AI language models support schizophrenia diagnosis: What does this mean for the future of psychiatry?

Photo courtesy of Unsplash

AI language technologies could support the future diagnosis and monitoring of schizophrenia. Researchers at the UCL Institute of Neurology demonstrated an automated diagnostic tool that allows the merging of neurological, behavioural, and phonological accounts of schizophrenia. 

Schizophrenia is characterised by hallucinations, disruptions in thought and emotional processes, as well as disorganisation of conceptual memory. This research examined 26 patients and healthy controls on a word association task, focusing on differences in cognitive maps formed by those two groups. Participants were asked to name as many animals or words starting with the letter “p” as possible in the span of 5 minutes. To validate the results, additional brain scans of regions associated with the formation of cognitive maps were performed. Indeed, as expected, the selection of words guided by semantic similarity was reduced in schizophrenic patients compared to healthy individuals.

The current diagnosis of schizophrenia almost solely relies on the personal accounts collected from the patient and their relatives, which proves to be a challenge for medical professionals. The rapid development of AI could address an urgent need for innovative solutions that would provide structured and automated tools supporting diagnostic procedures. A study published in the PNAS journal addresses exactly those needs, identifying small differences between the similarity of words chosen by schizophrenia patients and healthy individuals.

What, however, might this mean for future clinical practices? Scientists hope that AI language models could provide support for medical professionals to conduct structured phonological analysis, allowing for higher accuracy in diagnosis. Schizophrenia affects around 24 million people worldwide, and around 685,000 in the UK. Furthermore, accurate diagnosis allows for improved predictions on the further development of given disorders as well as the implementation of personalised treatment.

The implications of the study provide an optimistic future not only for patients but also for the scientific community. Moving forward, we will be able to investigate not only behavioural and neural developments in schizophrenia but also the phonological and conceptual changes. Dr. Matthew Nour (UCL Queen Square Institute of Neurology and University of Oxford) commented on this discovery: “We are entering a very exciting time in neuroscience and mental health research. By combining state-of-the-art AI language models and brain scanning technology, we are beginning to uncover how meaning is constructed in the brain, and how this might go awry in psychiatric disorders.”

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

Although the findings are promising, there is still a long way to go before this technology can be implemented in clinical settings. AI success heavily depends on intensive training on large data sets. To ensure the precision of this novel diagnostic tool, further research should be focused on testing larger participant samples. However, with growing interest in the implementation of AI in medicine, Dr. Nour says we could look forward to the technology being implemented in clinical settings within the next decade.