A Novel Hybrid System for ASD Diagnosis: Integrating EEG Image Analysis, Deep Learning, and Swarm Intelligence

This paper introduces a hybrid deep transfer learning approach for the diagnosis of Autism Spectrum Disorder (ASD) utilizing electroencephalography (EEG) data. The proposed methodology transforms EEG signals into multi-resolution images via wavelet transforms and employs a pretrained convolutional neural network (CNN) for feature extraction. To enhance the discriminative power of the model, Particle Swarm Optimization (PSO) is applied for feature selection, optimizing the input to a Random Forest classifier. The results demonstrate that the Random Forest classifier with 100 trees, following PSO feature reduction, achieves a test accuracy of 85.29% and a Kappa statistic of 0.7059, indicating substantial agreement with actual diagnoses. This study highlights the potential of combining deep learning with evolutionary algorithms to create an efficient and accurate diagnostic tool for ASD, which could complement clinical assessments. Future directions include expanding the dataset, exploring multi-modal data integration, and refining the optimization algorithms to further enhance diagnostic capabilities.