Depression is a common mental illness across the globe, with more than 300 million patients. Depression is a very popular disease not only in stressful advanced countries but also in developing countries. Meanwhile, it is said that there are more than 4 million patients with Parkinson's disease and more than 45 million patients with dementia in the world. These numbers continue to increase with the aging of the society. Symptoms of these three diseases are similar at early onset, but technical development for objective evaluation to distinguish them has been slow.

We developed techniques for screening and monitoring depression symptoms using the voice. Since long ago, attempts have been made to detect depressive symptoms by voice, but our technique is quite unique in that it is based on speech emotion recognition technology.

In depression, expression of the emotions is suppressed overall, and sadness lasts for a long time. We focused on this emotional change in depression, and succeeded in judging depression by capturing the change in emotion using the voice. Voice screening and monitoring have many advantages: they are noninvasive and can be measured from a remote place. As no special equipment is required, there is no need for large-scale infrastructure development, therefore installation and maintenance costs are very low.

Our technology has already reached a practical level, and we have accumulated various medical evidences. MIMOSYS is a system to monitor the state of mind using the voice with smartphones installed with our technology. A multilingual version is now open to the public as a social implementation study by the University of Tokyo. In Japan, Hitachi Systems and PST began to offer this technology through the cloud system. By using the cloud system, it became possible to monitor individuals from a remote place as a group, and utilization in the field of occupational medicine has already begun.

Currently, we are expanding the field of development to diseases other than depression, e.g. Parkinson's disease and dementia. Evaluation using the emotions may be a humane approach, but it is difficult to grasp the differences in detail. Therefore, in order to expand target diseases, we developed a new method that does not involve emotion. Prototypes of the algorithm are near completion, and we are now accumulating clinical data and verifying the algorithms iteratively. We believe that screening, monitoring and prognostic prediction of multiple diseases will become possible in the near future.

In this presentation, I will introduce a part of the voice pathophysiological analysis, what we can call an innovative technology for health care.