Automatic detection of attention deficit hyperactivity disorder using machine learning algorithms based on short time...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Attention deficit hyperactivity disorder (ADHD) is a common neurodevelopmental disorder in children and adolescents.
• Overlapping symptoms complicate traditional ADHD diagnosis.
• Electroencephalography (EEG) offers a non-invasive method for assessing brain activity and potential automated diagnosis.

Purpose of the Study:

• To investigate electroencephalogram (EEG) decomposition techniques for improved ADHD detection.
• To evaluate the efficacy of machine learning algorithms in distinguishing ADHD patients from healthy individuals using processed EEG data.

Main Methods:

• EEG signals were processed using Independent Component Analysis (ICA) to remove noise and artifacts.
• EEG signals were decomposed into subbands using Short-Time Fourier Transform (STFT) and Discrete Cosine Transform (DCT).
• Decomposed subbands and raw EEG signals served as input for machine learning classifiers (XGBoost, Logistic Regression, Random Forest).

Main Results:

• STFT decomposition methods demonstrated superior performance compared to DCT.
• The STFT-XGBoost model achieved 91% accuracy with 6 electrode placements (Fp1Fp2F3F4P3C3).
• Other electrode combinations and classifiers (Logistic Regression, Random Forest) also showed high accuracy, ranging from 88% to 89%.

Conclusions:

• Automated EEG analysis using STFT decomposition and machine learning shows significant potential for ADHD diagnosis.
• This technology could enhance early detection and personalized treatment strategies for ADHD.
• The proposed method offers a novel, non-invasive tool for future clinical application in ADHD diagnosis.