Diagnosis of ADHD in children from EEG signals using amplitude modulation features

Attention deficit hyperactivity disorder (ADHD) is a common neurodevelopmental disorder in children and accurate diagnosis of this disorder is very important in preventing later complications in them. It has been shown that ADHD can be diagnosed with high accuracy by analyzing electroencephalogram (EEG) signals. This study introduces a new framework for diagnosing ADHD based on amplitude modulation (AM) features of EEG signals.

We utilized a public EEG dataset from 61 ADHD and 60 healthy control (HC) subjects. Signals were decomposed into five frequency bands using discrete wavelet transform (DWT). From each band, we extracted a combined feature set comprising eight conventional features and six AM features. A two-stage feature selection (t-test filter followed by sequential forward selection) was applied. The diagnostic performance was evaluated using multiple classifiers (KNN, SVM, AdaBoost, ANN, DT, and NB) employing Monte Carlo cross-validation.

The integration of AM features yielded a significant and consistent improvement across all classifiers and mother wavelets. The highest accuracy of 99.46% was achieved using a KNN classifier with features derived from the coif2 wavelet. Analysis of the optimal feature subset revealed a high density of AM features in the gamma and beta bands and a predominance of discriminative information from frontal brain regions. The framework outperformed recent state-of-the-art methods on the same dataset.

This study establishes AM features as a powerful and interpretable EEG biomarker for ADHD, directly quantifying disrupted cross-frequency neural coupling associated with the disorder. By achieving superior diagnostic accuracy with simpler and more interpretable models, the proposed framework enables more objective and clinically useful diagnostic tools.