Classification of auditory ERPs for ADHD detection in children

Area of Science:

• Neuroscience
• Biomedical Engineering
• Pediatrics

Background:

• Attention deficit hyperactivity disorder (ADHD) is a common childhood neurodevelopmental condition.
• Current ADHD diagnosis relies on behavioral symptoms and clinical evaluation, with electroencephalography (EEG) as a supplementary tool.
• Objective and efficient diagnostic methods are needed to improve early detection and intervention for ADHD.

Purpose of the Study:

• To develop and evaluate a machine learning approach for classifying EEG records of children with ADHD and control subjects.
• To assess the potential of portable EEG systems as an objective tool for ADHD diagnosis.

Main Methods:

• EEG signals were recorded from 47 children (22 with ADHD, 25 controls) using a portable headset during a 2-tone oddball paradigm.
• Features were extracted from auditory event-related potentials (ERPs), frequency bands, chaos quantification, and bispectral analysis.
• Support Vector Machine (SVM) and Trees algorithms were employed for classification.

Main Results:

• The SVM and Trees algorithms achieved the highest performance, with 86.36% accuracy and 95.45% sensitivity.
• The study identified relevant EEG features that differentiate children with ADHD from controls.
• Portable EEG systems demonstrated potential for objective ADHD assessment.

Conclusions:

• Portable EEG systems, analyzed with machine learning, can effectively complement standard ADHD clinical assessments.
• This approach offers an objective, time-efficient, and accessible method to support early ADHD diagnosis.
• High accuracy and sensitivity in EEG-based classification are crucial for reducing misdiagnosis and enabling timely interventions.