LAUSR

Early detection of Alzheimer's disease (AD) has become increasingly important. Healthy monitoring technology focusing on behavioral changes is a promising approach in this vein. Among such technologies, handwriting features measured by digital tablet devices have attracted attention as potential indicators for detecting AD and mild cognitive impairment (MCI). However, previous studies have mainly investigated features in single tasks, and it remains unclear whether combining the features of multiple tasks could improve the performance of detecting AD and MCI. In this study, we investigated features in five representative tasks used in neuropsychological tests collected from 71 seniors including some diagnosed with MCI and AD. We found that our three-class classification model improved diagnosis accuracy by up to 11.3% by combining features of multiple tasks, for a final accuracy of 74.6%. We also suggested that drawing behaviors during multiple tasks might be useful for estimating disease progression simply by utilizing the labels of disease groups.