Truace
Subscribe

Automated autism spectrum disorder detection using EEG signals and time-frequency visibility graphs

Early and objective screening of Autism Spectrum Disorder (ASD) remains challenging because conventional diagnosis primarily relies on behavioural assessment and clinical observation. To address this limitation, this study proposes a dual-domain computational framework for automated EEG-based ASD classification by integrating complementary time-frequency analysis with Horizontal Visibility Graph (HVG)-based network modelling. Four time-frequency decomposition techniques, namely the Short-Time Fourier Transform (…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-0801· cite & verify
Automated autism spectrum disorder detection using EEG signals and time-frequency visibility graphs
"Mary Lou McDonald & Aengus Ó Snodaigh meet D12 Autism Support Parents Groupin Dublin" by Sinn Féin is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.

In brief

On 2026-08-15, a peer-reviewed study described a dual-domain computational framework for automated ASD detection from resting-state EEG, combining time-frequency analysis with Horizontal Visibility Graph modelling and machine learning classification.

The reported 93.54% accuracy and 98.17% AUC suggest a potentially objective, interpretable screening aid to complement behavioural assessment, but the source only reports internal stratified 10-fold cross-validation without external clinical deployment, prospective validation, or patient outcome data.

Main points

  1. Study integrated four time-frequency methods — STFT, DWT, WVD, and SLT — with Horizontal Visibility Graph network modelling of EEG.
  2. 17 graph-theoretic descriptors were extracted from HVG networks and evaluated with conventional classifiers including a Soft Voting Ensemble.
  3. Wilcoxon signed-rank test confirmed superiority of DWT-based representation over STFT, WVD, and SLT approaches.

The gain

A DWT-HVG framework with Soft Voting Ensemble enabled automated classification of resting-state EEG for ASD screening with 93.54% accuracy and 98.17% AUC in stratified 10-fold cross-validation.

The rundown

Researchers applied four time-frequency decomposition techniques to resting-state EEG and converted the representations into Horizontal Visibility Graph networks to capture non-stationary dynamics.

From those networks they extracted 17 graph-theoretic descriptors and tested conventional machine learning classifiers, with the DWT-HVG plus Soft Voting Ensemble outperforming other combinations on accuracy, sensitivity, specificity, F1-score and AUC.

Sources

  1. Peer-reviewedNeuroscience2026-08-15

The debate