Investigating the validity and reliability of Electrovestibulography (EVestG) for detecting post-concussion syndrome (PCS) with and without comorbid depression

Abdelbaset Suleiman, Brian Lithgow, Behzad Mansouri, Zahra Moussavi

Research output: Contribution to journalArticleResearchpeer-review

8 Citations (Scopus)


Features from Electrovestibulography (EVestG) recordings have been used to classify and measure the severity of both persistent post-concussion syndrome (PCS) and major depressive disorder. Herein, we examined the effect of comorbid depression on the detection of persistent PCS using EVestG. To validate our previously developed EVestG classifier for PCS detection, the classifier was tested with a new blind dataset (N = 21). The unbiased accuracy for identifying the new PCS from controls was found to be >90%. Next, the PCS group (N = 59) was divided into three subgroups: PCS with no-depression (n = 18), PCS with mild-depression (n = 27) and PCS with moderate/severe-depression (n = 14). When moderate/severe depression was present, PCS classification accuracy dropped to 83%. By adding an EVestG depression feature from a previous study, separation accuracy of each PCS subgroup from controls was >90%. A four and three-group (excluding mild-depression subgroup) classification, achieved an accuracy of 74% and 81%, respectively. Correlation analysis indicated a significant correlation (R = 0.67) between the depression feature and the MADRS depression score as well as between the PCS-specific feature and Rivermead Post-Concussion Questionnaire (RPQ) (R = −0.48). No significant correlation was found between the PCS-specific feature and the MADRS score (R = 0.20) or between RPQ and the depression feature (R = 0.12). The (PCS-specific and depression-specific) EVestG features used herein have the potential to robustly detect and monitor changes, relatively independently, in both persistent PCS and its depression comorbidity. Clinically, this can be particularly advantageous.

Original languageEnglish
Article number14495
Number of pages11
JournalScientific Reports
Issue number1
Publication statusPublished - 1 Dec 2018

Cite this