Novel machine learning algorithm system can accurately predict risk of developing psychosis by analyzing speech patterns of individual
The IBM Research team for Computational Psychiatry and Neuroimaging developed a method to accurately evaluate the psychiatric state of a patient using speech samples. New study involves a larger cohort and a different evaluation protocol, using which the system was able to effectively predict the onset of psychosis within two years in at-risk subjects with 80% accuracy. The study was the result of an international collaboration with Mt. Sinai School of Medicine, Columbia University, UCLA and the Universities of Melbourne, Australia, and Buenos Aires, Argentina and it was published in the journal World Psychiatry in June 2018.
As a part of the study, the system examined transcripts of interviews with subjects identified as at-risk of developing psychosis, about one quarter of which went on to develop a psychotic disorder over the following two years. The subjects were asked to explain how well they understood a short story they were given to read and the algorithm examining the speech patterns of the subjects was able to determine, with 83 percent accuracy, whether a person would go on to develop a psychosis. The broad applications for the technology are certainly in IBM’s sights with Guillermo Cecchi, from the IBM Research team, suggesting that similar computer-driven diagnostic analysis of speech patterns is being studied for conditions including Parkinson’s, Alzheimer’s and depression.
Furthermore, the team aims to develop diagnostic app to track a person’s speech patterns and suggest risk warnings for a variety of neural disorders. “In five years, what we say and write will be used as indicators of our mental health and physical wellbeing,” explains lead author on the study Cheryl Corcoran, from Mount Sinai School of Medicine. “Patterns in our speech and writing analyzed by cognitive systems will provide tell-tale signs of early-stage mental disease that prompt us to seek treatment.”