"An automatic speech analysis approach for early screening of Alzheimer’s Disease"
Alzheimer's disease (AD) is the sixth leading cause of death in the United States and affects at least 5 million individuals. Currently, there is no cure for AD. So the early detection of the disease before irreversible brain damage and mental decline has occurred is critical to prevent or delay the progression of the disease. Biomarkers and medical images are believed to offer promising paths to indicate early stages of AD, however, the expense of the exams makes them hardly ideal pre-screening approaches.
Another prominent sign of AD is language dysfunction. Some aspects of language are affected at the same time or even before the memory problems emerge potentially leading to early diagnosis. Therefore, we propose an end-to-end natural language processing pipeline to automatically identify AD subjects using short narrative samples from the DementiaBank dataset. Both statistical and neural models are applied to extract linguistic features from narrative transcripts. Classifiers are then built upon those features to distinguish AD subjects from elderly cognitive normal controls. Additionally, we also aim to identify and interpret the key linguistic characteristics of AD, including phonemic and semantic factors.