Towards a Machine Learning Model to Classify Cognitive Ability Using EEG Data and Virtual Spatial Navigation Task Scores in Intellectually Disabled Adults

It is crucial that tools for assessing cognitive change over time in individuals with Intellectual Disabilities address their unique needs, necessitating the development of tailored assessment methodologies. This study presents a novel machine learning approach that predicts cognitive ability using electroencephalogram data and virtual spatial navigation task performance measures, aimed at assessing cognitive change over time in adults with Intellectual Disabilities. We developed a Virtual Reality-based spatial navigation task, co-designed with young adults with Intellectual Disabilities, to ensure cognitive accessibility and engagement. The Virtual Reality task, developed targeting the PiCO Neo 3 headset, involves navigating through a virtual environment with varying levels of visual cues. Electroencephalogram data, collected using a wireless 32 channel electroencephalogram device, captures neural activity associated with navigation task performance measures. The resulting dataset integrates electroencephalogram features, navigation task performance measures, and Montreal Cognitive Assessment-Basic scores. Both traditional neural network-based approaches and Convolutional Neural Network techniques were employed to train models with the goal of classifying Montreal Cognitive Assessment-Basic ranges. Preliminary results indicate high accuracy in distinguishing between high and low Montreal Cognitive Assessment-Basic scores for the Convolutional Neural Network based models. These results demonstrate the potential of this integrated approach for the assessment of cognitive change over time in people with an Intellectual Disability.