New tools based on AI can help diagnose schizophrenia

A language-based computational assay of conceptual disorganization in schizophrenia.


Schizophrenia is a debilitating neuropsychiatric disorder whose core clinical features are thought to reflect abnormalities in internal conceptual representations. Blood tests and brain scans play a tiny part in psychiatric diagnosis and are almost exclusively used to complement talking with patients and others close to them.

However, this lack of specificity makes it difficult to monitor treatment and get a deeper knowledge of the origins of mental disease.

A new study by the UCL Queen Square Institute for Neurology aims to understand how automated language analysis could help doctors and scientists diagnose and assess psychiatric conditions. Scientists have developed new tools based on AI language models to characterize subtle signatures in the speech of patients diagnosed with schizophrenia.

26 participants with schizophrenia and 26 control participants were given five minutes to complete two verbal fluency tests, which involved naming as many words as possible that fit into the ” animals ” category or began with the letter “p”.

The scientists employed an AI language model trained on vast amounts of online content to represent word meaning similarly to humans to analyze the responses provided by participants. They investigated whether the AI model could anticipate the phrases that subjects recalled on the spur of the moment and whether this predictability was altered in schizophrenia patients.

They discovered that patients with the most severe symptoms of schizophrenia had the most significant discrepancy between the answers supplied by controls and those predicted by the AI model.

According to the researchers, the difference may be related to how the brain develops connections between memories and concepts and stores this knowledge in so-called “cognitive maps.” In a second section of the same study, the scientists used brain scanning to analyze brain activity in brain regions involved in learning and retaining these “cognitive maps,” they found evidence to support their theory.

Lead author, Dr Matthew Nour (UCL Queen Square Institute of Neurology and University of Oxford) said“Until very recently, the automatic analysis of language has been out of reach of doctors and scientists. However, this situation is changing with the advent of artificial intelligence (AI) language models such as ChatGPT.”

“This work shows the potential of applying AI language models to psychiatry – a medical field intimately related to language and meaning.”

The team from UCL and Oxford now plan to use this technology in a larger sample of patients across more diverse speech settings to test whether it might prove useful in the clinic.

Journal Reference:

  1. Matthew M. Nour, Daniel C. McNamee, Yunzhe Liu et al. Trajectories through semantic spaces in schizophrenia and the relationship to ripple bursts. PNAS. DOI: 10.1073/pnas.2305290120