Automatic analysis with 93% accuracy whether to become schizophrenia in the future
by on
Schizophrenia is a disease for which treatment has not yet been established, and investigation of the cause and development of preventive measures are urgently needed. Under these circumstances, researchers at Harvard University and Emory University automatically analyze the patient's statement before schizophrenia and other mental disorders, and predict with a 93% accuracy the risk of future illness. I developed a machine learning tool.
A machine learning approach to predicting psychosis using semantic density and latent content analysis | npj Schizophrenia
(PDF file) https://www.nature.com/articles/s41537-019-0077-9.epdf
eScienceCommons: The whisper of schizophrenia: Machine learning finds 'sound' words predict psychosis
https://esciencecommons.blogspot.com/20019/06/the-whisper-of-schizophrenia-machine.html
The whisper of schizophrenia: Machine learning finds 'sound' words predict psychosis | EurekAlert! Science News
https://www.eurekalert.org/pub_releases/2019-06/ehs-two061319.php
Many schizophrenia and mental disorders begin to develop prodromal symptoms around the age of 17 and develop in the 20s. There is a 25-30% chance of developing schizophrenia and other mental disorders for people with prodromal symptoms, and experienced clinicians have future mental disorders with 80% accuracy through structured interviews and cognitive tests Is said to be predictable.
The method using machine learning developed this time is to find out the pattern that will become ill in the future from the conversation of people with prodromal symptoms. It has long been known that there are certain patterns in the remarks of people with mental disorders.
Surveys using machine learning reveal words that people with 'depression' tend to use-GIGAZINE
In this research, first, we let the computer learn the 30,000-user conversations exchanged on the overseas bulletin board reddit . Then, using a neural network called Word2Vec that vectorizes words, 'position' is assigned to words in a semantic space based on the meaning of words. At this time, words with similar meaning are arranged so that the distance is closer than words with distant meaning.
In addition, researchers at Emory University Wolff lab have developed a program to analyze the semantic density of used words and succeeded in quantifying how much information is embedded in each sentence.
After generating the baseline 'normal' data from the reddit user's conversation, the researchers applied the above techniques to 40 patients who actually had a clinician. Comparing the data automatically analyzed by the program with the reference normal sample and the long-term data of the patient who became mental disorder, the use of the words regarding 'sound' and 'voice' is higher than the reference It has been shown that the probability of becoming schizophrenia and mental disorder increases in the future. The developed program could predict the risk of future schizophrenia with 93% accuracy.
by on xay
The results are expected to streamline existing diagnostic methods, recognize new variables, and improve prediction accuracy.
It is one of the symptoms of schizophrenia that abnormal sounds are heard, but even people in the pre-stage who develop schizophrenia often use sound-related words, even experienced clinicians It is a point that you do not notice. “Determining such subtle differences in conversation is like looking at microscopic bacteria with the naked eye.” “The automated analysis technology we developed finds these hidden patterns very much. It is a highly sensitive tool, ”says Neguine Rezai, the author of the dissertation.
Related Posts:
