Alterations in Brain Entropy in Autism Spectrum Disorders

Background: Biological systems typically exhibit complex behavior with nonlinear dynamic properties. Nonlinear signal processing techniques such as sample entropy is a novel approach to characterize the temporal dynamics of the brain. Estimating entropy, the state of uncertainty of a system, is especially important in clinical populations such as autism spectrum disorders (ASD) as higher entropy would entail brain disease. Considering the models of atypical brain network connectivity in ASD, sample entropy would provide a novel dimension to understand brain organization.

Objectives: The main objective of this study is to examine the alterations in brain entropy in resting state functional MRI data in children with ASD. We hypothesize increased brain entropy in children with ASD compared to typically developing (TD) children.

Methods: Resting state fMRI data collected from 22 high-functioning children with ASD and 18 age-and-IQ-matched TD children were preprocessed using the CONN toolbox. This consisted of motion correction, normalization to MNI template (3mm isotropic voxels), spatial blurring (6mm), low-pass filtering (.008 < f < .08 Hz), CompCor (a method for identifying principal components associated with white matter and cerebrospinal fluid, CSF), and scrubbing using the ART toolbox. White matter and CSF were entered as nuisance variables along with motion parameters and outliers of head motion detected by ART. Sample entropy was then calculated in a whole-brain voxelwise manner for all subjects with the following parameters based on a previous study (Sokunbi et al., 2013; r = 0.46 and m = 2). Group differences were assessed using a two-sample t-test, and brain-behavior correlations were calculated using sample entropy, Social Communication Questionnaire (SCQ), and Social Responsiveness Scale (SRS) scores from all subjects.

Results: The main results are: I) significantly increased sample entropy in ASD, relative to TD in bilateral insula and right calcarine sulcus; II) significantly reduced sample entropy in ASD, relative to TD in left middle temporal gyrus; and III) positive correlations of average sample entropy in clusters of significant group differences (insula, calcarine sulcus) with SCQ (r = 0.71, p < 0.001) and SRS (r = 0.63, p < 0.001) scores across all subjects respectively were found.

Conclusions: Higher sample entropy found in insula and calcarine sulcus in ASD children indicate increased randomness of a system, meaning the dynamic system activity is less predictable and less organized. This might also reflect abnormal brain response in ASD, particularly in visual and salience networks (Samson et al., 2012; Uddin et al., 2013). These findings are consistent with altered sample entropy previously reported in clinical populations such as schizophrenia (Sokunbi et al., 2014), and ADHD (Sokunbi et al., 2014). The correlation of increased sample entropy with ASD symptoms in our study underscores the clinical implications of this neurobiological index.