Assessing Cognitive Function by a Virtual Reality Serious Game System in Schizophrenia Patients at Stable Phase: A prospective nonrandomized clinical trial

Abstract Background: Cognitive impairments are enduring characteristics and core deficits in patients of schizophrenia, particularly impairments in working memory and executive function. It is very important to accurately assess the true cognitive function of schizophrenia patients, but the existing assessment tools have shortcomings in the aspects of ecological assessment, immersion and fun assessment. Virtual reality (VR) technologies and serious games have shown potential in these aspects. Methods: We conducted a prospective non-randomized controlled trial to explore the application of a VR serious game system, integrating VR technology and serious game, in assessing cognitive functions in patients with SZ. Forty-two patients with SZ and sixty-five healthy controls (HCs) were enrolled. The system recorded and scored the participants' performance in the VR serious game. We compared the performance between patients with SZ and HCs, and further explored its association with the scores by Brief Assessment of Cognition in Schizophrenia (B-CATS). Further, machine learning models were established to classify patients with SZ and HCs. Results: Significant differences were observed in the performances of VR serious game between the two groups. The performances of VR serious game was consistent with the scores by B-CATS, particularly in the SZ group. Machine learning models effectively classified patients with SZ and HCs based on results of VR serious game, with Receiver Operating Characteristic (ROC) curve areas of 0.804 using Logistic Regression Machine model (LRM) and 0.819 using Support Vector Machine (SVM) model. Conclusion: These findings emphasized the potential of utilizing VR serious game system as a cognitive assessment tool for patients with SZ at stable phase.