Abstract

Dementia, particularly Alzheimer’s disease (AD), presents major diagnostic and therapeutic challenges, with traditional assessments often limited by cost, accessibility, and ecological validity. Virtual reality (VR) has emerged as an innovative approach for evaluating and treating cognitive impairments by simulating real-world environments and tasks. A systematic search of PubMed, Google Scholar, and Scopus identified 412 studies. After duplicate removal and screening, 20 full-text articles met the inclusion criteria. Studies were assessed for VR-based diagnostic and therapeutic applications in individuals with mild cognitive impairment (MCI) and AD. VR tools included immersive navigation tasks (e.g., virtual supermarket, city navigation), cognitive training platforms (e.g., ANTaging, SCT-VR), and hybrid assessments such as the VARABOM test. Machine learning-integrated VR systems achieved 97–100% diagnostic accuracy in distinguishing MCI/AD from healthy controls, outperforming standard tests like the MoCA. VR-based interventions improved spatial navigation, visuospatial memory, executive function, and episodic recall, with some demonstrating transfer effects to hippocampal function. VR metrics, including navigational errors and time-to-completion, served as sensitive biomarkers for early detection and progression monitoring. Comparative studies found VR superior to paper-and-pencil tasks and non-VR activities in enhancing engagement and ecological validity, with adverse events being infrequent and mild. VR-based assessments and interventions demonstrated high diagnostic sensitivity, therapeutic potential, and user acceptability in MCI and AD populations. Their integration into clinical practice might enable earlier detection, targeted cognitive training, and ecologically valid monitoring, positioning VR as a transformative tool in cognitive health care.

Key words: Machine learning, dementia, mild cognitive impairment, Alzheimer’s disease, virtual reality, systematic review.