AI

AI's capability to spot schizophrenia through latent language cues reveals new diagnostic options

Researchers at the UCL Institute for Neurology have created new tools that can identify minor speech patterns in people with schizophrenia using AI language models. The study, which was written up in PNAS, tries to comprehend how psychiatric illnesses may be identified and evaluated by medical professionals and scientists using language analysis that has been automated.

Currently, procedures like blood tests and brain scans play a very minor role in psychiatric diagnosis, with practically all of the time spent conversing with patients and those closest to them. This lack of specificity, however, limits a deeper comprehension of the origins of mental disease and the oversight of treatment.

In two verbal fluency tasks, 26 participants with schizophrenia and 26 control participants had to list as many words as they could that either fit into the category of "animals" or began with the letter "p" in the allotted five minutes. The scientists utilized an artificial intelligence language model that has been trained on massive volumes of internet material to represent word meaning in a manner comparable to that of humans to analyze the responses provided by participants.

They investigated whether the AI model could predict the words that subjects spontaneously remembered and whether this predictability was affected by patients with schizophrenia. They discovered that patients with more severe symptoms showed the biggest difference in the AI model's ability to anticipate answers from control participants vs those supplied by those with schizophrenia.

According to the researchers, this discrepancy may be related to the way that the brain develops associations between memories and concepts and stores this information in so-called "cognitive maps." A second section of the same study, in which the scientists used brain scans to detect brain activity in the areas of the brain involved in learning and retaining these "cognitive maps," provides support for this notion.

Dr. Matthew Nour, the lead author, from the University of Oxford and the Queen Square Institute of Neurology stated, "Until very recently, clinicians and scientists were unable to perform automatic language analysis. This situation is altering, though, with the introduction of artificial intelligence (AI) language models like ChatGPT.

This research demonstrates the promise of using AI language models in psychiatry, a discipline that has close ties to language and meaning. Schizophrenia is a debilitating and widespread psychiatric condition that affects over 685,000 people in the UK and around 24 million people globally.

According to the NHS, the disease can include hallucinations, delusions, muddled thinking, and behavioral changes. To determine whether this technology might be helpful in the clinic, the UCL and Oxford team now want to test it on a wider sample of patients in a variety of speaking settings.

According to Dr. Nour, the field of neuroscience and mental health research is about to enter a very exciting period. We are starting to understand how meaning is created in the brain and how this might go wrong in psychiatric diseases by fusing cutting-edge AI language models with brain-scanning technology. The use of AI language models in medicine is incredibly popular. Within the next ten years, I anticipate these tools will start to be used in clinics if they prove reliable and safe.