Automated detection of autism based on the electrical signals of the brain (EEG)

Early screening is essential for effective intervention and rehabilitation in children with Autism Spectrum Disorder (ASD). EEG signals, offering real-time and sensitive monitoring, were used in this study to develop two proposed methods for ASD diagnosis. A dataset of 52 samples (22 with ASD and 30 controls) was collected using 19-channel sleep EEG recordings. Method 1: This method utilized energy differences between the left and right brain hemispheres. Preprocessing steps included decimation, involved noise removal FIR-BPF (< 0.5 Hz and > 49 Hz to reject unwanted frequencies) and artifact subspace reconstruction, and amplitude normalization. EEG signals were decomposed into five frequency bands, and peak channel envelopes were calculated for each hemisphere. Features—maximum, mean, and minimum energy values—were extracted using a sliding window approach with varying overlap ratios (12.5%–87.5%). The SVM classifiers with Linear (L), RBF, and Quadratic kernels were used and, the highest performance was achieved in the theta band: accuracy 91.7%, sensitivity 91.4%, and F1 measure 91.6%, with SVM-L using maximum energy features. Method 2:This approach analyzed two EEG dataset from Iraq and Poland. Preprocessing involved noise removal (< 0.5 Hz and > 49 Hz to reject unwanted frequencies) and artifact subspace reconstruction. EEG channel power spectral density features and brain topography maps were processed using deep feature extractors (e.g., AlexNet, GoogLeNet). ANOVA was applied for feature selection, followed by classification using SVM and Linear Discriminant Analysis (LDA). The alpha band demonstrated the best performance. The mean accuracy reached 98% with an standard deviation (SD) of 2.3% (Iraq, GoogLeNet + ANOVA + SVM-L) and 96.5% with an SD of 3.6% (Poland, AlexNet-FC6 + ANOVA + SVM-L). Both methods highlight the potential of EEG based systems as cost-effective and accurate diagnostic tools, providing valuable decision-support for healthcare professionals in ASD diagnosis.