Could AI help early detection of schizophrenia?
It typically manifests in people aged around 20 years old, and is easy to diagnose as soon as it does, yet very difficult to treat. Still, researchers have recently demonstrated that certain anticipatory signs of schizophrenia can be detected in language before the illness truly manifests.
To accomplish this, they input conversational language excerpts from 30 000 internet users into an artificial intelligence (AI) system, which compared them to the language used by young subjects known to be at risk of psychosis. The results were that the young subjects who later developed schizophrenia presented lower semantic density and a tendency to use words associated with sound.
By combining these two variables, the scientists were able to accurately predict schizophrenia manifesting in 93% of cases! Needless to say, this could be a highly valuable tool for detecting this illness earlier and thus treating it more successfully.
1) Mina Gheiratmand, Irina Rish, Guillermo A. Cecchi, et al. Learning stable and predictive network-based patterns of schizophrenia and its clinical symptoms. npj Schizophreniavolume 3, Article number: 22 (2017).