New study – “90% diagnostic accuracy”

11 Aug

According to study author Christine Ecker in today’s Guardian:

We know already that people with autism have differences in brain anatomy and some regions are just bigger and smaller or just different in shape…[o]ur technique can use this information to identify someone with autism.

The study used 20 non autistic controls and 20 autistic people – all adults – and found ‘significant differences’ in the grey matter areas of the brain which control behaviour and language. This is nothing new in itself, differences in brain structure have long been known about in regards to autism. Whats new in this study is the method – and resultant accuracy – of the detection of autism.

In the experiment, Ecker showed that her imaging technique was able to detect which people in her group had autism, with 90% accuracy. “If we get a new case, we will also hopefully be 90% accurate,” she said. The research, supported by the Medical Research Council, Wellcome Trust and National Institute for Health Research, is published today in the Journal of Neuroscience.

If this is established as a viable method (Carol Povey of NAS states that further testing is still required) then it’ll be the first true objective test for autism ever developed. So far, as everyone knows, autism is diagnosed based on the opinion of a clinician (or team of specialists). Whilst they will probably still play a role, this test offers an objectivity that would be unparalleled. It would also have the interesting effect of making the DSM diagnosis largely obsolete.

12 Responses to “New study – “90% diagnostic accuracy””

  1. Visitor August 11, 2010 at 17:45 #

    Small point: I think “90% diagnostic accuracy” in your heading should be in quotes Kev.

  2. Kev August 11, 2010 at 18:25 #

    No sooner said than done 😉

  3. Clay August 11, 2010 at 19:01 #

    Basically, I think it’s a good thing. It would be good to have objective proof, given the many errors in diagnoses currently.

  4. Joseph August 11, 2010 at 19:13 #

    That can’t be right. I’ve never heard of a 90%-accurate classifier built using 40 data points.

    I’m sure it’s 90% accurate in the data set that was used to build it. But what this tells me is that the result is extremely over-fitted.

    That’s the fundamental problem with all these claims about screening accuracy.

    In the future, I believe accurate screening instruments and classifiers won’t come out of the conventional scientific process. They will probably come from machine learning competitions, where it’s much harder to “cheat” so to speak.

  5. Corina Becker August 11, 2010 at 20:09 #

    It’s been around the news on twitter a lot today…

    A very interesting study, so I would like to see a larger study done.

  6. Norton Gunthorpe August 11, 2010 at 20:31 #

    From the paper:

    Autism affects multiple aspects of the cere-bral anatomy, which makes its neuroanatomical correlates inherently difficult to describe. Here, we used a multiparameter classification approach to characterize the complex and subtle gray matter differences in adults with ASD.

    SVM achieved good separation between groups, and revealed spatially distributed and largely non- overlapping patterns of regions with highest classification weights for each of five morphological features.

    Our results confirm that the neuroanatomy of ASD is truly multidimensional affecting multiple neural systems…

    This level of sensi- tivity compares well with behaviorally guided diagnostic tools whose accuracies are on average 80%.

    Naturally, one would ex-further demonstrate that the classification is driven by autistic symptoms, the test margins of individuals with ASD were correlated with measures of symptom severity (Ecker et al., 2010).

    We found that larger margins were associated with more severe impairments in the social and communication domain of the ADI-R. The classifier therefore seems to use neuroanatomical information specifically related to ASD rather than simply reflecting nonspecific effects introduced by any kind of pathology.

  7. Liz Ditz August 12, 2010 at 00:02 #

    There have been a couple more blog posts done about what the study really said, and what the twitter stream reported

    First off, the Kings College press release is pretty sedate

    http://www.kcl.ac.uk/news/news_details.php?news_id=1426&year=2010

    then

    @CebmBlog Autism & the brain scan: the real predictive value http://bit.ly/9BK1u8 positive predictive value brain scan is actually 4.5%

    http://wacogne.posterous.com/brain-scans-diagnose-autism-when-is-a-test-no

  8. RAJ August 12, 2010 at 13:30 #

    “We found that larger margins were associated with more severe impairments in the social and communication domain of the ADI-R. The classifier therefore seems to use neuroanatomical information specifically related to ASD rather than simply reflecting nonspecific effects introduced by any kind of pathology”

    The failure of ASD research is examining only ASD subjects compared to ‘Normal’ controls. Adult stroke patients with gross brain pathology as a group are also seen as being ‘socially impaired’. Does that mean the adult stroke patients are ‘Autistic’? Probably so, if you score their impairments using any of the ‘Gold Standard’ diagnostic tools such as ADI-R.

    http://www.ncbi.nlm.nih.gov/pubmed/9745234

    A diagnosis of ASD by ADI-R was conferred on Romanian orphans abandoned at birth and subjected to severe emotional deprivation.

    http://www.ncbi.nlm.nih.gov/pubmed/18093025?

    There is no structural anomoly in the brain in ASD that hasn’t also been found in a variety of neurodevelopmental and neuropsychiatric conditions.

    The study might have been more informed if they had control groups that included schizophrenia, mental retardation (without ASD), ADHD, Downs Syndrome, Fragile X etc..

    Non-specific abnormal brain pathology is associated with diverse disabling neurodevelopmental and neuropsychiatric conditions. What this study is really stating is that abnormal brain pathology is associated with risk for neurodevelopmental problems not specifically ASD.

  9. Tom August 12, 2010 at 16:21 #

    RAJ says: “The study might have been more informed if they had control groups that included schizophrenia, mental retardation (without ASD), ADHD, Downs Syndrome, Fragile X etc..”

    It might be helpful if RAJ bothered to read the paper. They evaluated 19 subjects with ADHD as a neurodevelopmental control. The classification scheme they used correctly assigned 15 of the 19 to the ADHD category.

    From the paper:
    “Bilaterally, the ASD classifier did not allocate the majority of ADHD subjects to the ASD category. This indicates that it does not perform equally well for other neurodevelopmental conditions, and is more specific to ASD. To further demonstrate that the classification is driven by autistic symptoms, the test margins of individuals with ASD were correlated with measures of symptom severity (Ecker et al., 2010). We found that larger margins were associated with more severe impairments in the social and communication domain of the ADI-R. The classifier therefore seems to use neuroanatomical information specifically related to ASD rather than simply reflecting nonspecific effects introduced by any kind of pathology.”

  10. BA August 12, 2010 at 17:48 #

    It is certainly preliminary research but very interesting. Here’s a brief post I have on it and diagnostics generally.

    http://www.psychologytoday.com/blog/radical-behaviorist/201008/time-throw-out-the-dsm

  11. Savannah Logsdon-Breakstone August 12, 2010 at 23:34 #

    While it might be promising, I do want to point out that at a 40 point sample, it is hardly definitive or near a point where it could be used in diagnosis. Much much more research- including a much larger sample- would need done before we approached that point.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: