Dementia can creep on slowly and develop without much notice in many patients. Monitoring these changes is not easy, typically requiring professional experts to interpret faint symptoms gauged through standard questionnaires. Patients undergoing such testing tend to get used to it and end up improving their test taking ability, skewing the results.
Researchers at Osaka University and Nara Institute of Science and Technology in Japan have now developed a similar, but purely computer-driven approach that asks a user questions and then comprehensively monitors the responses. As the user answers questions, the software identifies different parameters of the speech, such as its tone, speed, intonation, and usage of verbs and nouns. At the same time, a camera watches the person’s face and measures a number of visual parameters as well.
All this data is fed through a machine learning system that was trained with recordings of individuals confirmed to suffer from dementia and those that seem to be free from it.
In an initial study of the system involving 12 healthy people and 12 people diagnosed with dementia (according to the criteria in Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV), the system had a 90% accuracy rate relying only six questions each only two to three minutes long.
Study in IEEE Journal of Translational Engineering in Health and Medicine: Detecting Dementia Through Interactive Computer Avatars…
Via: Osaka University…