AI-Based Immersive AR/VR Platform for Alzheimer’s Disease Symptom Tracking and Monitoring

Background: In pre-clinical Alzheimer’s disease (AD), the entorhinal cortex and hippocampus are some of the first brain regions affected. The accumulation of amyloid beta (Aβ) and phosphorylated tau (p-tau) neurofibrillary tangles first occurs in the entorhinal cortex before spreading to the hippocampus. As neurodegeneration progresses, rapid hippocampal atrophy drives cognitive decline. Given that these pathological changes impair learning, spatial navigation, memory, and executive function, reliable cognitive assessment is critical for detecting AD in its early stages. While traditional paper-and-pen cognitive screening tools are effective for measuring cognitive function, their limited ecological validity reduces their ability to assess how cognition is applied in real-world situations. Conversely, while biological biomarkers offer valuable insights into AD pathology, they are limited by cost, invasiveness, and an inability to measure functional performance. Recently, virtual reality (VR) has been emerging as a promising tool for cognitive screening, capable of detecting subtle cognitive shifts before traditional clinical assessments. Furthermore, preliminary studies suggest that individuals with AD can safely tolerate intensive VR-based cognitive training. Prior research has explored the potential of VR for cognitive assessment and rehabilitation, including AI-driven adaptive difficulty, cognitive stimulation methods, and functional assessments modeled after instrumental activities of daily living (iADL). However, these approaches have yet to be integrated. We intend to develop an AI-integrated VR platform capable of modulating task difficulty in real time based on patient performance and physiological changes. This platform will be used to determine which virtual environments and tasks are most effective for the early detection of AD-related neurodegeneration, and to evaluate whether dynamic task adjustment enhances cognitive rehabilitation.

Methods: We conducted an extensive literature review spanning the pathophysiology of AD, existing VR-based cognitive assessment and rehabilitation tools, and AI models suitable for integration into the proposed platform.

Results: Based on our review, we extracted 15 tasks for integration into our AI/VR system and developed a preliminary structural framework. Tasks reflecting iADLs were identified as the most effective for improving cognitive function, as they engage multiple cognitive domains simultaneously and simulate real-life challenges. Specifically, embedding path integration components into virtual iADL tasks may offer a highly sensitive marker for early cognitive impairment. Furthermore, pairing VR-based assessments of cognitive function with blood-based p-tau biomarkers could optimize early AD detection by linking behavioral evidence of cognitive impairment with biological markers of pathology. Five AI models were identified for potential use in modulating task difficulty based on real-time user sensorimotor and task-completion data, as well as supporting rehabilitation.

Conclusion: Our findings demonstrate that user performance on spatial navigation and path integration tasks correlates with early AD biomarkers, highlighting their potential for the early detection of decline and the tracking of cognitive rehabilitation. In future phases of this project, we plan to collaborate with a development team to implement these features into our proposed AI-integrated VR platform. We will evaluate the platform’s efficacy in identifying early-stage AD and improving cognitive function by correlating performance metrics with patient Aβ and p-tau levels, ultimately aiming to build an intensive training system that measurably enhances patient cognitive capacity.