Multimodal Machine Learning Model For MCI Detection Using EEG, MRI and VR Data

Brain volume decrease is associated with neurode-generation and aging, which can manifest in some cases as mild cognitive impairment (MCI) leading to Alzheimer's disease (AD) [1]. Thus, detecting MCI at early stages is considered crucial but also challenging due to not only its subtle symptoms but also the need for safe and effective detection methods. In this matter, virtual reality (VR) environments can simulate real-world scenarios that challenge various cognitive functions such as memory, attention, and spatial awareness. Besides, magnetic resonance imaging (MRI) scans offer detailed images of brain structures and can reveal changes in brain activity patterns associated with MCI. Electroencephalography (EEG) based approaches also offer a non-invasive and cost-effective means of detecting early-stage MCI by capturing changes in brain activity and connectivity patterns associated with cognitive decline. While EEG and MRI combined with VR simulations are valuable tools for predicting MCI, advancements in machine learning (ML) facilitate feature extraction from biomedical and physiological signals, particularly in anomaly detection and classification tasks. In this study, we present a novel method leveraging a multimodal model to differentiate MCI from healthy control (HC) subjects using multimodal data comprising EEG, MRI, and VR data.