International Meeting for Autism Research: Neural Signatures Predict Autism Diagnosis

Neural Signatures Predict Autism Diagnosis

Friday, May 13, 2011
Elizabeth Ballroom E-F and Lirenta Foyer Level 2 (Manchester Grand Hyatt)
10:00 AM
M. D. Kaiser, J. A. Eilbott, R. H. Bennett, D. R. Sugrue and K. A. Pelphrey, Child Study Center, Yale University, New Haven, CT
Background: Using functional magnetic resonance imaging (fMRI) to assess brain activity during the visual perception of biological motion in 4- to 17-year-old children with autism (ASD), their unaffected siblings, and typically developing (TD) children, we recently discovered three types of “neural signatures of autism” including state, trait, and compensatory markers (Kaiser et al., 2010, PNAS). The state regions represent areas of dysfunction unique to the children with ASD. The identification of the state markers presents the possibility of predicting an autism diagnosis based on activity to biological motion in specific brain regions.

Objectives: The current study sought to predict group membership in a new set of children with and without autism using the state markers from our prior study.

Methods: Eleven children with ASD (mean age = 11.8±3.06 years) and 16 TD children (mean age = 11.58±2.65 years) participated in the study. Diagnosis of ASD was confirmed with ADOS, ADI-R and expert clinical judgment. Children viewed point-light displays of coherent or scrambled biological motion. Data collection is ongoing.

Results: We replicated our prior finding of hypoactivation in a set of brain regions involved in the perception of biological motion in children with ASD relative to TD children. We conducted a discriminant function analysis using activity within the previously identified state regions to identify a linear function of regional beta weights that maximally separated children with ASD from TD children. Six brain regions (right and left fusiform gyrus, right posterior superior temporal sulcus, right amygdala, left ventrolateral prefrontal cortex, and ventomedial prefontal cortex) were entered simultaneously. Three-fourths of the children with or without autism were correctly classified as having or not having autism. This analysis provided best-estimate probabilities of group membership for each participant based on the optimal weightings for the state regions. We then evaluated the receiver operating characteristics (ROC) of this probability value for the state regions of activity. The “gold-standard” for our ROC analyses was the diagnostic criteria used to define the ASD group (i.e. the ADOS, ADI-R, and expert clinical evaluation). This analysis revealed an area under the curve (AUC) value of .73 (95% confidence interval = .54 - .92) for the regions of state activity with an associated asymptotic significance level of .043. This AUC value of for the regions of state activity indicates that there is a .73 probability that the group differences in the imaging data from this set of brain regions allows correct identification of a person as having ASD. A probability score from the discriminant function analysis of .50 is associated with a sensitivity of .82 and specificity of .63 for ASD.

Conclusions: We previously identified disruption in brain regions involved in social perception in ASD. Here we replicated these results in a new sample and then utilized the previously identified regions reflecting the state of having autism to successfully predict diagnosis in a new group of children with and without ASD. Our results illustrate the predictive power of fMRI as a diagnostic tool.

| More