Virtual reality gameplay classification illustrates the multidimensionality of visuospatial neglect

Abstract

Abstract Brain injuries can significantly impact mental processes, and lead to hidden disabilities not easily detectable. Traditional methods for assessing these impacts are imprecise, leading to unreliable prevalence estimates and treatments with uncertain effectiveness. Immersive virtual reality has shown promise for assessment, but its use as a standalone tool is rare. Our research focused on developing and validating a standalone immersive virtual reality classification system for unilateral spatial neglect, a condition common following brain injury characterised by inattention to one side of space. Our study involved 51 brain injury inpatients and 30 controls, all engaging with “The Attention Atlas,” an immersive virtual reality game for testing visual search skills. Our classification system aimed to identify patients with neglect, “minor atypicality” (indicative of inattention not consistent enough to be labelled as neglect), or non-neglect. This categorisation was based on a simple mathematical definition, utilising gameplay to describe spatial orientation (to the left or right side) and attentional challenge (indicative of search inefficiency). These metrics were benchmarked against a normative model to detect atypical visual search, which refers to gameplay beyond the usual bounds. The combination of neglected side, orientation, and challenge factors were used to categorise neglect. We discovered a strong correlation between atypical visual search patterns and neglect risk factors, such as middle cerebral artery stroke, parietal injuries, and existing neglect diagnoses (Poisson regression incidence rate ratio = 7.18, 95% CI = 4.41-11.90). In our study, immersive virtual reality identified neglect in one-fourth of the patients (n = 13, 25.5%), minor atypicality in 17.6% (n = 9), and non-neglect in the majority, 56.9% (n = 29). This contrasts with standard assessments, which detected neglect in 17.6% (n = 9) of cases and had no intermediate category. Our analysis determined six categories of neglect, the most common being left hemispace neglect with above-median orientation and challenge scores. Traditional assessments were not significantly more accurate (accuracy = 84.3%, P = .06) than a blanket assumption of non-neglect. Traditional assessments were also relatively insensitive in detecting immersive virtual reality-identified neglect (53.8%), particularly in less severe cases and those involving right side inattention. Our findings underline the effectiveness of immersive virtual reality in revealing various dimensions of neglect, surpassing traditional methods in sensitivity and detail, and operating independently from them. To integrate immersive virtual reality into real-world clinical settings, collaboration with healthcare professionals, patients, and other stakeholders is crucial to ensure practical applicability and accessibility.