Enhancing ADHD Diagnosis Using Deep Learning: A Robust Model for Analyzing EEG Signals in Children

Abstract

Inattention, hyperactivity, and impulsivity are the hallmarks of Attention-Deficit/Hyperactivity condition (ADHD), a common neurodevelopmental condition that affects both children and adults. More objective and effective procedures must be investigated because traditional diagnostic methods mostly rely on subjective behavioural assessments and questionnaires. This study looks into using DL approaches to improve the precision and efficacy of diagnosing ADHD. We analyzed various cognitive conditions using a dataset of EEG signals from 121 children, which includes unbiased data from both ADHD and control groups. We extracted key features using Euclidean distance for subsequent model training. A good result of 85% on the test set was achieved by the suggested model architecture, which includes the Conv1D, MaxPooling1D, GRU, Flatten, Concatenate, Dense, and Dropout layers. The model displayed high precision, recall, and F1-scores, indicating its robustness in identifying ADHD symptoms. The lightweight architecture, with a total of 141,057 parameters, ensures suitability for real-time applications, potentially assisting clinicians in making more accurate and timely diagnoses. Future work will focus on expanding the dataset, exploring advanced architectures, and integrating multimodal data to further enhance predictive capabilities.