In Japan, where the population is aging further, the number of people with dementia continues to increase, and countermeasures against dementia are one of the important social issues in Japan.Diagnosis of dementia is usually based on a medical history, imaging tests, and multiple cognitive tests to measure memory and arithmetic skills.However, these tests are highly specialized, and there are problems such as the need for training for medical personnel who perform the tests and the time required.

　Project Professor Taishiro Kishimoto of Keio University School of Medicine Hills Future Preventive Medicine Wellness Joint Research Laboratory and FRONTEO Co., Ltd. have developed a "conversational dementia diagnosis support AI program" using natural language processing (NLP).This uses AI to detect the possibility of dementia based on free conversations between the elderly and medical staff.

　In this research, we succeeded in determining the prevalence of dementia with an accuracy of 0.90 (Note 1) from free conversations that do not use specific tasks.This program is easy to implement and can reduce the burden on both medical professionals and patients.In addition, as a technology that can avoid the "learning effect" in which the test subject remembers the contents of the test and the accuracy of the test decreases by repeating it, which was an issue with conventional cognitive function tests, it can be applied to screening tests. is expected to

　The results of this research will be announced on August 2022, 8 (UK time).Scientific Reportswas published in the magazine.

1.Research background and outline

　In Japan, where the population is aging, the number of people with dementia continues to increase, and is estimated to reach 2025 million (about 730 in 5 elderly people) by 1.^*.Measures against dementia are one of the important social issues in Japan.Diagnosis of dementia is usually based on a medical history, imaging tests, and multiple cognitive tests to measure memory and arithmetic skills.However, these tests are highly specialized, and there are problems such as the need for training for medical personnel who perform the tests and the time required.

　Dementia affects various functions such as memory and attention, but it is also known to affect language functions.Therefore, we recorded free conversations and conducted research to determine the possibility of dementia using natural language processing (NLP) technology. A total of 135 conversations were transcribed from 432 collaborators, decomposed into morphemes and parts of speech, vectorized, and machine-learned.As a result, we succeeded in determining the prevalence of dementia with an accuracy of 0.90, a sensitivity of 0.88, and a specificity of 0.92.This accuracy could be achieved with the number of words obtained from 3-5 minutes of speech.

* Ministry of Health, Labor and Welfare: Comprehensive Promotion of Dementia Measures, https://www.mhlw.go.jp/content/12300000/000519620.pdf

2.Research results, significance, and future development

　The results of this research are based on the detection of dementia based on free conversations between the elderly and medical personnel, and it is possible to identify dementia without conducting tests such as memory and calculation.As a technology that can avoid the "learning effect" in which the subject remembers the contents of the test by repeating the test and the accuracy of the test decreases, it is expected to be put into practical use for screening tests.

　Based on the results of this research, from April 2021 to March 4, FRONTEO will run AI for the "Conversational Dementia Diagnosis Support AI Program" with Professor Masaru Mimura of the Department of Psychiatry and Neuroscience, Keio University School of Medicine as a clinical trial coordinator. A clinical trial was conducted for practical application as a medical device.Currently, based on the results of the clinical trial, we are proceeding with the process of obtaining pharmaceutical approval.

3.Remarks

　This research was conducted by the Japan Agency for Medical Research and Development (JAMED), a research and development project for medical devices and systems that realize future medical care. This was done with the support of "Development of a device that delivers real-time evaluations".

4.paper

English title:Identifying neurocognitive disorder using vector representation of free

conversation

Japanese translation of the title:Evaluation of dementia by vector representation of free conversation

Author name:Toshiro Horigome, Kimihiro Hino, Hiroyoshi Toyoshiba, Tohisa Shindo, Katsura Funaki, Yoko Eguchi, Momoko Kitazawa, Takusen Fujita, Masaru Mimura, Taishiro Kishimoto

Journal title:Scientific Reports

TWO:10.1038 / s41598-022-16204-4

URL:https://www.nature.com/articles/s41598-022-16204-4

[Glossary]

(Note 1) Accuracy refers to the percentage of correctly recognizing the presence or absence of a disease.Here, it refers to the percentage of subjects who were able to correctly recognize dementia patients or healthy elderly people among all subjects.A perfectly correct answer is 1.00.In addition, sensitivity refers to the percentage of people with dementia who were correctly recognized as having dementia, and specificity refers to the percentage of healthy elderly people who were correctly recognized as healthy elderly people.

*Please contact the following in advance when interviewing.

*This release has been sent to the MEXT Press Club, the Science Press Club, the Health and Labor Press Club, the Kosei Hibiya Club, and the science departments of each company.