Advancing Early Autism Detection in Children through Machine Learning-Assisted Spectrogram Analysis

The diagnosis of Autism Spectrum Disorder (ASD) is characterized by ongoing difficulties with verbal and non-verbal communication, repetitive habits, and social interaction. Electroencephalography (EEG) is one of the most popular techniques for investigating neurological disorders such as autism because of its low cost, excellent temporal resolution, and general availability. However, the analysis of EEG recordings typically generates extensive data with dynamic characteristics, traditionally requiring visual inspection by trained clinicians for ASD detection. This approach is labor-intensive, costly, subjective, prone to errors, and lacks consistency. The aim of this research is to create an effective diagnostic method for the identification of ASD using time-frequency spectrogram images obtained from EEG signals. The process involves preprocessing raw EEG signals using filtering, followed by the Short-Time Fourier Transform (STFT) to convert them into spectrogram images. Machine learning (ML) techniques are then used to evaluate these images. The spectrogram images are processed to extract features, which are then inputted into five distinct ML classifiers, including k-Nearest Neighbor (kNN), Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), and Naive Bayes (NB), to perform classification. The system achieved an accuracy of 92.7% when employing kNN. Its simplicity and high performance suggest it could serve as a valuable decision support tool for healthcare practitioners in diagnosing ASD.