Electroencephalogram (EEG) Based Prediction of Attention Deficit Hyperactivity Disorder (ADHD) Using Machine
Jun Won Kim1, Bung-Nyun Kim2, Johanna Inhyang Kim3
1Department of Psychiatry, Daegu Catholic University School of Medicine, Daegu, Republic of Korea.
Neuropsychiatric Disease and Treatment
|February 18, 2025
Summary
Machine learning analysis of electroencephalogram (EEG) data shows promise for diagnosing Attention Deficit Hyperactivity Disorder (ADHD). This non-invasive method achieved high accuracy, identifying key brainwave patterns for ADHD diagnosis.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Biomedical Engineering
Background:
- Attention Deficit Hyperactivity Disorder (ADHD) diagnosis faces challenges in accuracy and timeliness.
- Electroencephalogram (EEG) offers a non-invasive method to study brain activity.
Purpose of the Study:
- To evaluate the effectiveness of combining EEG data with machine learning for enhancing ADHD diagnostic accuracy.
- To explore the potential of machine learning models as objective tools for ADHD diagnosis.
Main Methods:
- 168 participants (107 ADHD, 61 neurotypical) underwent EEG recording.
- EEG data analyzed across five frequency bands (delta, theta, alpha, beta, gamma).
- Extreme Gradient Boosting (XGBoost) classifier used with Leave-One-Subject-Out (LOSO) cross-validation.
Main Results:
- The XGBoost model achieved 90.81% test accuracy and an F1-score of 0.9347.
- Data augmentation generated 2434 ADHD and 1060 neurotypical EEG segments.
- SHAP analysis identified middle beta frequency features (O1 electrode) as significant for classification.
Conclusions:
- EEG-based machine learning models demonstrate potential for accurate and interpretable ADHD diagnosis.
- The study highlights the novelty of combining SHAP analysis, data augmentation, and LOSO cross-validation.
- Further research with larger, diverse datasets is recommended for clinical validation.
