Classification of attention deficit hyperactivity disorder using machine learning on an EEG dataset.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Developmental Psychology

Background:

• Attention Deficit Hyperactivity Disorder (ADHD) is a common neurodevelopmental disorder in children.
• ADHD's etiology is complex, involving genetic, environmental, and neurological factors.
• Electroencephalography (EEG) records brain electrical activity, reflecting cognitive processes.

Purpose of the Study:

• To analyze EEG signals for classifying Attention Deficit Hyperactivity Disorder (ADHD) in children.
• To evaluate the effectiveness of machine learning algorithms in ADHD diagnosis using EEG data.
• To explore potential for classifying ADHD subgroups and severity.

Main Methods:

• Analyzed EEG signals from 121 children (ADHD and control groups) under cognitive tasks.
• Extracted features using Euclidean distance, a common metric in machine learning.
• Trained four supervised machine learning algorithms: linear regression, random forest, extreme gradient boosting, and K-nearest neighbor (KNN).

Main Results:

• The K-nearest neighbor (KNN) algorithm achieved the highest classification accuracy among the tested algorithms.
• Performance varied across different frequency bands of the EEG signals.
• The study identified KNN as a promising method for ADHD classification.

Conclusions:

• Machine learning, particularly KNN, shows significant potential for ADHD classification using EEG data.
• Hyperparameter tuning can further enhance the accuracy of the KNN algorithm.
• This approach could be extended to classify ADHD subgroups and assess disorder severity.