Dissociating Individuals with Asperger SYNDROME FROM High Functioning AUTISM USING A Probabilistic Pattern Recognition Approach

Background: Autism is a spectrum condition (ASC) with multiple causes, co-morbid conditions, and a wide range in type and severity of symptoms expressed.  Traditionally, ASC can be further subdivided by some into individuals with Asperger syndrome (AS), who have normal language milestones, and individuals with high-functioning autism (HFA), who have a history of delayed language development.  Behavioral evidence suggests that by adulthood these groups are largely indistinguishable at the phenotypic level.  However, the extent to which these groups differ at the level of brain anatomy is largely unknown.

Objectives: To test if individuals with AS and HFA can be distinguished on the basis of grey matter neuroanatomy using a probabilistic pattern recognition approach.

Methods: Structural MRI data was collected on 38 well-characterized male adults with an ADI-R confirmed diagnosis of ASC (mean age = 25 yrs, mean FSIQ = 105), which included 23 individuals with AS (i.e. phrase speech acquired prior to 36 months) and 15 individuals diagnosed with HFA (i.e. history of delayed language acquisition).  Both subgroups were matched for age and FSIQ.  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 AS and HFA using a probabilistic pattern recognition (PR) approach, and to (2) find a spatially distributed pattern of regions with maximal discriminative power.

Results: Overall, the PR approach achieved above-chance separation between Asperger’s and HFA group and was able to classify individuals at an overall accuracy of 68.1% (p < 0.04) and a true/false positive rate of (TP/FP) 72.0/64.2 % respectively.  Discrimination of the Asperger’s subgroup was driven by a combination of volumetric and geometric features and revealed spatially distributed patterns of regions with maximal discriminative weights.

Conclusions: This study provides preliminary evidence for the hypothesis early language acquisition serves as a marker for distinct brain phenotypes within the autism spectrum, and that individuals with AS may be dissociated from individuals with HFA on the basis of spatially distributed patterns of grey matter abnormalities.