Natural language analysis AI "Continue to lead AI medical devices" FRONTEO/President Masahiro Morimoto/Hiroshi Toyoshiba Life Science AI CTO | Venture visit

A venture that enhances its presence as a player in the pharmaceutical industry.We will visit the manager of the hot venture, talk about the start of the business, the desire for the business, and the future prospects.

Started clinical trials for the world's first language AI medical device

――In March this year, we announced that it will start clinical trials for the “conversational dementia diagnostic support AI system”.It is the world's first clinical trial for language AI medical devices (edition: Interviews were held on April 20, announced on the 26th of the same month).

Morimoto: Dementia diagnosis is usually made by a specialist, based on the results of neuropsychological tests (MMSE, etc.).Our systems are developing cognitive dysfunction from 5 to 10 minutes of daily conversations between doctors and patients, and is developing as replacement of conventional tests.

Toyoshiba: We are using it at medical institutions such as "it takes a long time to test" or "there is no specialist and cannot be diagnosed", and we are thinking of using it as information for doctors to make final judgments.increase.

Morimoto: Since only conversation data is used, we expect to use it in remote medical treatment.

Toyoshiba: In the clinical trial, the patient will be asked for neuropsychology and consultation.We read the conversation of the interview into the system as text data to show whether the cognitive decline is suspected.The results are compared to the neural psychological tests to see if the prediction accuracy is properly prerequisite.

―― In terms of sales, we partnered with the Republican Industries last year.Are you thinking about developing it in health?

Morimoto: If it is approved, we will sell it to a medical institution, and we will have the Republican Industries sell in Japan.Neural psychological tests are also covered by insurance, so we will basically use the AI system within the scope of insurance.

At the moment, we are developing it for domestic, but we also want to expand in Asia, the United States and Europe.In the future, we are considering using it outside insurance, and I would like to expand it to the use of apps such as apps that make ordinary people check their condition.

Because you know the site, you can get the information that the site needs

――The papers search AI “AmanoGawa” has commercialized two AIs that support drug discovery AI and drug discovery research.There are many approaches to say AI drug discovery in a bite, but what do you think of your superiority in that?

Toyoshiba: In our case, there are researchers with not only engineers but also specialized biology knowledge in the company, so the strength is to select patients who use target search or use biomarkers.I think there is.

I think there are many places where artificial intelligence is used to optimize compounds, but I thought it would be the first trend.Because the optimization of compounds does not require so much biology knowledge.On the other hand, the area we are dealing with requires domain knowledge (knowledge of industry and business), and we believe that we can provide information that needs research sites based on it.。

Morimoto: In the field of drug discovery, we first analyzed medical paper.The exploration of the dissertation is an indispensable task to take a closer look at the disease mechanism in drug discovery research, and is supported by AI.In the past, drug discovery researchers were doing it by hand, but it was possible to automate it because of the understanding of the work.

――The pharmaceutical company has already introduced FRONTEO AI by Takeda Pharmaceutical and Chugoku Pharmaceuticals.How is the reaction from the introduction destination, including academia?

Toyoshiba: Pharmaceutical companies provide a passway map -like relationship between molecules and genes related to diseases through Cascade Eye, but there are not many systems that output in such a way.

Our natural language analysis AI is good at finding relationships from sentences, but from a pharmaceutical company, "(the result that came out) is a distance from the image of humans, and for researchers.It will be a new perspective. "Our AI thinks that we have a particular strength in terms of discovery.

――What is the response so far in the life science business?

Morimoto: We can also use our proposal to new applications and functions, and I feel the response.

There is a community gap between software developers and site (doctors, nurses, researchers, etc.), which requires delicate tuning to fill it.Our AI is a "micro AI" that can produce results with good accuracy even with small computer power.The advantage is that it is easy to combine with existing workflows because there is no need to prepare a large system.I think this is a rare feature of other companies' systems, and I think that the range of utilization will be expanded in the future.

Toyoshiba: At first, you were aware of technical aspects, such as improving accuracy.However, while talking to customers, it has become clear that how to work with the system is the key.In that sense, I feel that good things are being made because I have researchers with domain knowledge.

"Fairness" for life science

――Please tell us what you entered healthcare.

Morimoto: FRONTEO originally started a business in the field of Legaltech.We introduced the technology to collect and present "evidence" to determine the whereabouts of trials and surveys, and launch a company to realize fairness under the law.

After that, the sources of lawyers and criminal investigators show that there is a lot of language information such as sentences and conversations, and while the analysis is more efficient, the AI engine "KIBIT" specialized in natural language processing.Was born.

It was around 2013 that I started approaching non -regaltech fields.I think that there are other fields that require an expert's judgment like Legaltech, and while introducing our skills, I wrote in the life science field to the judgment of doctors and nurses.I knew that I was using conversation and had the need to make it more efficient.

Both Life Science and Ligartec will change their judgment depending on their expertise and experience.But both should be equally accepted.We decided to enter, thinking that our philosophy of "realizing fairness in the information society" was also possible.

――What was the reason why Toyoshiba -san participated in Fronteo?

Toyoshiba: I was originally a pharmaceutical company and participated in clinical trials to verify the effectiveness and safety of molecular targeted drugs using target search and genome biomarker.At that time, the most time was to read the dissertation and summarize with the measurement results.

I found Froneto when I was thinking that I wanted to learn NLP (natural language processing) well in order to improve this work, and thought that there was no company doing interesting NLP.。At that time, there were still many companies saying, "Let's do something using artificial intelligence," but Fronteo was already using AI at Ligartec.

I felt that there was a technique that could understand that "Fronteo transitions from A to B and B to C, but C to D has a contradiction".It is similar to a cascade that indicates the flow of molecular reactions.I thought that if the contradiction was understood on the way, I could lead to new discoveries.

―― After that, Mr. Toyoshiba has developed an AI system “Concept Encoder” used in the life science field.

Toyoshiba: There was already Kibit in Fronteo.Kibit came out of the needs of the lawsuit, "How to find evidence quickly", and is suitable for learning and analyzing on -site knowledge and tacit knowledge.On the other hand, in the life science area, it is necessary to objectively judge data instead of referring to the opinions of specific doctors.It was necessary to have an engine with a different perspective from Kibit, which led to the development of Concept Encoder.

Morimoto: At that time, we had a sense of crisis that we might not understand the needs of the field.So I wanted people familiar with the life science industry to participate.He felt the limit of forcing Kibit to the life science area, and he participated when he was trying to create a new AI engine.

Now, in the field of Legaltech and the patent analysis field, we are also promoting new products that combine the characteristics of Kibit and Concept Encoder.

You can become an AI drug discovery company

――What is the future prospects of Life Science AI?

Toyoshiba: The diagnostic support AI has led to the improvement of functions, such as further detailed predictions of cognitive function, early detection of recognition function, dialect, etc.I think.In addition, I would like to develop things that use automatic voice recognition.

Morimoto: I want to spread it to various central nervous systems with dementia.I would like to apply "COROBAN", which analyzes the nursing record of electronic medical records and predicts the risk of falling and falling.The drug discovery support AI is not only used by companies, but also believes that we will become an AI drug discovery company that targets drug discovery.

In the case of revenue, we would like to realize 10 billion yen in sales in the next five years, and at least 100 billion yen after entering the global expansion.

――What other data collected in each product do you consider?

Morimoto: Our collection data is a natural sentence, which is the subjective information of doctors and patients, but natural languages are rarely used by other companies, including overseas.It is not used even though there is a lot of information.

Concept Encoder can follow the progress and changes of subjective information, so we believe that it can be used to understand Patent Journey and maintain quality of life.I want to make use of natural languages as a new digital biomarker.

There are various regulations and rules in the pharmaceutical industry.For example, guidelines for sales information provision activities.You have to check the sentences (subjective information) written by MRs, such as daily reports and emails, and check if they violate the guidelines.We believe that this can be more efficient by taking advantage of our natural language processing (edition: On May 6, the MR business support AI system "GUIDELINE VIEWER" is started).I would like to contribute to solving problems with work efficiency.

――What kind of position do you aim for in the future?

Toyoshiba: I think that AI medical equipment is running the top as a leading company.I want to be so in the future, and I always say in the company, but I want to be a company that pulls the first -in class and pulls the industry.

(Listle, Mayu Kameda, photo is provided by Fronteo)