Detecting cognitive traits and occupational proficiency using EEG and statistical inference.

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Machine learning (ML) is frequently applied to electroencephalogram (EEG) data for detecting neurological conditions.
• Subject-independent classification of specific cognitive skills in healthy individuals using EEG remains a significant challenge.

Purpose of the Study:

• To compare the subject-independent classification performance of three distinct ML pipelines.
• To assess the efficacy of ML in distinguishing cognitive traits based on educational background using EEG.

Main Methods:

• Recorded 128-channel EEGs from 26 healthy volunteers performing arithmetic, logical, and verbal tasks.
• Compared three pipelines: supervised Riemann projections with logistic regression, and handcrafted power spectral features with LightGBM.
• Classified participants into mathematics and humanities specialist groups based on education and occupation.

Main Results:

• All three ML pipelines achieved balanced accuracy significantly above chance for classifying educational type (0.84-0.89).
• The pipelines successfully distinguished mathematical proficiency based on learning experience.
• Different pipelines offered varying trade-offs between classification performance and model explainability.

Conclusions:

• ML approaches show promise for recognizing individual cognitive traits from noninvasive EEG data.
• Subject-independent classification of cognitive skills is feasible with advanced ML techniques.
• EEG-based ML can differentiate cognitive profiles related to specialized learning.