Enhancing EEG-Based ASD Detection Using Wavelet Transforms and Hybrid Deep Learning Models

This chapter investigates the use of an integrated approach combining wavelet transform and a hybrid architecture to improve the accuracy and reliability of Autism Spectrum Disorder (ASD) diagnosis using electroencephalography (EEG) data. Wavelet transforms are employed to convert the EEG signals to images, which are then processed by an architecture with a pre-trained convolutional neural network for feature extraction and a classic classifier to capture intricate patterns of brain activity. The overall performance indicates room for improvement since the best classification accuracy was 69.047% by using LeNet combined with Random Forest. Part of this performance is justified by the quality of the database, which has missing data and little variability. However, this study already demonstrates good potential for an AI-supported approach in reducing false negative rates during the diagnostic process.