started. Combining information from many detectors across a large building can tell you about how a fire is spreading, but not about how a raging fire _might_ spread.
That people who have had some episodes labelled as 'schizophrenia' leaves common tell-tale signs detectable in this way is good to know, from a research point of view. But just like the 'fire analogies' above, where multiple similar looking fires, with similar results, can start in markedly different ways, the similar features in brains of people diagnosed with 'schizophrenia' only tell _part_ of the picture. As for how their brain came to be that way, such evidence can be likened to evidence that a fire in one room is a common cause of a fire in a neighbouring room. Things like 'chemical imbalances' are often touted as the 'cause' of things like bipolar disorder, or schizophrenia, rather than as a link in a causative chain. (This to me has always seemed as silly as saying that the pictures on your TV are _caused_ by electrical fluctuations in the aerial.)
In general, I think people working with mind and brain tend to overgeneralise, exaggerate, and oversell the consequences of their observations. This is further compounded when potential counter-evidence is 'defended against' and 'argued away', as happens between different factions of the mental health profession. People want things to be straightforward and simple, as if treating cuts and broken bones, and often inadvertently assume things are that simple before proceeding with studies whose results rely upon statistical reasoning which is contingent upon various assumptions uniformly holding across the population being studied... and then generally don't make clear their assumptions. People then read peer-reviewed research, and assuming a far simpler and more uniform picture than the evidence warrants.
John_Chalisque
You need to measure the performance of the model on data that wasn't used to train the model in the first place. Otherwise you will likely just get a model which recognizes known images, but doesn't have any predictive strength.
It just means system wants to destroy and murder those people for some reason.
Yeah, because they're fucking crazy!
I have Quadrophenia. P. Townsend
You extended U.S. stats to the world. You assume the rest of the world has as many mental disorders as the U.S. While most of those belonging to your major political party still stand behind Trump.
Don't know if you're joking/trolling. Schizophrenia affects 1% of the population and this rate is pretty much constant over the whole world and through history.
74% is better than random guessing (50%), but not by that much. This tells us that it might be possible to diagnose schizophrenia by MRI analysis, but it is far from a useful product.
Schizophrenia cannot be predicted by looking at the brains blood flow.
Africans who settle in Europe have a unusual high proportion of sufferers of schizophrenia.
Europeans who suffer from schizophrenia usually just hear voices in their heads talking about them or they believed that people on the television are talking about them. But Africans mostly hear voices telling them to do things to other people in other words they tend to be violent. nobody understands why Africans who suffer from schizophrenia are more often than not aggressive.
Incidence is actually lower in Western countries, including the US. Though rates are much higher in black immigrants to the West.
https://en.wikipedia.org/wiki/...
That's racist. Only straight white male rural Trump supporters have mental disorders. They are deplorable.
if our political narratives will change. A popular narrative is folks can solve their problems through sheer force of will. Often not even very much. But more and more science is finding breakdowns in human anatomy are the cause of many behavioral problems. Think of those studies that showed rats in a healthy community didn't have addiction problems.
As we find more and more that people aren't inherently predisposed to behavior and that their environment and physicalities dictate their behavior much more than we've liked to acknowledge I wonder if we'll see a breakdown in the old notions of fault and the old "Pull Yourself Up by Your Bootstraps" mentality. Let's not forget that that phrase is, after all, a literal impossibility.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
... IBM can very inaccurately predict schizophrenia...
Can you not see that already??? This is just all parallel construction in action.
Dont trust the parallel construction people.
give them a butcher knife and see if they stab the researcher
love is just extroverted narcissism
Something something about Trump calling South Americans criminals and rapists
These diagnoses will be very convenient after the camps get built. Imagine these corporations intentionally causing the problems they pretend to diagnose, the way McAfee used to WRITE viruses. Imagine a friend disappearing here and there whenever "AI" gets cranky.
Using a dangerous schizophrenia rate of 0.5% (the fraction varies by severity of schizophrenia) across a US population of 320 million, applying the test to everyone gives ~83.1 million false positives (normal people diagnosed as dangerous schizophrenics) while missing 368 thousand dangerous schizophrenics. How the hell can that be useful?
False positive rate != 1 - overall accuracy
Imagine a square (with an area of 1 arbitrary unit squared) divided by an off-center horizontal line and an off-center vertical line. Let's say the top rectangles represent cases where schizophrenia was detected (regardless of correctness), the bottom two cases where no schizophrenia was detected, the left pair cases where there actually isn't schizophrenia, and the right pair cases where the subject does have schizophrenia. The labels on our square look something like this:
FP TP
TN FN
If the lines are off-center (and they will be!) and don't intersect along the / diagonal (and they won't!), the FP probability and the FN probability will be different!
Also, even if we use the infamous =0.05, we end up overdiagnosing by a factor of about 10 instead of 100
CAPTCHA: symmetry
So if this test where to be administered, it would indicate that close to 100 million people in the US alone have schizophrenia. :-)
Though when I look at my co-works I sometimes wonder...
Letter To Iran
Schizophrenia affects 1% of the population and this rate is pretty much constant over the whole world and through history.
I tried to find a source for that, but failed. Do you have one?
It seems problematic to say that any rate is prevalent in the world, given that schizophrenia is a rather loose diagnosis where the definition varies based on locality.
In some parts of the world, having a mental health diagnosis carries a significant stigma and is avoided, unlike in other parts of the world.
And in some countries, like the US, allow MDs to give diagnoses in fields they have no qualifications for, and how well-meaning doctors knowingly set a dubious diagnosis to get insurance companies to pay for treatments and therapies that the patient otherwise would be refused. (This happened in my family, where a relative with celebral palsy could not get anger management therapy paid for with that diagnosis, even though it was the cause. But schizophrenia qualified, so after discussions with a doctor, that diagnosis was added, knowing that it was wrong. +1 for schizophrenia statistics.)
This is from a "Nature Partner Journal" (see http://www.nature.com/partnerpublishing/journals/), which is a line of open access content under the Nature brand, not Nature itself.
Normally when I read an article like this, I go to the Methods first to figure how much skepticism I'm going to have when I then check out the Results. There isn't a Methods section in this paper! I'm consequently not terribly impressed by the quality of this journal.
If "Schizophrenia" is something 'behavioral' in society, then 'Schizophrenia' can't be a disease, nor an illness, because stamping a label on people doesn't make them sick, nor ill.
So what is 'Schizophrenia'? A set of symptoms alone?
Or, a thesis? Superstition? Wishful thinking?
Somehow I strongly doubt the scientific rigor and truth value of knowledge in the field of psychiatry.
By taking 10000 people at random I was able to get 98.8% accuracy by always returning false.
Long time mental health worker here. Using an MRI to d/x schizophrenia is fairly needless, but
I understand that it has research interest/implications. For a lot less money and time, you can talk to some one for a few minutes and discern this quite readily. It's almost always that obvious. I would disagree regarding the parent poster about "dangerous schizophrenics." That is, most patients with this condition are
not imminently dangerous, although clearly, if actively psychotic, are at increased risk for violent behavior.
Where I also disagree, is the point many posters are making that 74% is useless as a screening tool. The difference is how healthcare views screening. Screening should be biased toward false positives. Diagnosis is not made on the basis of screening tools alone (at least it should not be). A positive screen is a signal for an expert to step in and make a judgment, which may involve collecting more data.
Trump supporters don't necessarily have mental disorders. Trump however...
So, what happens when you increase blood flow, and make sure these people are living in an environment with clean oxygen rich air? Do the symptoms go away? Is air pollution limiting the amount of oxygen that is getting to the brain and impacting people's ability to think? We know that lead poisoning does this, what about airborne heavy metals from other forms of pollution (mercury from coal and whatever is in diesel and gasoline exhaust)?
I thought I read some study about how the crime rate went down and intelligence went up when unleaded gasoline stopped being used.
i can't even believe the smash skum ha ha time. you think ibm is a tall tower with excrement flow. its not! it's not! flow blood in the damn urinary crank case with the stalled rotors. i have a hernia even looking at that land. who says the land is lazarus?
Bullshit. A study like this couldn't possibly encompass the scope of the entire population and its nuances. We have enough misdiagnosis, thanks.
Saying something is 74% accurate without stating false positive and false negative rates falls apart for rare diseases.
Here's an example: I have actually have a better method that can distingish between a control group and the real cases with 98.8% accuracy. I'm not kidding. All I do is I always say the person does not have the disease. Since 98.8% of people do not have it, I'm automatically correct 98.8% of the time.
Some drink at the fountain of knowledge. Others just gargle.
Patient: I'd like to get an MRI.
IBM: The MRI is complete and we've informed the federal government that you are 74% more likely to misuse a firearm. Any other tests? Or press the 'any' key to continue?
Shit, really? I bet people working with machine learning in research or industry have no idea about this caveat! Did you read the paper to check if they split the data into training and validation sets? Or are you just spouting inapplicable truisms that might or might not apply?
Yep, pretty sure that certain localities at various times have had a 0% schizophrenia rate but a 1% demon possession rate.
Now that we know it works in theory, can they train it with:
* A larger sample
* Schizophrenics who have been treated with various psych drugs (many of which affect brain function)
* Schizophrenics who have never been treated with pharmaceuticals (probably not very many of these out there)
* Non-schizophrenics who have been treated with various medications
If 74% is their first-run success rate, that is very promising.
If they focus on getting fMRIs from schizophrenics prior to drug treatments, they would have a better picture of schizophrenia as it exists in the wild---and a better idea of what undiagnosed schizophrenics look like. This seems like it would be the most useful data for reliably confirming a new diagnosis.
And maybe there should be public funding to try this with other disorders.
---
According to the latest ruleset, this post should be modded as Vorpal Flamebait +5.
then there's not much too it. It's either inherited, a random defect or something caused by their environment. Environment here does not mean upbringing. I'm talking about the literal physical environment. e.g. like how there was widespread violence in the US because we used leaded Gasoline.
You do bring up a good point, which is that as tech improves we'll be able to spot people who are inherently predisposed to certain types of behavior. There's a nice big body of dystopian sci-fi on this subject too (Minority Report & Psycho Pass come to mind).
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
I'm sure doctors can use this as a supplementary diagnostic aid, just another datapoint, but you'd be imcompetent if you used this as your sole diagnostic method.
If the flow of blood to the brain is zero then there is absolutely no chance that the person will develop schizophrenia.
Take that, so-called "scientists"!
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
I wish news reports would not mention "detection rates" without mentioning "false detection rates".
I can make a test with a 100% detection rate- it's the sentence "they have schizophrenia". Really shitty false detection rate though.
This was a test group of 95 people. We were told that the AI diagnosed 74 % of them correctly. Now how high was the score of the human doctors using all means available, except for the AI?
Let me quote the relevant parts of the story to you:
"The research team first trained its neural network on a 95-member dataset of anonymized fMRI images from the Function Biomedical Informatics Research Network which included scans of both patients with schizophrenia and a healthy control group. [...] From this data, the neural network cobbled together a predictive model of the likelihood that a patient suffered from schizophrenia based on the blood flow. It was able to accurately discern between the control group and those with schizophrenia 74 percent of the time."
That does read like they tested on the training set, doesn't it? Asshole.
it was still basically predictive analysis. The point is still valid regardless of how you're predicting the future. A psychic pre-cog saying you're going to commit a crime is functionally no different than the super computer's in Psycho Pass. It's just the author's preferred narrative convenience.
My point still stands. What happens to the narrative of personal responsibility in a world where the machinery of human beings has been solved. Where we know and understand every process down to the smallest level? Could you hold someone responsible for their actions when you can trace the exact path that led them to those actions? Especially for people clearly taking self destructive actions (petty criminals, drug users and the like)?
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
IBM should try and use Watson to predict their company revenue, as the current method seems to be producing exaggerated figures.
sz = send z-modem file transfer
rz = receive z-modem file transfer
[grin]
Ant(Dude) @ Quality Foraged Links (AQFL.net) & The Ant Farm (antfarm.ma.cx / antfarm.home.dhs.org).
I pray that this device be available to test school kids once a year or so. Being able to identify kids who will become ill in advance may well mean that we can moderate the severity of the disease at onset. A great deal of the alcoholism and drug use we deal with starts with a person who is fighting the onset of mental illness. Now if we can expand the ability of this sort of device we might save millions of lives as well as the huge money spent on mental illness every year. My belief is that the vast majority of people in prison are there due to mental health issues that they try to control with alcohol or dope. Imagine the suffering and the money that could be eliminated.
Mexicans are not South American whitey.
Using a dangerous schizophrenia rate of 0.5% (the fraction varies by severity of schizophrenia) across a US population of 320 million, applying the test to everyone gives ~83.1 million false positives (normal people diagnosed as dangerous schizophrenics) while missing 368 thousand dangerous schizophrenics. How the hell can that be useful?
You're unable to use any form of the word "schizophrenia" without compulsively preceding it with the word "dangerous".
Which is diagnostic, is it not?