Diagnosis of autism spectrum disorder based on complex network features

Ghasem Sadeghi Bajestani; Mahboobe Behrooz; Adel Ghazi Khani; Mostafa Nouri-Baygi; Ali Mollaei

doi:10.1016/j.cmpb.2019.06.006

Diagnosis of autism spectrum disorder based on complex network features

Comput Methods Programs Biomed. 2019 Aug:177:277-283. doi: 10.1016/j.cmpb.2019.06.006. Epub 2019 Jun 8.

Authors

Ghasem Sadeghi Bajestani¹, Mahboobe Behrooz², Adel Ghazi Khani³, Mostafa Nouri-Baygi⁴, Ali Mollaei⁵

Affiliations

¹ Center for Computational Neuroscience Research, Department of Biomedical Engineering, Imam Reza International University, Mashhad, Razavi Khorasan, Iran. Electronic address: g.sadeghi@imamreza.ac.ir.
² Software Department, Computer Engineering Faculty, Imam Reza International University, Mashhad, Iran.
³ Computer Engineering Department, Imam Reza International University, Mashhad, Iran.
⁴ Computer Engineering Department, Ferdowsi University, Mashhad, Iran. Electronic address: nouribaygi@um.ac.ir.
⁵ Artificial Intelligence Department, Computer Engineering Faculty, Islamic Azad University, Mashhad, Iran.

PMID: 31319956
DOI: 10.1016/j.cmpb.2019.06.006

Abstract

Background and objectives: Autism spectrum disorder (ASD) is a disorder in the information flow of the human brain system that can lead to secondary problems for the patient. Only when ASD is diagnosed by clinical methods can the secondary problems be detected. Hence, diagnosis of ASD at an early age and in young children can prevent its secondary effects.

Methods: By employing the visibility graph (VG) algorithm, the present study examines the C3 single-channel of EEG signals and presents the differences among the topological features of complex networks resulting from these signals. The average degree (AD) can be a method for the detection of normal and ASD samples.

Results: With an accuracy 81/67%, the ASD class can be discerned.

Conclusions: The current paper demonstrates that only by the usage of EEG signals of the brain's C3 channel and the topological features of its network (AD and related features, such as R_ADACC and R_ADMPL) can ASD and NC target classes be distinguished at an early age.

Keywords: ASD; Average degree; Complex networks; EEG; KNN classification; Visibility graph.

MeSH terms

Algorithms
Autism Spectrum Disorder / diagnosis*
Brain / diagnostic imaging*
Child
Child, Preschool
Cluster Analysis
Diagnosis, Computer-Assisted / methods*
Electroencephalography*
Female
Humans
Male
Reproducibility of Results
Signal Processing, Computer-Assisted