Separating Mental Disorders Using Vestibular Field Potentials

Abstract

Background: Absence of quantitative techniques for objectively diagnosing many brain changes associated with mental illnesses hampers early intervention and effective treatment. Known bidirectional neural pathways closely link the vestibular system and regions involved in emotion processing. Objectives: To assess whether Electrovestibulography (EVestG) can detect specific neural responses, using an ear probe and tilt chair, to provide a quantitative indirect measure assessment of brain regions and pathways frequently compromised in mental illnesses. Materials and Methods: EVestG data was collected on 38 subjects with major depression, 22 with schizophrenia, 36 with bipolar disorder and 57 matched healthy controls. Data was analyzed using the NEER algorithm to generate the average field potentials and firing patterns. Characteristic features were extracted followed by AdaBoost subset feature selection and classification for separating data into four classes. To remove the bias of working on small size population, we used 10-fold cross validation to select the best diagnostic features. The accuracy of the diagnostic features’ classification was tested using nonparametric statistical analysis. Results: EVestG signals were statistically different (P = 0.000 to 0.040) between the groups by using Kruskal-Wallis, and the best diagnostic accuracies for a four-way diagnostic group separation were on average (n = 100, 10 repeated 10-fold cross validations) 70.2% (SD = 9.6) using 10-fold cross validation. Conclusions: Comparing vestibular driven responses has the potential to be a valid and clinically useful diagnostic tool.