“Autism can be detected by a 15-minute brain scan,” the Daily Expresshas reported. The news is based on a study that investigated whether anatomical differences in the brain can be used to identify people with autism. It found that a brain scan and computer algorithm using five different measurements of brain shape and structure was up to 85% accurate in identifying the autistic spectrum disorder (ASD) in adults. These measurements could be used as a “biomarker” for autistic spectrum disorders, the researchers say.
This small preliminary study is a valuable contribution to the search for a better way of identifying autism, a condition that can be difficult to diagnose due to its wide range of causes, types and symptoms. However, it is not possible to say at present whether such a technique could replace or even aid current diagnostic methods in the near future. Far bigger studies comparing the brain scans of larger numbers of people with ASD and those without the condition are now needed to assess whether this scan is accurate enough for widespread use.
Where did the story come from?
The study was carried out by researchers from the Institute of Psychiatry at King's College London. Funding was provided by the Medical Research Council. The study was published in the peer-reviewed Journal of Neuroscience.
The study was widely reported in the media, with most stories concentrating on interviews and the information in the press release to interpret the scientific data featured in the published research paper. Few news reports discussed the relatively small size and preliminary nature of this exploratory study, or the need to test its methods in larger studies before it can be considered suitable for use in clinical diagnoses. The claim in the Daily Express, that autism can now be detected by a 15-minute brain scan, was incorrect.
What kind of research was this?
The autistic spectrum disorder (ASD) is made up of a range of different types of autistic conditions, with multiple causes and a wide range of symptoms. It is often associated with other behavioural disorders. These factors make it difficult to identify and describe the “neuroanatomy” (the internal neural structure of the brain associated with the condition). While previous research has highlighted several possible differences in the anatomy of particular regions of the brain in people with autism, these have only been studied in isolation.
This study aimed to test the theory that individuals with autism have “multidimensional” differences in brain shape, structure and volume and therefore that this “neuroanatomical pattern” can be used to identify ASD.
What did the research involve?
The researchers recruited participants through a clinical research programme, including 20 adults who had been diagnosed with ASD and a further 20 adults without the condition as a control group. All volunteers were right-handed males, between the ages of 20 and 68, and none had any history of medical disorders affecting brain function. The diagnosis of ASD was confirmed using accepted criteria. A further 19 adults diagnosed with Attention Deficit Hyperactivity Disorders (ADHD) were also recruited to act as a neuro-developmental control group, to see whether the method could differentiate between ASD and other neurodevelopmental disorders. This group was matched to the ASD group in gender, age and whether they were right- or left-handed.
The scientists used magnetic resonance imaging (MRI) to take scans of the brain’s grey matter in all three groups. A separate imaging technique was used to reconstruct these scans into 3D images. Using a computer algorithm, the images were then assessed and classified using five “morphometric parameters”. This means that the researchers looked at particular variations in the size, shape and structure of five different features of the brain’s grey matter, which are associated with ASD.
The results were assessed to see if the computer classification of people with ASD matched the clinical diagnosis.
What were the basic results?
Using this method, the study was able to identify individuals with ASD with a sensitivity (accuracy) of up to 90% (i.e. if a volunteer had a clinical diagnosis of ASD, there was a 90% probability that he was correctly assigned to the ASD category by the computer programme).
However, the accuracy of the results varied according to the measurements used. The computer diagnoses were more accurate using measurements from the left hemisphere of the brain, with individuals with ASD being correctly identified in 85% of all cases, when all five measures were taken into account. The highest accuracy of 90% was obtained using a measurement of cortical thickness in the left hemisphere.
On the right hemisphere, the assessments were not as accurate, with individuals with ASD being correctly classified in 65% of all the cases.
Specificity (correctly identifying that a person with no clinical diagnosis of ASD did not have the condition) was also very high. Of the control group, 80% were correctly classified as controls.
In the ADHD control group, information from the left hemisphere was used to correctly identify 15 of the 19 individuals with ADHD (78.9%), while four of these individuals (21%) were incorrectly allocated to the ASD group. Classifications using the right hemisphere were less accurate.
How did the researchers interpret the results?
The researchers say their approach confirms the hypothesis that the “neuroanatomy” of autism is “multidimensional”, affecting several different features of the brain. Their approach using “multiparameter classification” compares well with current diagnostic methods looking at behavioural signs and symptoms. They suggest that brain anatomy could be used as a “biomarker” to facilitate and guide behavioural diagnosis.
Conclusion
In this small preliminary study, researchers were able to correctly identify people with ASD with 90% accuracy, and individuals without ASD with 80% accuracy, using a variety of different measurements of the brain’s grey matter.
However, this study was only in 59 individuals in total. The findings need to be replicated in far larger studies before such a programme might be used to aid diagnosis in the clinical setting. In particular, it is necessary to clarify that this method can specifically differentiate between ASD and other neuro-developmental conditions. In addition, the implications of such a test for ASD would need careful consideration, including which people would be eligible for the test and whether it should be considered for use in children.
The researchers also note that:
- Differences in scanners may have affected the ADHD classification.
- The variation in accuracy between the right and left hemisphere needs further exploration.
- The classification algorithm was only used on high-functioning adults with ASD, so it is not known whether it would produce the same results in other groups with more severe ASD.
- The small sample size made it impossible to investigate any brain differences between autism and Asperger’s syndrome.
Overall, these are promising findings and further research is awaited with interest.
