Rest-fMRI based comparison study between autism spectrum disorder and typically control using graph frequency bands

Graph signal processing (GSP) is a subset of signal processing, allowing for the analysis of functional magnetic resonance imaging (fMRI) data in the topological domain of the brain. One of the most important and popular tools of GSP is graph Fourier transform (GFT), which can analyze the brain signals in different graph frequency bands. This paper has analyzed the resting-state fMRI (rfMRI) data of two sites using the GFT tool to discover new knowledge about autism spectrum disorder (ASD) and find features discriminating between ASD subjects and typical controls (TCs). The results were reported for both structural and functional atlases with different numbers of regions of interest (ROIs). For the ASD group, the signal energy concentrations in low and somewhat high frequency bands declined by increasing age in most well-known brain networks. The changes of signal energy levels in different graph frequency bands were less for ASD subjects in comparison to TC ones. This result seems to reflect the difficulty in dynamic switching, which in turn leads to lower behavioral flexibility in the ASD group. In the low graph frequency band, the segregation of brain ROIs and brain networks increased with the age of ASD subjects. For TCs, growing up led to the integration of brain ROIs and segregation of brain networks in low and high-frequency bands, respectively. In the low-frequency band, the growing process was accompanied by lower activation and higher isolation of ASD brain networks. In addition, the segregation of salient-ventral attention network and dorsal attention network of ASD subjects grew with age. The structural atlas results indicated the reduced segregation of ASD subjects' default mode network in the high graph frequency band. The cross-frequency functional connectivity analysis showed that high-frequency signals of the right precentral gyrus and right precuneus posterior cingulate cortex had connections with almost all the low-frequency ROIs so that all connections were dramatically different between ASD and TC. The results of different scenarios at different graph frequency bands demonstrate that the combinatorial usage of functional and structural data through GSP can open a new avenue to investigate ASD.

Automated summary

This paper utilizes graph signal processing tools to study autism spectrum disorder, finding that as age increases, energy concentration of brain signals in ASD patients decreases, leading to difficulty in dynamic switching and lower behavioral flexibility. There are significant differences in brain activity patterns between ASD and typical controls in different graph frequency bands.