A Study on Evaluation of Dementia Progression Using CNN and Digit Images

Recently, the number of elderly persons with dementia has increased. The official report of the Japanese government says that about 20 percent of elderly persons will develop dementia in 2025. Therefore, early detection and prevention of dementia is an important issue from the viewpoint of geriatric medicine and social welfare. Currently, there are various methods for diagnosing dementia like Mini-Mental State Examination (MMSE). These are methods for oral and formal testing. Thus, subjects often become nervous and then, the accurate result is not acquired. This is one of the problems of these approaches. Therefore, the demand for a more straightforward test for dementia evaluation has been growing. In this paper, the authors discuss a scheme for the dementia evaluation method using CNN and handwritten digit images. We collected actual digit images in collaboration with welfare houses in Mie Pref. and discuss the effectiveness of the proposed method.