Publication details
EEG microstate-based classification model for assessment of the state anxiety in healthy humans
| Authors | |
|---|---|
| Year of publication | 2024 |
| Type | Conference abstract |
| Citation | |
| Description | Anxiety as an occasional emotion of worry and tension in an uncertain situation is a common harmless experience in our lives. On the other hand, an excessive fear with neurovegetative symptoms within an anxiety disorder might impair human abilities, behavior, productivity, and quality of life, if not detected and treated early. In the current work we developed a novel method of EEG-based anxiety detection and classification. The proposed method incorporates features of functional EEG microstates, which were only scarcely studied in anxiety disorders. Additional features in the time, frequency and time-frequency domain were also employed. A machine learning classifier was trained and evaluated on these features using a public EEG dataset of healthy participants (Database for Anxious States based on a Psychological Stimulation; DASPS). Using Ridge regression, a total of 100 and 102 features were selected for two-level and four-level classifications, respectively. The best classification results were obtained using the SVM classifier with a linear kernel based on Nu-Support Vector Classification, reaching an accuracy of 98% and 96% for two-level and four-level classifications, respectively. The proposed model outperformed other existing methods that assess anxiety on the DASPS dataset in terms of accuracy, sensitivity, specificity, and precision. |