ReSmart-15: A Soft Information Gain Based Questionnaire for Early Dementia Detection

Hyeseong Park¹, Myung Won Raymond Jung¹, Ji-Hye Kim² and Uran Oh³*

Published: July 01, 2022

DOI: 10.55162/MCET.03.060

View Pdf

Abstract  

As the number of people affected by dementia increases, there is a need for diagnosing potential symptoms of early dementia, and questionnaires such as the mini-mental state examination (MMSE) are widely used for early dementia detection. To build a more effective questionnaire, we propose ReSmart-15, a dementia detection questionnaire that includes daily behavior-based questions in five categories (i.e., attention, spatial ability, spatiotemporal ability, memory, and thinking ability). To evaluate the effectiveness of each question in detecting early dementia, information gain can be used to rank their contributions. However, the current information gain-based method requires hard classification results such as whether the patient had been diagnosed with early dementia or not. In this paper, we propose a “soft information gain” based ranking system where each patient is diagnosed with an early dementia probability (from 0 to 1), not with a hard decision of early dementia (0 or 1). We conducted an experiment to test the effectiveness of ReSmart-15 compared to MMSE and found that the top 2 questions were from ReSmart-15, and 60 percent of the ReSmart-15 questions were in the top 10.