American researchers have succeeded in detecting 93% of individuals at risk of psychosis thanks to the analysis of conversations by machine learning.
Schizophrenia is a serious brain disorder resulting from an imbalance in the chemical systems of the brain. This disease generally appears around the twenties and affects 0.7% of the world’s population, including 600,000 people in France. In order to be able to intervene as early as possible, researchers are trying to better understand the pathology and its risk factors.
Knowing who will develop psychosis
A new study, published in the journal npj Schizophrenia could make it possible to move forward on the subject. Thanks to artificial intelligence analyzing language, American researchers succeeded in predicting 93% which individuals would develop psychosis during their lifetime.
When he is in a state of psychosis, the patient struggles to differentiate what is true from what is not. During this phase, he suffers from what is called a psychotic episode where he experiences disillusions and hallucinations. During a crisis, the patient will tend to act inappropriately, to speak incoherently, to suffer from insomnia, anxiety and sometimes even to fall into depression. The warning signs of psychosis generally begin in late adolescence during a prodromal phase, that is, the period during which a set of prodromal symptoms, usually mild, announce the arrival of the main phase of the disease.
Through questionnaires and cognitive ability tests, trained physicians can usually predict by 80% which patients entering the prodromal phase will develop psychosis. They are therefore now working on different approaches to improve this prediction rate. Among them is machine learning, a form of artificial intelligence by which computers “learn through experience”. This system searches for patterns in a known database and decides which match specific features.
Machine learning to spot clues hidden in language
Researchers from Emory University in Atlanta, Georgia, and Harvard University, Massachusetts decided to use this technique to spot hidden clues in language use.
They therefore programmed their machine to identify the linguistic norms of daily conversation thanks to the dialogues of 30,000 users of the Reddit forum. They then analyzed the words with software. After training the machine to establish a “normal database”, the team fed it with conversations extracted from questionnaires completed by 40 volunteers during a study on the diagnosis of young people at risk of psychosis.
The researchers then compared the analysis of these conversations with the database. Result: the participants who subsequently developed psychoses tend to use words evoking sound (they speak of voices, among other things) and have a weaker semantic field than the others: they use similar words more frequently.
“A microscope for the warning signs of psychosis”
“Trying to hear these subtleties in conversations with people is like trying to see microscopic germs in your eyes,” says Neguine Rezaii, from the Department of Neurology at Harvard University. Thus, machine learning “is like a microscope for warning signs of psychosis.”
“If we can identify which individuals are at risk earlier and practice preventive interventions we may be able to reverse the trend (…). There is interesting data showing that treatments like cognitive-behavioral therapy can delay the onset psychosis and perhaps even reduce the frequency of seizures”, adds Professor Elaine Walker who also participated in the study.
Schizophrenia is the most common psychosis
Schizophrenia is the most common chronic psychosis, psychotic crises can also occur in a context of deep depression, manic episode in a bipolar person or abandonment crisis in a borderline patient. It generally evolves by acute phase in the first years, it generally stabilizes with residual symptoms of intensity variable depending on the patient. This of course depends on the quality of the patient’s psychosocial support, their access to care and their acceptance of their treatment.
Indeed, being aware of one’s illness and benefiting from rapid care from the first psychotic disorders helps enormously to heal. Because it is possible to cure schizophrenia. Thereby, according to Insermabout a third of patients are in lasting remission after a few years of treatment and can resume a normal social, emotional and professional life.
.
