A Fusion-Based Machine Learning Approach for Autism Detection in Young Children Using Magnetoencephalography Signals

Aug 5, 2026

Upload by Jiven

This study targets early biomarkers of autism in young children by recording magnetoencephalography (MEG) signals from thirty children aged 4–7 years with autism and thirty age- and gender-matched typically developing controls while they watched cartoons. Neural oscillations are characterized using amplitude features (power spectral density, PSD) and phase features (preferred phase angle, PPA). Machine learning classification achieves higher accuracy with PPA than PSD (88% vs 82%). A fusion method combining PSD and PPA yields 94% feature-level fusion and 98% score-level fusion, revealing autism-discriminatory oscillation patterns and supporting autism pathophysiology insights.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Abstract

Aim and dataset

Feature design (PSD and PPA)

Machine learning performance

Fusion strategy results

Implications for autism pathophysiology

Journal of Autism and Developmental Disorders [https://doi.org/10.1007/s10803-022-05767-w](https://doi.org/10.1007/s10803-022-05767-w) A Fusion‑Based Machine Learning Approach for Autism Detection in Young Children Using Magnetoencephalography Signals Kasturi Barik1 · Katsumi Watanabe2 · Joydeep Bhattacharya3 · Goutam Saha1 Accepted: 16 September 2022 © The Author(s) 2022 Abstract In this study, we aimed to find biomarkers of autism in young children. We recorded magnetoencephalography (MEG) in thirty children (4–7 years) with autism and thirty age, gender-matched controls while they were watching cartoons. We focused on characterizing neural oscillations by amplitude (power spectral density, PSD) and phase (preferred phase angle, PPA). Machine learning based classifier showed a higher classification accuracy (88%) for PPA features than PSD features (82%). Further, by a novel fusion method combining PSD and PPA features, we achieved an average classification accuracy of 94% and 98% for feature-level and score-level fusion, respectively. These findings reveal discriminatory patterns of neural oscillations of autism in young children and provide novel insight into autism pathophysiology. Keywords Autism spectrum disorder · Brain oscillations · Preferred phase angle · MEG · Classification · Biomarker Introduction Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder that influences the brain's information processing during infancy. Autism is characterized by disabilities in social association and communication, which generally exhibits repetitive behaviours and a restricted range of interests (Frith, 2008) . Its prevalence rate in children is reported to vary from 0.23% in India (Rudra et al. , 2017), 1.7% in the UK (Baio, 2014), to 2 .5% in the USA (Xu et al., 2018) . ASD influences individuals in various manners and * Joydeep Bhattacharya [j.bhattacharya@gold.ac.uk](j.bhattacharya@gold.ac.uk) Kasturi Barik [kasturibarik.phd@iitkgp.ac.in](kasturibarik.phd@iitkgp.ac.in) Katsumi Watanabe [katz@waseda.jp](katz@waseda.jp) Goutam Saha [gsaha@ece.iitkgp.ernet.in](gsaha@ece.iitkgp.ernet.in) 1 Department of Electronics and Electrical Communication Engineering, Indian Institute of Technology Kharagpur, Kharagpur, India 2 Faculty of Science and Engineering, Waseda University, Tokyo, Japan 3 Department of Psychology, Goldsmiths, University of London, London, UK can vary from mild to extreme (Amaral et al. , 2017) and causes a significantly debilitating effect on the quality of life (Farley et al., 2009). There is no remedy for ASD, yet early identification followed by suitable intervention can reduce symptom severity, uphold improvement in behaviour and learning of an autistic child. Of note, predominant behavioural symptoms of ASD emerge later in the developmental phase, so there is a critical need for identifying early signs of ASD (Wolff et al., 2018) . This study aims to identify early neural markers of ASDin young children from their resting state neuromagnetic brain responses. A growing body of literature suggests that ASD is often associated with disruptions in the features of large scale brain oscillations (Billeci et al., 2013 ; Simon & Wallace, 2016), and this has been reported as the core feature of ASD pathophysiology in very young children (Gabard-Durnam et al. , 2019) . Electroencephalography (EEG) and Magnetoencephalography (MEG) are traditional neuroimaging techniques that record macroscopic brain activity with millisecond precision in a non-invasive fashion. Here, the MEG signal is preferred as it is reference-free. We recorded brain responses from young children between 4 and 7 years of age using a MEG device specially customized forchildren while they were watching cartoons of their choice. There were sixty children, equally divided into two groups: children with ASD and age-matched typically developing (TD) children. One of the dominant theories of autism is ba

FAQs

What data and age group were used in the study?

The study recorded MEG from thirty children with autism aged 4–7 years and thirty age- and gender-matched typically developing controls.

How were neural oscillations represented for classification?

Neural oscillations were represented using amplitude features (power spectral density, PSD) and phase features (preferred phase angle, PPA).

What classification accuracy did the fusion method achieve?

Combining PSD and PPA features produced 94% accuracy for feature-level fusion and 98% for score-level fusion.

Aug 5, 2026

Upload by Jiven