Early detection of cancer with AI image diagnosis Challenge of start-up companies January 31, 17:06

Endoscopes are used for diagnosis leading to early detection of cancer.

Japanese manufacturers lead the world with over 90% market share.



AI (artificial intelligence) is responsible for the doctor's technique to find cancer from the captured images.

While major companies are competing to develop the technology, a start-up company in Tokyo has taken up the challenge with a novel idea.



Early detection of stomach cancer is a particularly difficult challenge.

I asked him about his thoughts and strategies.


(Economic Department reporter Daiko Nagoshi)

Japan leads the world in endoscopy

Japan leads the world in endoscopes, with a total share of 98% of the world's share, centered on the three major Japanese companies.



Diagnosing cancer from images taken with an endoscope is said to be extremely difficult, and it is said that even specialist doctors need 10 years of experience.



Under such circumstances, technological development is progressing with AI (artificial intelligence) for “diagnosis”.



While major companies such as Olympus, Fujifilm, and NEC are competing in the development and commercialization of related technologies, Tokyo-based startup "AI Medical Service" entered this field in 2017 and is working toward practical application. I have progressed.

Taking on the challenge of gastric cancer, which has a high degree of difficulty

President Tomohiro Tada (51), who is also a specialist in gastrointestinal endoscopy, has accumulated clinical experience in more than 20,000 cases of endoscopy so far, while discovering the potential of AI and continuing development.

President Tomohiro Tada


: "Colorectal cancer is stage 3 advanced cancer with a 5-year survival rate as high as 80% if treated appropriately. On the other hand, if gastric cancer is stage 3, the 5-year survival rate is less than 50%. No matter how hard we try, we cannot save more than half of the cases, but 98% of stomach cancers can be completely cured if detected at an early stage.”

“However, early detection of stomach cancer is difficult, and it is said that 20% to 30% of stomach cancer is missed. Because of the high value of AI for diagnosis, we are focusing our research and development on it.”

AI that surpasses human power

In order to improve the accuracy of cancer detection by AI, it is said that it is necessary to learn many cases anyway.



President Tada has been developing an AI system with support from the New Energy and Industrial Technology Development Organization (NEDO) in 2019.

We partnered with more than 100 medical institutions nationwide to receive endoscopic images, and continued steady work to have the AI ​​memorize them.



In order to increase the amount of learning, we believe that videos are more efficient than still images, so we have developed equipment that can take high-definition images and provide them to partner medical institutions.



Over the past five years, we have collected over 200,000 cases.



Then, it is said that it turned out that the accuracy can exceed human ability.



For example, there was a case like this.

The same 140 cases were diagnosed by endoscopists and AI.



Among the 140 cases, 75 gastric cancer lesions are hidden.



As a result, the detection rate of specialists was 37.2%, and AI exceeded this at 58.4%.



In addition, it takes about 4 seconds for a medical specialist to distinguish between images, while AI takes only 0.02 seconds for each image.

President Tada


“In 2019, when we succeeded in researching and developing an AI that uses video to detect stomach cancer in real time, I thought that AI surpassed humans. You can spend time facing patients.Not only that, but it also expands the capabilities of doctors to provide better medical care.It's not just that AI and doctors work together to improve medical care. I think"

To be used by many medical institutions

We are also planning novel ideas to make it easier for many medical institutions to introduce AI diagnostic imaging.



One of them is that it can be retrofitted to existing endoscopes.



Endoscope equipment was packaged by each major manufacturer, including monitors and computer terminals, and the AI ​​function could not be introduced unless it was developed by the manufacturer.



President Tada made it possible to install it on endoscopes of any manufacturer, and it is characterized by being able to use it simply by installing the software on a general-purpose personal computer.

Another is the subscription fee.



We will keep the initial cost low and pay a fixed monthly fee.



President Tada's AI system is currently in the process of applying for national marketing approval.



In addition, from this year, in collaboration with the University of Tokyo, we have opened a course to consider clinical and technical issues in disseminating AI diagnostic imaging in society, and are accelerating research and development.

President Tada


“I am an endoscope user, so I was particular about the ease of introduction. AI changes in a year or two, so I decided to make it an external AI. so we can quickly provide the latest

“In addition, by keeping costs down with general-purpose computers, it will be easier for many people to use them. We can fill the technical gap between endoscopy specialists and non-specialists, eliminate regional disparities, etc., and can be used anywhere in the country. I think it will lead to equal access to medical care.”

I want to lead to early detection of cancer

Endoscopic diagnosis by AI is positioned as an aid to doctors.



After the AI ​​has determined the possibility of cancer, a doctor will make a final diagnosis after a pathological examination.



With AI support, it is possible to quickly enter the next diagnostic step that leads to early detection.



President Tada expects this to be effective in regions where there are only a limited number of endoscopy specialists.



Speaking of AI, there is certainly a negative image of taking away human jobs.



Under such circumstances, President Tada emphasized the cooperation between humans and AI.



Through the interview, she felt that her words were persuasive, as she is an endoscopist herself.



It also conveys his passion for helping as many people as possible to detect cancer at an early stage, both in Japan and around the world, which are facing the major challenge of resolving the medical underpopulation in areas.

Economic Department Reporter


Daiko Nagoshi


Joined in 2017


After working at Fukuoka Broadcasting Station