Identification of neural connectivity signatures of autism using machine learning

Abstract

Alterations in interregional neural connectivity have been suggested as a signature of the pathobiology of autism. There have been many reports of functional and anatomical connectivity being altered while individuals with autism are engaged in complex cognitive and social tasks. Although disrupted instantaneous correlation between cortical regions observed from functional MRI is considered to be an explanatory model for autism, the causal influence of a brain area on another (effective connectivity) is a vital link missing in these studies. The current study focuses on addressing this in an fMRI study of Theory-of-Mind (ToM) in 15 high-functioning adolescents and adults with autism and 15 typically developing control participants. Participants viewed a series of comic strip vignettes in the MRI scanner and were asked to choose the most logical end to the story from three alternatives, separately for trials involving physical and intentional causality. The mean time series, extracted from 18 activated regions of interest, were processed using a multivariate autoregressive model (MVAR) to obtain the causality matrices for each of the 30 participants. These causal connectivity weights, along with assessment scores, functional connectivity values, and fractional anisotropy obtained from DTI data for each participant, were submitted to a recursive cluster elimination based support vector machine classifier to determine the accuracy with which the classifier can predict a novel participant's group membership (autism or control). We found a maximum classification accuracy of 95.9% with 19 features which had the highest discriminative ability between the groups. All of the 19 features were effective connectivity paths, indicating that causal information may be critical in discriminating between autism and control groups. These effective connectivity paths were also found to be significantly greater in controls as compared to ASD participants and consisted predominantly of outputs from the fusiform face area and middle temporal gyrus indicating impaired connectivity in ASD participants, particularly in the social brain areas. These findings collectively point toward the fact that alterations in causal connectivity in the brain in ASD could serve as a potential non-invasive neuroimaging signature for autism.

Keywords: autism; classification; effective connectivity; fMRI; machine learning; theory-of-mind.

Figure 1

A flow-chart depicting the Recursive Cluster Elimination based Support Vector Machine (RCE-SVM) procedure.

Figure 2

Graph showing classification accuracy, sensitivity and specificity obtained by simultaneously using the following features: behavioral scores, functional connectivity, effective connectivity and fractional anisotropy obtained from DTI. The X-axis shows number of clusters/number of features and the Y-axis, the performance (classification accuracy, sensitivity and specificity). ^*indicates significance (p < 0.05 corrected).

Figure 3

Mean of nineteen paths which was most important for giving maximum classification accuracy for autism and control groups. All paths had significantly decreased connectivity (p < 0.05 corrected using Bonferroni method for 18 paths; for one of the paths p < 0.05 uncorrected) in the Autism group as compared to controls. The bars represent standard errors.

Figure 4

The nineteen paths whose effective connectivity values were top-ranked features for classification of the two groups (Autism and Controls) with the maximum accuracy (Left panel: participants with autism; and right panel: control participants). The width of the arrows represents the path strength and the color of the path indicates its rank obtained during classification with 1 being the most significant and 19 being the least significant.

Figure B1

Connectivity maps showing Granger causality path weights for all possible connections between 18 ROIs. Top: Autism, Bottom: Controls. (A) posterior to anterior paths, (B) anterior to posterior paths, (C) left to right paths, (D) right to left paths.

Figure B2

Connectivity maps showing z-scores of functional connectivity path weights obtained using Pearson's correlation for all possible connections between 18 ROIs. Top: Autism, Bottom: Controls. (A) anterior–posterior paths, (B) bilateral paths.