Development of voice biomarkers in psychiatry and neurology

Abstract

We developed voice biomarkers for depression, Parkinson disease, dementia, and mild cognitive impairment. For depression, Parkinson disease, and dementia, we tested our original index, multiple regression analysis, logistic analysis, and machine learning, all of them were sufficiently accurate. In the case of mild cognitive impairment, good accuracy was obtained by comparing the change of voice in a single task. The voice biomarkers for mild cognitive impairment were made publicly available on the Web. In addition, voice biomarkers for depression and Parkinson disease can also be used to differentiate between these two disorders. In the future, it is necessary to study voice acquisition methods for differentiation of the three diseases including mild cognitive impairment.