Hybrid DWNN -RF Framework for EEG-Based Autism Spectrum Disorder Detection: A Low-Complexity and Interpretable Approach

Título: Hybrid DWNN -RF Framework for EEG-Based Autism Spectrum Disorder Detection: A Low-Complexity and Interpretable Approach

Autores: Flávio Secco Fonseca, Jordana Leandro Seixas, Adrielly Sayonara de Oliveira Silva, Fábio Matheus Spindola da Cunha, Maiara Marçal, Ana Carolina Soares de Melo, Maíra Araújo de Santana, Cecilia Cordeiro da Silva, Clarisse Lins de Lima, Arianne Sarmento Torcate, Giselle Machado Magalhães Moreno, Juliana Carneiro Gomes Cassemiro & Wellington Pinheiro dos Santos

Resumo: Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition characterized by challenges in social interaction, communication, and repetitive behaviors. The diagnostic process for ASD remains challenging due to the heterogeneity of symptoms and reliance on subjective clinical evaluations, often leading to delayed interventions. Early and pre-cise diagnosis is critical for enabling timely therapeutic support and improving developmental outcomes. In this study, we propose a novel hybrid methodology for ASD detection using electroen-cephalographic (EEG) signals. Our approach integrates Deep-Wavelet Neural Networks (DWNN) for feature extraction with Random Forest classifiers for robust decision-making. DWNNs leverage predefined wavelet filter banks, eliminating the need for parameter tuning and reducing computational complexity while capturing both temporal and spectral patterns in EEG data. This architecture not only ensures efficient feature extraction but also operates sustainably without requiring GPU acceleration, making it suitable for resource-constrained environments. Experimental results demonstrate the effectiveness of our hybrid model, achiev-ing an accuracy of 82.0%, a kappa index of 0.640, sensitivity of 96.0%, and specificity of 68.0% when using a Random Forest with 500 trees. These findings highlight the potential of DWNN-based frameworks in providing accurate, reliable, and interpretable solutions for ASD diagnosis, paving the way for scalable and sustainable AI-driven healthcare applications.

Palavras-chave: Autism Spectrum Disorder (ASD); Electroencephalography (EEG); Deep-Wavelet Neural Network; Random Forest Classifier; Health Informatics; Neuroengineering.

Páginas: 7

Código DOI: 10.21528/CBIC2025-1176337

Artigo em PDF: CBIC_2025_paper1176337.pdf

Arquivo BibTeX:
CBIC_2025_1176337.bib