Objectives: The objective of this research was therefore to characterize the complex and subtle structural pattern of gray matter abnormalities in adults with ASD on the basis of multiple morphometric parameters, and to disentangle spatially distributed patterns of regional differences with potentially different neuropathological underpinning. Furthermore, we aimed to examine the predictive value of individual morphometric parameters for group membership (i.e. diagnostic value).
Methods: Structural MRI data was collected on 20 well-characterized male adults with ASD (mean age = 33 yrs, mean FSIQ = 103), and 20 age/IQ matched healthy controls. All individuals with ASD met algorithm cut-offs for ASD on both the ADI & ADOS. For each participant, a set of 5 morphological parameters including both volumetric and geometric features were obtained at each spatial location on the cortical surface (i.e. vertex) was obtained using FreeSurfer software. This set of measures was then used to (1) discriminate between individuals with ASD and controls using a Support Vector Machine (SVM) analytical approach, and to (2) find a spatially distributed pattern of regions with maximal discriminative power.
Results: Overall, SVM achieved good separation between ASD and control group and was able to identify individuals with ASD at a sensitivity and specificity of up to 90% and 80% respectively using cortical thickness measures. In addition, SVM revealed spatially distributed, independent patterns of regions with maximal discriminative power for each of the five morphometric features describing brain volume and geometry. For all parameters, the left hemisphere provided higher classification values than the right hemisphere.
Conclusions: Our results confirm the hypothesis that the neuroanatomy of ASD is truly multi-dimensional, i.e. affecting multiple brain regions with a differential involvement of individual areas. These differences also provided significant predictive power for group membership, and could thus be used as a potential biomarker for ASD to facilitate and guide the behavioural diagnosis. The spatial patterns detected using SVM may also help further exploration of the specific genetic and neuropathological underpinnings of ASD, and provide new insights into the most likely multi-factorial aetiology of the condition.
1 Ecker et al. (2009). Investigating the predictive value of whole-brain structural MR scans in autism: a pattern classification approach. NeuroImage [Epub ahead of print]