Artificial intelligence applied to the study of different areas of the brain will be able to predict people’s emotional disorders, including anxiety and depression, making clinical diagnoses ever more precise. The research group coordinated by Alexander Grecucci of the Department of Psychology and Cognitive Sciences and of the Interdepartmental Center of Medical Sciences of the University of Trento has used for the first time a method to build a predictive brain model, capable of correctly classifying the anxiety of the study participants, even without have information about their psychological state.
Kekko dei Modà: «My struggle with depression: I couldn’t bend my legs and get out of bed»
You coordinate the Clinical and Affective Neuroscience Lab at the University of Trento, what do you do?
«We study the emotional life of the human brain to understand and explain our ability to process emotions, regulate them and channel them into actions and decisions, both in normal and pathological conditions. We know that failures to perceive, modulate and express our emotions are at the heart of a great many psychological disorders, from anxiety to personality disorders. In the laboratory there are researchers from various disciplines, physicists, engineers, psychologists, and in coordination, in addition to myself, there are Irene Messina and Federica Meconi».
What technologies are used?
«Neuro imaging tools, such as functional and structural magnetic resonance, electroencephalography and more recently neurostimulation methods to understand where these functions are located. In the latest research we are also focusing on the study of personality – a complex construct of cognitive, behavioral, affective patterns – whose disorders are still poorly understood today, such as narcissistic, borderline and antisocial disorders, which can also develop criminal acts. Furthermore, through advanced technologies, such as artificial intelligence algorithms, we have analysed, and are now able to understand, the whole circuit that predicts people’s level of anxiety, the intensity of which is different from individual to individual in response to an experienced event” .
Let’s come to the methodology you used.
“We record the activity or structure of the brain using traditional neuroscience methods, magnetic resonance imaging, to have brain images of each person and how the brain works; subsequently the raw data was entered into a computer equipped with a combination of machine learning algorithms, which processes that data, providing predictive models capable of understanding which areas of the brain are affected by anxiety, even in individuals we do not know about Nothing”.
And how did the AI analysis process take place?
«We use methods that come from a branch of machine learning, already used in various scientific but also technological fields, such as the algorithms used by social networks or e-commerce sites to predict the behavior of its users. In our laboratory we thought of the brain as a mathematical object, whose structure can be broken down into smaller units, to understand their connection in order to predict a specific mental state. Machine learning allowed us to understand the functioning of this mathematical object, the brain, and later we used them to predict a specific mental state, for example anxiety».
Sounds like the plot of a science fiction movie.
«In class, in fact, I often show a sequence from the film Transcendence with Johnny Depp, in which some scientists try to record his brain data to transfer his mind to a computer and our method is very similar. We record brain states and by applying the predictive models filter we try to understand if we can predict emotions and personality».
Do you think this method is valid for everyone?
«Until now, neuroscientific studies have allowed us to elaborate an ideal average of the participants, but which does not take into account individual differences, while our models capture the brain characteristics of single individuals, so having learned from those data, they are able to predict the characteristics of an unexamined person. The algorithm, in fact, has learned to tolerate variations on a statistical basis».
How accurate are these predictive models?
«We are around 80/90%. Human error is much wider, because a clinical diagnosis is made with questionnaires, tests and interviews, but an objective marker is not available, while our aim is to create biomarkers that have diagnostic objectivity from the point of view of a precision medicine.
What would be the benefit?
«Refine treatment and therapeutic success. In the future, these models could be used to diagnose early development of future psychological problems or disorders, as other studies are already doing with Alzheimer’s, for example. In a new study on adolescent disorders in collaboration with the University of Bordeaux we are carrying out research on 600 young people».
You gave the example of a film, thinking of Spielberg’s Minority Report, it was possible to predict violent behavior, and therefore crimes. Possible?
“Science fiction has sometimes been shown to be pre-science. These models can lead to even very precise predictions of an individual’s future, but it is hoped that they will be used preventively and not in brain control or to pre-arrest them. Instead, I think it would be very interesting to understand whether a boy of 2023 could develop depression in the next ten years, to intervene early, the best weapon to guarantee a good quality of life for people and to reduce the costs of the health system”.
© breaking latest news