Discriminating between ADHD adults and controls using independent ERP components and a support vector machine: a
Andreas Mueller1, Gian Candrian1, Venke Arntsberg Grane2
1Brain and Trauma Foundation Grisons, Poststrasse 22, 7000 Chur, Switzerland.
Nonlinear Biomedical Physics
|July 21, 2011
Summary
Independent component analysis of event-related potentials (ERPs) effectively differentiates adult ADHD patients from controls. This machine learning approach achieves high accuracy, aiding in ADHD diagnosis.
Area of Science:
- Neuroscience
- Psychiatry
- Machine Learning
Background:
- Numerous event-related potential (ERP) studies exist for attention-deficit hyperactivity disorder (ADHD), primarily in children.
- Independent Component Analysis (ICA) separates mixed ERPs into independent source signals representing functional processes.
- Support Vector Machines (SVM) are machine learning classifiers used for pattern recognition.
Purpose of the Study:
- Investigate the use of independent ERP components for differentiating adult ADHD patients from controls.
- Select the most informative feature set for classification.
- Validate the predictive power of the SVM classifier on an independent ADHD sample.
Main Methods:
- Age-matched adults (75 ADHD, 75 controls) performed a visual go/no-go task.
- ERPs were decomposed into independent components using ICA.
- A selected set of independent ERP component features was used for SVM classification.
Main Results:
- A 10-fold cross-validation achieved 91% classification accuracy.
- Validation on an independent ADHD sample yielded 94% classification accuracy.
- Latency and amplitude measures from components related to inhibitory and executive functions were key differentiators.
Conclusions:
- ERPs, when analyzed with advanced methods like ICA and SVM, can significantly aid in ADHD diagnosis.
- This approach offers a promising tool for objective ADHD assessment in adults.
