Detection of personal and situational anxiety based on electroencephalography using machine learning

The purpose of research . Nowadays electroencephalography is successfully used for diagnostics and rehabilitation of psychological and cognitive disorders. At the same time, practical healthcare meets with the difficulties related to the weakness in objective assessment of the severity of anxiety states and the requirement to consider the individual characteristics of patients. These features are difficult to formalize, but can be taken into account when implementing machine learning methods. The purpose of the study is to evaluate the possibility of using machine learning technolo- gies to identify the level of anxiety correlated with the Spielberger-Khanin method, according to electroencephalography data. Methods . To identify anxiety in the main and control groups, the Spielberger-Khanin method was used, which allows for differential measurement of anxiety as a personal trait. Alpha rhythms were recorded using 6-channel electroen- cephalography. The CatBoost library, which implements a gradient boosting algorithm with loss function minimization, was used as a machine learning tool. Results . The experiment involved 92 respondents, divided into three groups based on the testing results. Before test- ing, all subjects were in a state of rest for 1.5 minutes with the EEG recording turned on. Then they were asked to take a test to determine personal anxiety with synchronous recording of alpha rhythm parameters. Based on the results of testing, respondents were divided into three groups in accordance with the level of a certain anxiety. The study revealed a relationship between different levels of anxiety according to the Spielberger-Khanin scale and the type of electroen- cephalogram in subjects, which makes it possible to move from testing patients to recording and interpreting EEG. Conclusion . The study revealed a positive relationship between different levels of personal anxiety according to the Spielberger-Khanin scale and the type of electroencephalogram in subjects with an accuracy of up to 14%. It is shown that when determining anxiety, it is possible to replace the test based on the Spielberger-Khanin method for determining anxiety by EEG using machine learning technology.