Decoding Behavior: Utilizing Virtual Reality Digital Marker and Machine Learning for Early Detection of Mild Cognitive Impairment

Abstract

The imperative for early mild cognitive impairment (MCI) detection is underscored by the limitations of traditional biomarkers, high cost and invasiveness, and they often fail to capture behavioral changes in MCI patients associated with impaired instrumental activities of daily living (IADL). This study introduces a cost-effective, non-invasive alternative using digital markers, virtual kiosk test, which involves performing IADL tasks such as ordering food via a kiosk in virtual reality (VR) to detect MCI at an early stage. Involving 20 healthy controls and 31 MCI patients, four key behavioral features within VR digital markers effectively differentiate groups: hand movement speed, proportion of fixation duration, time to completion, and the number of errors. A machine learning model demonstrated high effectiveness with 93.3% accuracy, 100% sensitivity, 83.3% specificity, 90% precision, and a 94.7% F1-score in group differentiation. Findings suggest that observing behaviors via the virtual kiosk test within 5 minutes can be an efficient approach for early MCI detection, acting as reliable VR digital markers.