Automated analysis of natural speech in personalized treatment for major depressive disorder

Listening to a person's speech is a key component of clinical assessment in psychiatry. Speech rhythm and melody reflect the speaker's mood in the moment. The content and form of speech reflect the speaker's thought process, their interests and motivations. Here we propose to apply analysis of speech to help select effective treatment for individuals with depression. Several types of treatment, including antidepressant medication and psychotherapy, can effectively treat depression, but it is difficult to predict which treatment will work for whom. In the framework of a large Canadian network, we will test whether an artificial intelligence classification of a person's speech can improve on clinical intuition in selecting a treatment that a given person with depression is likely to benefit from. Specifically, we will test whether information drawn from a five-minute speech sample could predict the outcomes of treatment with psychotherapy, magnetic brain stimulation, antidepressant and add-on medication. We will also examine speech analysis as an objective measure of depression improvement. We will develop an automated speech assessment tool that will facilitate the use of speech assessment in further research and clinical applications.