Autism Diagnosed With a Fifteen Minute Brain Scan

kkleiner writes "A new technique developed at King's College London uses a fifteen minute MRI scan to diagnose autism spectrum disorder (ASD). The scan is used to analyze the structure of grey matter in the brain, and tests have shown that it can identify individuals already diagnosed with autism with 90% accuracy. The research could change the way that autism is diagnosed – including screening children for the disorder at a young age."

Re:Unacceptable false positive rate

[abigsmurf]

aaaaand it says that in the article.

Bleh, I'm not used to actual good reporting on "a new totally amazing test!!!" stories in the media.

Much higher

[Mirey]

its actually much higher than that. What you're quoting is that 1 in 10 people with autism and given a false negative. Its actually much worse. Out of 10,000 children, 1980 would be found positive, out of which only 90 would have the disease. So only about 5% of people who tested postive would actually be autistic. It says this in TFA.

I'm not exactly impressed...

[Anubis IV]

They're trumpeting this 90% statistic, but what about the false positive rate? Contrary to standard /. procedure (I know, I know, I'm sorry), I decided to read TFA for an answer, and this is what I found out.

Let’s think of 10,000 children. Of these 100 (1%) will have autism, 90 of these 100 would have a positive test, 10 are missed as they have a negative test: there’s the 90% reported accuracy by the media.

But what about the 9,900 who don’t have the disease? 7,920 of these will test negative (the specificity in the Ecker paper is 80%). But, the real worry though, is the numbers without the disease who test positive. This will be substantial: 1,980 of the 9,900 without the disease. This is what happens at very low prevalences, the numbers falsely misdiagnosed rockets. Alarmingly, of the 2,070 with a positive test, only 90 will have the disease, which is roughly 4.5%.

So it only has a 4.5% true positive rate. Great.

Re:Unacceptable false positive rate

[AlexiaDeath]

Part of the official diagnosis criteria of any mental disorder is that it needs to cause a significant problem with normal functioning. If it does not cause a problem for you, you wont fit the criteria. Most people in IT have some aspie traits, but you need to know a real Aspie to know what it really means. My significant other is an officially diagnosed aspie and he is severely impaired by it. Things that to normal people do without thinking are hard for him.

Re:Or..

[Ihlosi]

Does that sound to you like basically any kid under the age of 10?

Err .. no. And if you ever did a side-by-side comparison, you'd know.

"Any kid under the age of 10" doesn't scratch wallpaper off the wall until their fingers are bloody, for example. Or spend an hour or two bouncing in circles and shouting "La-DEE la-DEE la-DEE" at the top of their voice.

Base rate fallacy

[martyros]

Not really -- the problem is with the base rate fallacy [wikipedia.org]. Suppose that there's a test that will tell you whether or not you have a disease with 99% accuracy: if you have it, you're 99% likely to test positive; if you don't have it, you're 99% likely to test negative.

Now, you get a test and it's positive. What's your probability of having the disease?

The answer is, "There's not enough information to answer the question." The missing piece of information is the "base rate".

Suppose that 50% of the people have the disease. Then in testing 1 million people, 500K will have the disease, of which 495K will come back positive (true positive), and 5K the test will come back negative (false negatives). 500K will not have the disease, of which 495k will come back negative (true negative), and 5k will come back positive (false positive). If the test came back positive, you're either a true positive or a false positive. Since there are 500K positives, and 495K of those are true positives, your chances of having the disease are 99%.

Suppose instead that 1% of people have the disease. Then in testing 1 million people, 990K will not have the disease, and 10K will have it. Of the 990K, 980K will come back negative (true negative) and 10K will come back positive (false positive). Of the 10K, 9900 will come back positive (true positive), and 100 will come back negative (false negative). There are 19,900 who tested positive, of which only 9900 (less than half) actually have the disease. So if you tested positive, your chances are about 50%.

So even if the test itself is very accurate (and I think 99% is pretty accurate), if the base rate is low enough (and in autism I believe it's still less than 1%), a positive reading may not be conclusive. You'd have to correlate it with other symptoms to make sure.

suribe

[suribe]

read "Why autism can't be diagnosed with brain scans" at http://www.guardian.co.uk/science/blog/2010/aug/12/autism-brain-scan-statistics [guardian.co.uk]

Re:Is there enough Helium?

[Dunbal]

I don't understand your reasoning.

      That's because you're not a health care professional. Put your personal feelings aside for a second and look at it this way:

      I have a population that might have a disease. The symptoms of the disease are not obvious, and there's no easy way to tell who has it and who doesn't.

      I have a very specific genetic test (Western Blot) that can tell me if that person's blood contains viral DNA. The test is specific because I am identifying DNA from the virus in question. You can't get more specific than that. However this test does have the possibility of false negatives - the person may have the virus, but I might have gotten a sample that for some reason contains no viral particles. Or the virus may be latent, living in T-cells in the bone marrow right now and not circulating in peripheral blood. NOT finding the DNA does not mean you don't have the virus. Finding the DNA confirms that indeed you have the virus. Also, this test costs a lot of money, and uses a lot of resources - less nowadays than 10 years ago, but still.

      On the other hand, I have a very sensitive test, the Enzyme Linked ImmunoSorbent Assay. This test identifies antibodies to the HIV virus down to very very dilute concentrations. It is extremely cheap. It can be done everywhere there's a centrifuge to obtain plasma from a blood sample. However because it identifies antibodies and not the virus itself, it is not a specific test. It's possible that a person have HIV antibodies without having the virus - because they were exposed to it enough to trigger an immune response, but never got an infection (the virus never "took hold"), or by some freak of nature, they have a SIMILAR antibody to something else that is reacting to the test. So thus the false positives - the test says you have it when really you don't.

      What we doctors do is we tend to SCREEN the population with the simple, inexpensive test first. Why? Because it's simple and inexpensive. That gives us a new population that absolutely captures all the diseased people, and also contains some false positives. We tell people that there may be a problem but they shouldn't worry about it just yet - but we need to run another test to be sure. YOU DO NOT TREAT PEOPLE BASED ON A PRIMARY SCREEN!

      To this new population you administer the second, expensive, SPECIFIC test. This lets you "weed out" those people who are false positives. You tell them congratulations, everything is ok and they probably shouldn't worry. You just keep an eye on them for a year or so to make sure they were indeed false positives and weren't people who for some reason gave a false negative on the second test. Maybe you repeat the test the year after just to be safe. And the rest, the real positives, end up with the diagnosis and the treatment.

      So what happens is you administer the expensive test to fewer people, saving time and resources, without letting anyone with the disease slip through the cracks. In the case of autism, presumably the primary screen would be the clinical signs - does the child exhibit autistic behavior and fit the criteria? IF that is the case, or IF the physician isn't sure, then the secondary screen (MRI) would be performed. My doubts are about the specificity of the MRI - will it show false positives - people with MRI changes that don't have autism? If so, then it's possible people will be misdiagnosed and that's a no-no.

      As for social stigma, uh, that's what doctor-patient confidentiality is about. I certainly won't tell anyone. It's not my problem if I tell you "listen we ran this test and you tested positive for HIV. It doesn't mean you have HIV though, we need to do a different test to be sure" and then you run around telling everyone you have HIV. I find that patients usually understand things when you explain them properly.

Re:Real Humiliation

[MarkRose]

I went through exactly that. Being as smart as I am, people just figured that I shouldn't have problems. But when reading body language and figuring out social boundaries is a strenuous mental exercise, and doesn't come naturally or work subconsciously as it does with most people, it's exhausting, and very frustrating when you keep screwing up, unable to figure out all the rules. The humiliation never completely goes away, but you get used to it after a while.

I'm 28, and I just figured out I have Asperger's Syndrome about a month ago. Not knowing until now has caused me a huge amount of grief. If I had known in kindergarten, it would have helped. Even then, I wasn't relating to the other kids -- and I never knew why I couldn't make friends. It wasn't a lack of trying.

"Well functioning" individuals with autism spectrum disorders can get better. After a while, we build up the "rules" for social interaction. The mental effort never goes away, but like learning to play chess, the basics do come more naturally after a while. It'll never be like riding a bike. To this day, I have trouble continuing a casual conversation. I'll never really connect with anyone that isn't a nerd. I'm okay with that.

In a sense, we are actors, life is a stage, and we do all our own stunts. The biggest problem well functioning individuals on the autistic spectrum face is coming across too normal, so that people attributed our odd behaviours as intentional and not to an innocent lack of understanding. We can learn, but because we usually highly intelligent, it's not obvious we need guidance or help.

Discovering that I am on the spectrum has brought a lot more of the humility I had already begun to learn in an effort to relate to people. My high intelligence made me arrogant as a kid. I used to look down on people if they weren't as smart as I am. It took me a while to recognize they had talents in areas I didn't. Now I know why my abilities are so different than those of a normal person.

I am blogging about being an aspie [aspieaspects.com], too. I'll probably repost this there later.