Machine learning and Serious Game for the Early Diagnosis of Alzheimer’s Disease

• Published in: Simulation & gaming 2022-08, Vol.53 (4), p.369-387
• Main Authors: Mezrar, Samiha, Bendella, Fatima
• Format: Article
• Language: English
• Subjects: Alzheimer's disease; Artificial Intelligence; Classification; Clinical Diagnosis; Clinical research; Clinical trials; Cognition & reasoning; Cognitive ability; Cognitive functioning; Cognitive impairment; Cognitive skills; Computer & video games; Diagnosis; Educational software; Episodic memory; Executive Function; Games; Impairment; Intellectual Disability; Machine learning; Medical diagnosis; Memory; Short term memory; Signs and symptoms; Teaching Methods; Thinking Skills; Usability
• ISSN: 1046-8781; 1552-826X
• DOI: 10.1177/10468781221106850

Background and Aim Aging people can suffer from cognitive impairments with a range of symptoms, including memory, perception, and difficulty in solving problems called Alzheimer’s disease (AD). The early detection of Mild Cognitive Impairment (MCI), which can develop AD, plays a major role in the management of patients to slow the decline in cognitive function, as treatments are effective at an early stage of the disease course. For this purpose, advanced computer technologies can provide a tool for the early detection of AD and prediction of disease progression. This article presents a serious game, including 16 mini-games that aimed at detecting AD or MCI in the mild stage. Based on gamification techniques and machine learning (ML), by overcoming the limitations of traditional tests. This gamified cognitive tool, entitled AlzCoGame, evaluates the main cognitive domains considered to be the most pertinent indicators in diagnosing cognitive impairments: working memory, episodic memory, executive functions, Visio-spatial orientation, concentration, and attention. Results and Conclusion Six predictive ML models have been implemented using the AlzCoGame dataset. We used the K-fold cross-validation and classification metrics to validate the model's performance. Based on the results of the pilot study, the best overall performance was obtained by the RF classifier with average Sensitivity = 0.89, Specificity = 0.93, Accuracy = 0.92, F1-Score = 0.91, and ROC = 0.91. We can deduce that including machine learning techniques and serious games could help improve certain aspects of the clinical diagnosis of cognitive impairment. Moreover, clinical trials are required to prove the impact of this gamified program on cognitive skills and evaluate usability measures.