Combining serious games and machine learning to assess the cognitive state of seniors using the cogniplat platform

Early identification of Mild Cognitive Impairment (MCI), a potential precursor to Alzheimer’s Disease (AD), is crucial in decelerating the progression of cognitive decline within elderly adults. This study presents a collection of serious games aimed at detecting MCI more efficiently and with potentially less bias, overcoming the limitations of traditional, time-consuming, and often late diagnostic methods. This paper specifically examines the effectiveness of the Cogniplat platform, enhanced with machine learning, as a novel tool for assessing the cognitive state of seniors using their performance data from these games. The proposed methodology involves formulating machine learning models to evaluate the predictive effectiveness of gameplay-derived features on the Cogniplat platform. The study with 27 senior participants demonstrates that in-game data from two sessions can accurately classify and predict cognitive conditions, with the Random Forest model outperforming other algorithms in distinguishing normal cognitive function from MCI, achieving a 99% accuracy rate. This preliminary study also provides initial insights into the measurement validity of the approach, focusing specifically on aspects of construct, criterion, and external validity. The research emphasizes the advantages of integrating serious games with machine learning, a method that not only enriches the cognitive assessment experience but also provides more precise and timely insights.