The Symptom Classification of Alzheimer’s Disease Based on Machine Learning: A Functional Near-infrared Spectroscopy Study

This paper demonstrates a technology for early diagnosis of Alzheimer's disease through a portable fNIRS device so that Alzheimer's disease, which has become a serious social problem in the aging society, can be easily screened and appropriately treated to patients. To prove the hypothesis, the brain signals of normal people, mild cognitive impairment patients, and Alzheimer's disease patients obtained from the fNIRS device were classified through machine learning using an artificial neural network model. Participants in the experiment performed behavioral tasks based on working memory, and changes in cerebral blood flow occurring in the prefrontal cortex of each participant during the task were collected using a portable fNIRS device. Using the collected data, features based on functional network analysis of the brain were extracted, and Alzheimer's disease diagnosis performance was evaluated using machine learning algorithms trained through the extracted features’ data. As a result, normal people, MCI patients, and Alzheimer's disease patients could be classified through 2-class classification. Through this result, this paper is able to show the possibility of early screening of Alzheimer's disease severity by using portable fNIRS device.