Patterns of Information Flow in Autism Canonical Brain Network by Transfer Entropy approach

The functional brain activity network is the result of the harmonious activity of different regions. Each harmonic activity requires feedback from different areas of activity to adjust itself. As a result, any disorder in this harmony can affect the functional brain activity network. In this study, we use transfer entropy (TE) to examine the information transfer patterns of brain activity in Healthy Controls (HC) and individuals with Autism Spectrum Disorder (ASD). The results indicate that the pattern of information transfer between brain canonical networks is different in these two groups, ASDs and HCs. The results of our study demonstrate that the HC group exhibits a higher volume of feedback in comparison to the ASD group. In addition, the major sources of information flow in HCs are the Default Mode Network and Visual Network, but the hubs in the ASD group are the Frontoparietal Network and Limbic. Also, the backbone extracted from the TE graph of brain regions of interest shows us the same modularity class in the flow of information in both groups, but we observe the separated fragments in ASDs. Finally, the question that can be answered is whether the pattern of information transfer can be used as a biomarker in diagnosing ASDs vs. HCs. To explore this, we employed XGBoost, which achieved a mean classification accuracy of 91% in distinguishing individuals with ASDs from HCs across cross-validations.