This study introduces a novel hybrid framework using electroencephalogram (EEG) signals for accurate, non-invasive Autism Spectrum Disorder (ASD) diagnosis. The advanced machine learning approach significantly improves early detection rates, aiding clinical assessment.
Current diagnostic methods are time-consuming and exhibit significant variability, hindering early detection.
There is a critical need for objective, non-invasive biomarkers for early ASD identification.
Purpose of the Study:
To develop and validate a hybrid feature fusion framework for non-invasive Autism Spectrum Disorder (ASD) diagnosis using electroencephalogram (EEG) signals.
To enhance diagnostic accuracy and reliability by integrating multi-domain EEG features and ensemble learning.
To establish a scalable, data-driven approach for early ASD detection.
Main Methods:
Utilized electroencephalogram (EEG) signals, specifically P300 event-related potentials (ERPs), from the BCIAUT-P300 dataset.
Extracted 22 features across time, frequency, and time-frequency domains using Wavelet Transform, power spectral density, higher-order statistics, and principal component analysis.
Implemented a hybrid fusion strategy combining optimal feature selection (PCA, HOS, PSD, FDA, CWT) and ensemble classifiers (SVM, XGBoost, LDA, Random Forest).
Main Results:
Achieved high classification performance with 97.7% accuracy, 96.8% sensitivity, and 98.5% specificity.
The hybrid fusion framework significantly outperformed traditional single-feature or single-classifier diagnostic methods.
Demonstrated the robustness and generalizability of the integrated multi-domain EEG features and ensemble learning approach.
Conclusions:
The proposed hybrid EEG-based framework offers a precise and reliable method for non-invasive ASD diagnosis.
This approach supports early ASD detection, aligning with biomedical informatics goals for clinical deployment.
The study highlights the potential of combining advanced signal processing and machine learning for improved neurodevelopmental disorder diagnostics.
