ASC Diagnosis - An Analytical Method?

[GaryS]

I was reading this recent paper this morning, it's very heavy going but essentially using some posh MRI scanning, very smart image analysis together with inexplicable statistics to compare to traditional ADI scoring for individuals on the spectrum together with a NT control group come up with some encouraging possibilities in diagnosis and research: I wonder what everyone feels about the possibility of a series of physical tests (MRI scans) that could objectively diagnose all flavours of autism.

 

Extract of Publication Discussion

Autism affects multiple aspects of the cerebral 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 nonoverlapping 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. The discriminating patterns detected using SVM may help further exploration of the genetic and neuropathological underpinnings of ASD.

 

SVM = Support Vector Machine - a statistical analytic approach to understanding complex data

 

For the geeks... "Correct" Classification of ASD using SVM Left brain hemisphere 85% (Sens 90%, Spec 80% p=0) Right 65% (Sens 60% Spec 70% p=0.03)

[coolblue]

Two questions: firstly, how do they define 'autism'? Do they mean that their method can discriminate between the various flavours of autism shown by different individuals, ie it can say this person has autism and not something else, or at a group level - a group of people has autism as distinct from something else?

 

Secondly, why do they need an MRI scan to tell them which bits of the brain are different? You can make a well-educated guess by just watching someone's behaviour. If they don't make eye contact, have a language problem, are clumsy and can't make decisions, you *know* that they are going to have differences in occipital, temporal, parietal, cerebellar and frontal areas. What you don't know is whether those behavioural differences are due to brain differences or whether the brain differences are due to the behaviours.

 

cb

[GaryS]

The ASD individuals were defined using Micheal Rutter's ADI-R interview system, and as the methodology seems to be in it's infancy I couldn't say where they are going with it. My interest is from a diagnostic perspective; as the reason we have different flavours is because the current diagnostic metric is behavioural assessment which will by it's very nature result in very subjective and segmented diagnoses with multiple (observed) classifications. IMO the only way this will be overcome is to determine a single and more importantly reproducible, parameterised metric (which may have a complex route) to determine the presence or absence of causation.

 

The point is that your well-educated guess whether someone is on the spectrum or not is not the same as my well-educated guess (or that of anyone else, however qualified or experienced). The process of having all of these professionals, all of these procedures, all of these behavioural tests result in the possibilities of an erroneous misdiagnoses (positive or negative error) screams for a more reliable and objective methodology.

[coolblue]

 

 

The point is that your well-educated guess whether someone is on the spectrum or not is not the same as my well-educated guess (or that of anyone else, however qualified or experienced). The process of having all of these professionals, all of these procedures, all of these behavioural tests result in the possibilities of an erroneous misdiagnoses (positive or negative error) screams for a more reliable and objective methodology.

 

The point is not whether or not someone is on 'the spectrum', as assessed by a behavioural measure, but whether or not 'the spectrum' exists or not - in terms of behavioural measures.

 

The well-educated guess about differences in brain areas isn't going to show the same subjectivity, because you can tell from someone's behaviour what brain areas are involved. What you can't tell is what caused those brain differences - and you're not going to be able to figure that out from an MRI scan.

 

cb

[GaryS]

The point is not whether or not someone is on 'the spectrum', as assessed by a behavioural measure, but whether or not 'the spectrum' exists or not - in terms of behavioural measures.

 

I have no idea what that means It's a given the spectrum does exist, otherwise there wouldn't be so many different diagnosis with overlapping behaviours.

 

The well-educated guess about differences in brain areas isn't going to show the same subjectivity, because you can tell from someone's behaviour what brain areas are involved. What you can't tell is what caused those brain differences - and you're not going to be able to figure that out from an MRI scan.

 

cb

 

That's rather sweeping generalisation don't you think? Perhaps we are going to re-tread the genome-phenome path again but you have previously agreed that there are many morphological aspects that are similar across many of the forms of autism. These guys have perhaps found a method of using a selection of these morphologies under MRI and using complex predictive techniques to reduce the information into a straight forward metric. Now I'll admit that it is a little bit of a leap to suggest that this may have sufficient discriminatory power to become diagnostic (I am asking the main author of this paper this very question) but I was interested to hear from members what they thought of such a possibility.

[bid]

I'm taking part in research into autism and language processing, using MRI amongst other methods. I discussed this with the researcher the last time I attended a research session.

 

The question I raised is this: people such as myself, with no learning difficulties and a high IQ, learn various coping strategies which in themselves may presumably alter brain function, etc. So I'm sceptical how variants like this could be accomodated into some diagnostic tool using MRI.

 

I just can't see how an MRI scan of someone at one end of the spectrum will be the same as the scan of someone at the other end of the spectrum, especially once you start to consider learning difficulties, and say epilepsy, which tends to have a higher incidence amongst those with learning difficulties anyway.

 

Given that the only way a diagnostic tool using MRI could be developed would be to analyse a huge database of scans of people on the spectrum, I can't see how you could end up with a workable model. This was more or less what 'my' researcher seemed to think too.

 

Bid

[GaryS]

Perhaps you're right in terms of having to create a database of scans, perhaps this is part of the research you are taking part in, I don't know. Additionally, perhaps there are differences at between the various flavours of autism at this level, the research I'm quoting did show that ADHD did not have the same characteristics as the ASD so perhaps what you are saying is actually possible.

 

There is a great deal of research that has shown a variety of morphological changes in those diagnosed with autism, these researches used five assessments of a couple of these anatomical changes - cortical thickness and an estimation of surface area.

 

A set of five morphometric parameters per vertex were used as input to the multimodal classifier. Three parameters [1–3: average convexity or concavity, mean (radial) curvature, and metric distortion] accounted for geometric features at each cerebral vertex and two parameters (4, 5: cortical thickness and surface area, respectively) measured volumetric features.

 

If Declan Murphy is prepared to put his name to this work then I'm afraid I don't think "your" MRI researcher has researched enough.

[bid]

Perhaps you're right in terms of having to create a database of scans, perhaps this is part of the research you are taking part in, I don't know.

 

No, the research I'm taking part in is hopefully going to advance the understanding of how those on the spectrum process language, to be used for therapy/education. It's most definitely not to do with using MRI as a diagnostic tool.

 

'My' researcher did talk about this idea of a database of scans that would then be analysed to create the diagnositic tool.

 

HTH

 

Bid

[GaryS]

Thats a shame - I do wish research didn't have so many rules sometimes and that there was more sharing of data.

 

Is the research part of the AIMS project?

Edited March 29, 2011 by GaryS

[coolblue]

I have no idea what that means It's a given the spectrum does exist, otherwise there wouldn't be so many different diagnosis with overlapping behaviours.

 

Indeed there's a spectrum in terms of overlapping behaviours. But since it's a bit fuzzy round the edges, it's highly likely that the brain scans are going to be a bit fuzzy round the edges too. Someone diagnosed with ASD or possibly ADHD is quite likely to have an ASD?ADHD? scan too.

 

 

cb

[coolblue]

 

 

If Declan Murphy is prepared to put his name to this work then I'm afraid I don't think "your" MRI researcher has researched enough.

 

 

But Declan Murphy's model of autism is (if I've understood him correctly) that it results from genetic variations that lead to fundamental structural abnormalities in the autistic brain. I don't get the impression he's much taken with the model that says that brain structure can be the outcome of an interaction between a genetic predisposition, brain function and behaviour - a model that is equally plausible.

 

Why does he have to be right and bid's researcher wrong? There are some pretty eminent people doing research into language processing, and some of them might possibly be in a position to disagree with Declan Murphy.

[bid]

Thats a shame - I do wish research didn't have so many rules sometimes and that there was more sharing of data.

 

Is the research part of the AIMS project?

 

No, the research project I'm taking part in is a collaboration between the Autism Research Centre, the Cognition and Brain Science Unit and Anglia Ruskin University.

 

Bid

[GaryS]

Indeed there's a spectrum in terms of overlapping behaviours. But since it's a bit fuzzy round the edges, it's highly likely that the brain scans are going to be a bit fuzzy round the edges too. Someone diagnosed with ASD or possibly ADHD is quite likely to have an ASD?ADHD? scan too.

 

 

cb

 

I don't think you can make that assumption. Medicine is littered with similar symptomatic conditions that have specific aetiologies, and the whole purpose behind finding a differential diagnosis. The two flavours you have selected actually do have quite different outcomes in this research; the results separate ASD (the ones tested) from ADHD which have a similar distribution of results as NT scans.

 

But Declan Murphy's model of autism is (if I've understood him correctly) that it results from genetic variations that lead to fundamental structural abnormalities in the autistic brain. I don't get the impression he's much taken with the model that says that brain structure can be the outcome of an interaction between a genetic predisposition, brain function and behaviour - a model that is equally plausible.

 

Why does he have to be right and bid's researcher wrong? There are some pretty eminent people doing research into language processing, and some of them might possibly be in a position to disagree with Declan Murphy.

 

I believe you're correct in your assessment of Prof Murphy's thoughts on autism but have no idea what he feels about genetic predisposition. My own feeling is that it is a convenient logic parking space when everything can't be explained.

 

True, Murphy could be wrong I was expressing an opinion on the likelihood of that possibility

[BethK]

Problem you'll hit with something like this, is that the spectrum is vastly different from one end to the other. So an MRI won't 'determine' wat we want to know. So far, the only real way to determine is by a behavioral assessment. It's like that US 'breakthrough' with the mice, a human brain is way more complex, so in theory, if they were to develope a link with a drug (that would have to be taken for a full lifetime, if we're honest), it won't be anything that would take effect for a good 20 odd years.

[GaryS]

A step closer - or is it?

 

ADHD, Autism May Sometimes Share Gene Mutations

 

Just love this...

"What we are beginning to understand is that genes don't code for disorders as they appear in the diagnostic manual. Genes code for proteins that are needed for brain structure and brain function," Schachar said. "Many of the genes we discovered involved in both ADHD and [autism spectrum disorders] affect the development of the nervous system."

Bin DSM V then

Edited August 13, 2011 by GaryS