Identification of Alzheimer's Disease Patients Based on Oral Speech Features

Screening Alzheimer's disease (AD) patients quickly and non-invasively is of great challenge in the field of clinical medical. In this study, a method based on oral speech features for AD patients identification was proposed. AD (27 people), MCI (Mild Cognitive Impairment, 42 people) and HCs (Healthy Controls, 25 people) were recruited to make a detailed description of the Cookie Theft picture. Linguistic features and acoustic features were extracted manually and automatically respectively from the speech. Based on these features, Support Vector Machine (SVM) classifier was adopted to model and identify AD patients. The results based on linguistic features and acoustic features reached an accuracy of 94.2% and 93.62% respectively. The results suggested that a validated oral task could be further used with automatic algorithm in AD identification. This study is the first study to classify Chinese AD patients with linguistic features and acoustic features, sending important message for rapid AD early screening based on a quick ecological oral task.