Finding critical features for predicting quality of life in tablet-based serious games for dementia

Jeehoon Cha, Jan-Niklas Voigt-Antons, Carola Trahms, Julie Lorraine O'Sullivan, Paul Gellert, Adelheid Kuhlmey, Sebastian Möller, Johanna Nordheim

In: Quality and User Experience 4 1 Pages 1-20 Springer 2019.


While the number of dementia cases is steadily increasing, as of today no medication has been developed to cure its underlying causes. Instead, the focus in treatment has shifted to improve quality of life (QoL) for people with dementia (PwD). To this end, some non-pharmacological treatments such as exercising, socializing, and playing games have received increasing attention. PflegeTab is a tablet-based application developed for this purpose. It includes a number of services such as cognitive training games, everyday activity training games, emotional applications, and a biographical picture album. In the present paper, we explore the possibility of QoL prediction for PwD using data collected while nursing home residents played games in PflegeTab ($$N = 81$$N=81). Using features generated from the data and applying linear discriminant analysis for classification, our approach obtained an average accuracy of 74.80% on predicting QoL ratings when measured by Monte Carlo cross-validation. Furthermore, this paper investigates which features were dominant for the classification (prominent features were e.g. time needed for task completion) and briefly discusses how the results might be utilized for managing general QoL of PwD.

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz