A Study of Spatial-Temporal Characteristics of Autism Fmri by Combining Hidden Markov Model and Dynamic Graph Theory

Abstract

Autism spectrum disorder (ASD) is a prevalent neurodevelopmental condition. Functional magnetic resonance imaging (fMRI) has emerged as a pivotal tool in studying ASD by measuring dynamic changes and evaluating the brain's intrinsic connectivity. Leveraging both spatial and temporal properties of fMRI can provide comprehensive understanding of brain activity, capturing its intricate structural relationships and dynamic functional changes. In this work, we uniquely integrated Hidden Markov Model (HMM) with dynamic graph (DG) theory to achieve the above goal. By employing dynamic functional connectivity, we captured the brain's dynamic shifts. Subsequently, the temporal variability of the topological structure was discerned through DG analysis, facilitating a comparison of dynamic differences between ASD subjects and healthy controls (HC). By utilizing the HMM, we identified the typical states of both ASD and healthy controls. We then examined the correlation between HMM indicators of these states and clinical autism scale scores to discern pattern differences. Both HMM and DG methods corroborated the findings using congruent mapping. Notably, the functional connectivity and spatial topological structure of the ventral attention network (VAN), visual network (VN), and default mode network (DMN) in ASD patients exhibited significant alterations. By employing diverse whole-brain analysis techniques, this study enhances the complementarity of information, offering potential insights for the diagnosis and treatment of ASD.

Note:
Funding declaration: This work was supported by the National Natural Science Foundation of China (grant numbers: 1237529, 82071913, and 22161142024).

Conflict of Interests: None.

Keywords: functional magnetic resonance imaging, Hidden Markov Model, dynamic graph theory, autism spectrum disorder