Using EEG to Discriminate Cognitive Workload and Performance Based on Neural Activation and Connectivity
MIT Lincoln Laboratory Lexington United States
Pagination or Media Count:
A major goal of noninvasive brain sensing is to ascertain both the workload and the efficacy of cognitive processing. Realizing this goal will assist in monitoring cognitive readiness under different levels of cognitive workload and fatigue. Our approach to discriminating a persons cognitive state is predicated on the idea that cognition depends on coordinated neural activations, operating over a range of frequencies, that link functional networks across multiple brain regions. Therefore, our approach focuses on characterizing neural activation and connectivity patterns across the brain within multiple frequency bands. In each band, neural activations are characterized using spatial distributions of power across EEG channels, and neural connectivities are characterized using the eigenspectra of EEG connectivity matrices. The connectivity matrices are constructed using two measures coherence and covariance. We use an auditory working memory task to vary cognitive workload by altering the number of digits held in memory during the simultaneous retention of a sentence in memory. Cognitive efficacy is assessed based on accuracy in recalling digits from memory. A Gaussian classifier is used to discriminate cognitive load and performance from EEG recorded during each experimental trial, and quantify discrimination accuracy with the area under the receiver operating characteristic curve AUC statistic. For cognitive load discrimination, AUC values of 0.59, 0.56, and 0.60 are obtained using power-, coherence-, and covariance-based feature sets, respectively. For cognitive performance discrimination, AUC values of 0.49, 0.62, and 0.63 are obtained for the same feature sets.
- Stress Physiology
- Anatomy and Physiology