Who is WithMe? EEG features for attention in a visual task, with auditory and rhythmic support

Area of Science:

• Cognitive Neuroscience
• Brain-Computer Interfaces
• Signal Processing

Background:

• Understanding attention is crucial for advancing cognitive science.
• Electroencephalography (EEG) is a key tool for studying brain activity related to attention.
• Cross-subject variability in EEG data presents a significant challenge for analysis.

Purpose of the Study:

• To identify EEG data representations most strongly associated with attention.
• To evaluate the ability of these representations to manage cross-subject variability.
• To compare the performance of various EEG features against established models.

Main Methods:

• Explored univariate (time domain, recurrence plots) and multivariate (global field power, functional brain networks) EEG features.
• Investigated persistent homology features for noise robustness and cross-subject variability.
• Evaluated feature performance using Support Vector Machine (SVM) accuracy on the WithMe dataset, benchmarking against deep learning models.

Main Results:

• Raw EEG time series data outperformed all tested specific data representations.
• Deep learning approaches, capable of learning optimal features, surpassed raw EEG data.
• The effectiveness of specific EEG representations varied in handling cross-subject variability.

Conclusions:

• Raw EEG data offers a strong baseline for attention analysis.
• Advanced deep learning methods show superior performance by learning optimal features.
• Further research across diverse experimental paradigms is needed to fully understand EEG representation utility.