Virtual reality exposure aided-diagnosis system for anxiety disorders: Long short-term memory architecture for three levels of anxiety recognition

Article type: Research Article

Authors: Handouzi, Wahida^a; | Maaoui, Choubeila^b | Pruski, Alain^b

Affiliations: [a] Laboratoire d’Automatique de Tlemcen (LAT), Tlemcen University, Tlemcen, Algeria | [b] LCOMS, Université de Lorraine, Metz, France

Correspondence: [*] Corresponding author: Wahida Handouzi, Laboratoire d’Automatique de Tlemcen (LAT), Tlemcen University, Tlemcen, Algeria. E-mail: [email protected]

Abstract: BACKGROUND:The COVID-19 pandemic has resulted in increased psychological pressure on mental health since 2019. The resulting anxiety and stress have permeated every aspect of life during confinement. OBJECTIVE:To provide psychologists with an unbiased measure that can aid in the preliminary diagnosis of anxiety disorders and be used as an initial treatment in cognitive-behavioral therapy, this article introduces automated recognition of three levels of anxiety. METHODS:Anxiety was elicited by exposing participants to virtual environments inspired by social situations in reference to the Liebowitz social anxiety scale. Relevant parameters, such as heart rate variability and vasoconstriction were derived from the measurement of the blood volume pulse (BVP) signal. RESULTS:A long short-term memory architecture achieved an accuracy of approximately 98% on the training and test set. CONCLUSION:The generated model allowed for careful study of the state of seven phobic participants during virtual reality exposure (VRE).

Keywords: Automatic anxiety recognition, support system, BVP signal, virtual reality, Liebowitz social anxiety scale, LSTM

DOI: 10.3233/BME-222542

Journal: Bio-Medical Materials and Engineering, vol. 34, no. 6, pp. 491-502, 2023