Researchers from the UCL Institute for Neurology and Oxford have developed AI models that can help mental health practitioners diagnose patients with schizophrenia.
Schizophrenia affects approximately 24 million people or 1 in 300 people worldwide and can be difficult to diagnose in its early stages. There is no blood test that can diagnose schizophrenia so psychiatrists make their diagnoses based mostly on conversations with the patient and those close to them.
In their experiment the UCL researchers asked 52 participants to complete verbal fluency tasks. Half of the participants had been previously diagnosed with schizophrenia with the other half acting as the control group.
They were given 5 minutes to come up with as many words as they could that either started with the letter “p” or belonged to the category “animals”.
The researchers then used an AI model to analyze the responses to see how predictable the words were that the participants came up with.
The model was trained on a large amount of internet text so it had a good idea of the kind of language to expect from the average person.
The control group of participants came up with words that were well aligned with what the model predicted. The participants with schizophrenia displayed reduced predictability which continued to decrease as the severity of their symptoms increased.
Our brains learn relationships between memories and ideas and store these in what the researchers call ‘cognitive maps’. These cognitive maps help us put together coherent thoughts and express them in words.
Researchers believe that schizophrenia patients struggle to do that because the areas in their brains responsible for creating and storing cognitive maps are deficient. The brain scans the researchers performed on the participants in this study confirmed the theory.
Lead author of the study Dr. Matthew Nour said “This work shows the potential of applying AI language models to psychiatry—a medical field intimately related to language and meaning.”
The researchers plan to expand the testing of their AI model to see if it could prove useful in a clinical setting for schizophrenia and other mental disorders.
AI models are really good at analyzing vast amounts of data and spotting things that humans simply can’t. We’ve previously seen how eye scans can detect symptoms of Parkinson’s, Alzheimer’s, or even schizophrenia.
Combining language models with visual AI models could lead to faster and earlier diagnoses of mental conditions. We’ve also seen promising developments in how AI is being used to create new drugs to treat them.