Multi-Tasking and Choice of Training Data Influencing Parietal ERP Expression and Single-Trial Detection-Relevance...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Cognitive Science

Background:

• Event-related potentials (ERPs) are crucial for brain-computer interfaces (BCIs) enabling communication and control for individuals with motor disabilities.
• Single-trial electroencephalography (EEG) classification in BCIs shows a strong link between ERP detection accuracy and their expression.
• BCIs represent a significant advancement in neuroscience and clinical applications, offering new methodologies.

Purpose of the Study:

• To investigate how the selection of training data and classifier transfer impacts the interpretability of single-trial ERP detection.
• To evaluate the effectiveness of ERP detection under different task conditions (simple vs. dual-task) and electrode configurations.
• To explore the potential of classifier transfer between simple and dual tasks for improved BCI performance.

Main Methods:

• Subjects performed a visual-motor oddball task under simple and dual-task conditions, involving motor-task relevant targets, irrelevant deviants, and frequent standards.
• Average ERP analysis and single-trial detection analysis were conducted using varying numbers of parietal electrodes (62, 6, and 2).
• Classifier transfer was examined between the simple and dual-task conditions.

Main Results:

• Parietal positive ERPs, particularly those evoked by target stimuli, showed stronger expression under the dual-task condition, suggesting increased task emphasis and cognitive coordination.
• Highest classification performance for targets was achieved irrespective of the number of parietal electrodes used (62, 6, or 2).
• Detection performance for targets over standards was higher under the dual-task condition compared to the simple-task condition when training on data from 2 parietal electrodes.

Conclusions:

• P300 and overlaying parietal positive ERPs can be successfully detected during concurrent motor activity, validating single-trial ERP detection.
• This research supports the use of single-trial ERP detection to infer a patient's attentional state during therapeutic interventions.
• Classifier transfer enhances classification performance, especially when training data includes diverse examples from dual-task conditions.