Event Related Potential Analysis Using Machine Learning to Predict Diagnostic Outcome of Autism Spectrum Disorder
| Autor: | Mayada Elsabbagh, Stefon J. R. van Noordt, Lina Abou-Abbas |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
genetic structures
medicine.diagnostic_test business.industry Brain activity and meditation Decision tree learning Decision tree Electroencephalography medicine.disease Machine learning computer.software_genre behavioral disciplines and activities Event-related potential Autism spectrum disorder medicine Autism Eye tracking Artificial intelligence business Psychology computer |
| Zdroj: | Bioengineering and Biomedical Signal and Image Processing ISBN: 9783030881627 BIOMESIP |
| Popis: | Identifying diagnostic biomarkers for autism spectrum disorder (ASD) is one of the challenges in autism research today. Recent studies using visual Event Related Potentials (ERPs) have identified abnormal patterns of brain activity in high-risk infants who go onto ASD diagnosis. In this study, we used well-established ERP components related to face processing. Features of these ERPs were used to explore the performance of machine learning algorithms in classifying ASD diagnostic outcomes. Data were used from the EEG Integrated Platform (EEG-IP). ERPs were recorded from six-months infants in response to static faces that dynamically changed between direct and averted eye gaze. Amplitude and latency measures of prominent ERP peaks including P100, N290 and P400 were derived across five scalp regions. Difference Scores between stimulus conditions (direct versus indirect eye gaze, toward versus away and face versus noise) were considered. Features were selected by weight correlation and used as inputs to three classifiers: k-Nearest Neighbor, Support Vector machines and Decision tree. Performance of these classifiers was compared. The results showed that the Decision tree classifier had the greatest average accuracy rate of 78.09% in classifying ASD diagnosis in high-risk infants. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|