Classification of Children With Attention-Deficit Hyperactivity Disorder Using Wigner-Ville Time-Frequency and Deep expEEGNetwork Feature-Based Computational Models

Attention-deficit hyperactivity disorder (ADHD) is a globally prevalent, chronic childhood neurodevelopmental disorder that may persist into adulthood. Early medical intervention can alleviate ADHD’s impact on children. While certain advanced technologies have emerged for ADHD diagnosis using electroencephalography (EEG) signals, a crucial trade-off arises between heightened computational demands and reduced success rates due to inadequate signal analysis and temporal mapping. To overcome this challenge, we introduce two approaches: a smoothed-pseudo Wigner-Ville time-frequency (T-F) feature-based support vector machine (spWV-SVMrbf) and a novel hybrid integration of expanded-EEGNetwork with long short-term memory (expEEGNetwork-LSTM). The spWV-SVMrbf utilized diverse band-energy information, whereas the expEEGNetwork-LSTM captured temporal nuances of the EEG signal, for classifying ADHD children. We evaluated our models’ efficacy on two publicly available ADHD datasets to test their applicability and reliability. The experimental outcomes reveal an approximately 1-9% enhancement in the classification efficiency compared to existing state-of-the-art machine and deep learning frameworks. Importantly, there was no significant difference found in the performance between the two datasets. Overall, our findings demonstrate that the spWV T-F energies, as well as the expEEGNetwork’s temporal learning, can elevate the precision of EEG-based ADHD diagnostic systems. This suggests the potential usefulness of our models in real-time applications.