Metric classification of early Parkinsonism in the space of electroencephalographic features

This paper considers the problem of metric classification of early Parkinsonism in the feature space of multi-channel signals of electroencephalography (EEG). The electroencephalography feature space includes both spectral characteristics and features of rhythmic disorganization. A model of logistic regression for the classification of early Parkinsonism is studied. The model was trained on the data obtained from the experimental EEG studies in a group of patients in the 1st stage of Parkinson’s disease and a control group of subjects. Analysis of the classification logistic model was carried out using the data from 38 subjects, including 22 subjects from the control group and 16 patients in the first stage of Parkinson’s disease. Dependencies of the recall on the functional value for the control group and the patients and classification accuracies are calculated.