Ensemble learning of attention-based BiLSTM networks for ADHD detection from EEG signals code

Early ADHD detection is vital for children's mental health. Traditional methods are subjective and inconsistent. This paper presents an ensemble learning approach using a Parallel Attention-Based BiLSTM (PABiLSTM) model on EEG data. Spectrograms, fractal dimensions, and recurrence plots processed via ResNet-50 enhance feature extraction and classification accuracy. Exhaustive experiments on two datasets results in the accuracy drops of 98.91% and 99.10% on respective datasets. Although deep learning applied to EEG data enhances the diagnostic accuracy, it also gives a more objective way to determine ADHD, and hence is a very promising way to develop better intervention strategies and treatment pathways.