Describing the Brain in Autism in Five Dimensions

Autism spectrum disorders (ASD) are a group of biologically based neurodevelopmental disorders characterized by impairments in three major domains:

1.       socialization

2.      communication

3.       behavior

These disorders include

·         autistic disorder (classic autism, sometimes called early infantile autism, childhood autism, or Kanner's autism),

·          Rett disorder,

·          childhood disintegrative disorder,

·         pervasive developmental disorder, not otherwise specified (PDD-NOS), and

·          Asperger disorder (also known as Asperger syndrome

A recent study by Ecker et al demonstrated  how a multiparameter classification approach can be used to characterize the complex and subtle structural pattern of gray matter anatomy implicated in adults with ASD. A set of five morphological parameters including volumetric and geometric features at each spatial location on the cortical surface was used to discriminate between people with ASD and controls using a support vector machine (SVM) analytic approach and to find a spatially distributed pattern of regions with maximal classification weights.

 On the basis of these patterns, SVM was able to identify individuals with ASD at a sensitivity and specificity of up to 90% and 80%, respectively. However, the ability of individual cortical features to discriminate between groups was highly variable, and the discriminating patterns of regions varied across parameters.

Morphometric feature Correctly classified (%) Sensitivity (%) Specificity (%) p Left hemisphere     All parameters 85 90 80 0*     Cortical thickness 90 90 90 0*     Radial curvature 72.5 65 80 <0.001     Average convexity 70 75 65 <0.004     Metric distortion 80 80 80 0*     Pial area 77.5 70 85 0* Right hemisphere     All parameters 65 60 70 <0.03     Cortical thickness 60 65 55 <0.01     Radial curvature 52.5 50 55 <0.30     Average convexity 50 40 60 <0.40     Metric distortion 57.5 45 70 <0.06     Pial area 45 45 45 <0.60

 The classification was specific to ASD rather than neurodevelopmental conditions in general (e.g., attention deficit hyperactivity disorder). These results confirm the hypothesis that the neuroanatomy of autism is truly multidimensional, and affects multiple and most likely independent cortical features. The spatial patterns detected using SVM may help further exploration of the specific genetic and neuropathological underpinnings of ASD, and provide new insights into the most likely multifactorial etiology of the condition.

@The Journal of Neuroscience August 2010