iCognito Neural Network for the Diagnosis of Mental Disorders

The main articles are:

• Neural networks (neural networks)
• Artificial Intelligence in Medicine
• Artificial intelligence in Russian medicine
• Mental disorders

2022: AI learning to recognize psychological disorders by reporting

The company -Skolkovo resident iCognito September 16, 2022 announced that on the basis of more than 50 million messages, users are trained AI to recognize shades of psychological states and human disorders in order to data build recommendations for psychotherapy on the basis of these. The first version of the product with an advanced AI component is scheduled to be launched in mid-2023.

iCognito specialists have launched a series of experiments on machine learning and improving recognition accuracy. The first version of the product will introduce mental state recognition capabilities on a selective set of steps and individualized AI recommendations for managing these states.

Until now, machine intelligence has only managed to recognize basic emotions by facial expression to voice , and text based on. big data Now experts have managed to train to algorithms social networks recognize aggressive and suicidal words and phrases.

The development of iCognito will make it possible to make much more accurate forecasts regarding psychological problems, allowing you to detect various depressive and alarming symptoms, problems of self-perception and decision-making, various crises in relationships, etc., the company emphasized. The product will be able to be used for primary diagnostics in medical institutions. By analyzing patient messages for specific markers of mental conditions and disorders, the AI will invite the patient to take a test to confirm their diagnosis and then refer them to the right specialist.

Also, this development is aimed at improving the quality of assistant robots around us.

Imagine that a chat bot assistant will be able to determine from a series of your messages what you are experiencing and why, and how long you have a state of anxiety or depression - and respond with a certain degree of emotional intelligence, recommend you steps to manage your condition, including overcoming mild depression and anxiety, - noted in iCognito.

According to the founder and director of iCognito Olga Troitskaya, while machine learning in the field of psychotherapy works with rather general needs. Progress in this area rests on the lack of high-quality data for training artificial intelligence.

"As of 2022, there is a large amount of unstructured text data - transcripts of psychotherapists or crisis line chats that contain live speech," said Olga Troitskaya. - Therefore, our research project began with the design of such chat bots that would imitate typical scenarios of psychotherapists' dialogues and automatically mark users' answers to questions that a psychotherapist usually asks. Over the 2.5 years of the iCognito project, more than 350 thousand people have downloaded, and we have accumulated a large database of text messages on depression, anxiety, stress and relationships. Since we initially collected the marked-up data, we obtained a very high-quality base for AI training. "

The ambition of the iCognito team is to offer AI recommendation algorithms to follow certain psychological self-help rules in recognizing a particular mental state, including managing those states for a long time. As of 2022, evidence-based psychotherapy has a wide range of self-help tools for mild to moderate forms of disorders. Thanks to these methods, you can independently work with the problem for a long time, filling out diaries, recording your states, performing certain exercises, and the algorithm will contribute to the processing of results and prompts for further steps.

iCognito products will be available through app stores, where some programs are already available for free.

2021: A study to establish the possibility of recognizing psychological symptoms and disorders by text

On November 9, 2021, iCognito announced two studies to establish the possibility of recognizing psychological symptoms and disorders by text. For research, open messages from psychological forums and messages left by users in iCognito computer psychotherapy programs were used.

The iCognito study was the first experiment to create a database in Russian, consisting of replicas of users on the topic of psychological states, since most scientific work is carried out in English. The study used the supervised learning method, where some of the information was marked up to teach the model, and the marked up information was used to predict new classes of data. The marking was carried out by psychologists according to the international classification of diseases DSM-5 in terms of symptoms of depressive and anxiety disorders, and the most frequent user requests and complaints were also allocated to individual classes.

The study found that the program can establish the presence of a depressive or anxiety disorder according to the user's message with a probability of up to 87%, as well as recognize individual symptoms: for example, irritability, suicidal tendency, negative attitude towards oneself, eating disorders or sleep. Artificial intelligence was able to understand whether a person is currently experiencing depression, anxiety or irritation. In addition, the model has learned to recognize the most frequent user requests: for example, problems related to work, money, study or relationships. It was also possible to determine whether a person suffers from loneliness or has a fear of communication (social anxiety).

It turned out that speaking about their psychological problems, users most often talk about their emotional states and practically do not mention the physiological symptoms of mental disorders - such as loss of concentration, sleep disorders, psychomotor disorders, muscle tension. The lack of data on these signs reduced the accuracy of their recognition and the reliability of the diagnosis. At the same time, the results indicate the prospect of continuing research on a larger amount of data and the wide possibilities for the development of artificial intelligence.

Recognition of mental states and disorders will allow machines to identify individuals at risk based on data left on social networks; identify the most severe conditions and provide targeted assistance. iCognito also plans to develop conversational artificial intelligence using these technologies - dialogue systems that will underlie smart and emotionally responsive assistant robots. The first introduction is planned to be carried out in its own computer programs of psychotherapy iCognito, which are implemented in the format of micro-sessions with a chat bot on smartphones.

Notes