Spontaneous BOLD Signal Fluctuation in Resting State Functional MRI Demonstrates Difference in Hurst Exponent Distribution in Adults with and without Autism Spectrum Conditions

Background: Functional magnetic resonance imaging (fMRI) time-series are typically complex with variable local autocorrelation. After adequate motion correction, resting state fMRI signals can be modeled as fractional Gaussian noise (Maxim et al., 2005) described by two parameters: the Hurst exponent (H) and the signal variance. The Hurst exponent relates to the fractal dimension and describes the self-similarity of signals. Typically for fMRI time-series the range lies in 0.5<H<1, thus signals are positively autocorrelated and have so-called long-memory (persistent) behaviors. The spatial distribution of H in resting state fMRI corresponds to anatomical structures with significant differences between grey and white matter (Wink et al., 2008). Differences between patterns of H in neuropsychiatric conditions compared to neurotypical controls have been explored in Alzheimer’s disease (Maxim et al., 2005) and attention-deficit/hyperactivity disorder (Anderson et al., 2006). Since long-memory processes measured by Hurst exponent could be related to long-memory oscillations from spontaneous neuron firing, and since autism spectrum conditions (ASC) may be underpinned by atypical neuron synchrony and connectivity, we predict differences in H in a cross-sectional study of adults with ASC compared to matched neurotypical controls.

Objectives: To assess the differences in the distribution of the Hurst exponent in spontaneous BOLD signal fluctuations from fMRI brain imaging in adults with ASC and matched controls.

Methods: 31 adult, right-handed males (18-45 years old) with a clinical and ADI-R confirmed diagnosis of ASC, and 33 age-, sex-, handedness- and IQ-matched neurotypical adults were scanned in a 3T MRI scanner by echo planar imaging in an eye-closed, awake, non-task resting state. Following preprocessing of the acquired images to correct for motion, maps of H were generated at each intra-cerebral voxel for each participant. These were then co-registered into the standard anatomical space of the Montreal Neurological Institute by affine transform and cross-section statistical analysis at the cluster level performed using permutation inference (CamBA v2.2.0 http://www-bmu.psychiatry.cam.ac.uk/software/).

Results: Statistical significance was set such one false positive cluster was expected under the null-hypothesis (equivalent p=0.003). At this level widespread increases in H were observed in participants with ASC relative to controls, particularly in the limbic systems and the anterior cingulate.

Conclusions: These data provide preliminary evidence that H may be substantively different in resting neural oscillations in ASC compared with matched controls. The regions identified have previously been implicated with ASC using other imaging techniques and it seems therefore plausible that changes in the persistence of fMRI time-series reflect changes in the underlying neural systems associated with the condition.