Note: Most Internet Explorer 8 users encounter issues playing the presentation videos. Please update your browser or use a different one if available.

Sex-Differential Fractal Complexity of Resting-State Brain Oscillations in Autism

Thursday, 2 May 2013: 09:00-13:00
Banquet Hall (Kursaal Centre)
M. C. Lai1, M. Lombardo2, B. Chakrabarti3, A. N. Ruigrok4, E. T. Bullmore4, S. Baron-Cohen5, M. A. Consortium6 and J. Suckling4, (1)Autism Research Centre, University of Cambridge, Cambridge, United Kingdom, (2)University of Cambridge, Autism Research Centre, Cambridge, United Kingdom, (3)Centre for Integrative Neuroscience and Neurodynamics, University of Reading, Reading, United Kingdom, (4)University of Cambridge, Cambridge, United Kingdom, (5)Autism Research Centre, Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom, (6)University of Cambridge, Institute of Psychiatry, University of Oxford, Cambridge, United Kingdom
Background: Fractal properties of resting-state fMRI BOLD signal oscillations have been shown to be atypical in adult males with autism spectrum conditions (ASC). Fractal scaling in males with ASC is shifted more towards randomness (Lai et al., 2010, Biol Psychiatry) suggestive of a ‘noisier’ resting-state signal in ASC. Given the emerging evidence of sex-differential characteristics of ASC, it remains to be seen how biological sex affects the fractal properties of resting-state oscillations as a potential fundamental characteristic of the neurobiology of ASC. 

Objectives: To investigate if the fractal properties of resting-state brain oscillations affect ASC differently according to biological sex.  

Methods: Four groups of right-handed adults (33 neurotypical males and 29 females, 25 males and 30 females with ASC; aged 18-49 years), matched for age, IQ and in-scanner head motion, were scanned with fMRI at 3T for 13 minutes (TR=1302 ms). Participants were only instructed to keep their eyes closed and stay awake. We measured at each voxel the Hurst exponent (H), which quantifies the fractal complexity of the resting-state fMRI time series, using a maximum likelihood estimator in the wavelet domain. With whole-gray matter (GM) H-maps, we used a linear support vector machine (L2-regularization and loss, ‘leave-one-subject-pair-out’ cross-validation, executed in LIBLINEAR) to classify individuals by their diagnostic status solely based on multi-voxel patterning of H-maps. Classification analyses were done within-sex in order to test predictive power of diagnostic classifications without the potential confound of sex. If sex is not a confounder we would expect similar levels of classification performance across each within-sex analysis. Classification performance was evaluated against the null hypothesis with a permutation test (1,000 iterations). Univariate analysis was also implemented with a voxel-wise two-way factorial design (factors of diagnosis and sex), with permutation inference at the cluster level (error cluster per image=1), to identify regions with a significant diagnosis-by-sex interaction.  

Results: Within males, multi-voxel pattern-information from GM H-maps were significantly above chance in accuracy of diagnosis predictions (76%, p=.001), sensitivity (70%, p=.03), specificity (82%, p=.001), positive predictive value (PPV, 79%, p=.001), and negative predictive value (NPV, 73%, p=.003). However within females, H-maps failed to provide pattern-information for significant classification performance above chance conditions (accuracy 42%, p=.89; sensitivity 37%, p=.95; specificity 47%, p=.57; PPV 41%, p=.90; NPV 42%, p=.80). Univariate factorial analysis revealed 3 clusters showing a significant diagnosis-by-sex interaction (p=.003), involving bilateral medial temporal lobes, lingual gyri, left insula and temporal pole, cerebellum and midbrain. The interaction was driven by lower H (i.e., shift-towards-randomness) in males with ASC compared to neurotypical males, whereas no difference was found in H between the female groups. In addition, neurotypical females had a significantly lower H than neurotypical males. 

Conclusions: Fractal scaling of resting-state fMRI brain oscillations provides sufficient multi-voxel pattern-information for predicting diagnostic status of ASC in adult males but not in females. These findings suggest importance for characterizing neural mechanisms for autism separately for each sex.

| More