Machine Learning Classification of Event-Related Brain Potentials during a Visual Go/NoGo Task

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning Applications

Background:

• Machine learning (ML) methods are increasingly used to analyze complex biological signals like electroencephalograms (EEG).
• ML excels at processing large datasets to identify patterns potentially missed by human analysis.
• Event-related potentials (ERPs), extracted from EEG, reflect brain activity in response to specific events and are crucial for understanding cognitive processes.

Purpose of the Study:

• To investigate the accuracy of ML algorithms in classifying brain electrical activity, specifically ERPs, evoked during a visual Go/NoGo task.
• To compare the performance of six different ML algorithms in distinguishing between trial types based on ERPs.
• To evaluate the impact of dimensionality reduction through parameterization on ML classification accuracy.

Main Methods:

• Six ML algorithms were employed to classify ERPs elicited during a visual Go/NoGo task.
• Raw EEG signals were used to train predictive models.
• A continuous-time subspace-based system identification algorithm was used to fit dynamic state space models, with transfer function parameters serving as data substitutes for dimensionality reduction.

Main Results:

• All tested ML algorithms achieved high accuracy in classifying ERPs associated with different trial types.
• Classification accuracy remained high even after the parameterization process, indicating the robustness of the ML models.
• The study demonstrates the effectiveness of ML in analyzing neural signals for cognitive event classification.

Conclusions:

• ML methods are highly effective for accurately classifying event-related potentials from EEG data.
• Dimensionality reduction via parameterization does not compromise, and may even support, accurate classification of neural signals.
• This approach holds significant potential for advancing the analysis of brain activity in various cognitive and clinical applications.