Hello everyone, My name is Kenshin Fujiwara and I am the CEO and founder of HACARUS Inc.
Through this series of blogs, I will discuss a wide range of topics of AI, from the history of AI to practical tips for the successful application of AI projects. I hope my blog posts will help you better understand AI and solve your business issues.
In the last blog, we studied a case study of AI implementation for diagnosis support at Kobe University. Today, we will take a look at another case of AI implementation for AI diagnosis support at Kyoto University.
AI for Uterine Cancer Diagnosis Studied by Kyoto University
Cervical cancer is the most common gynecological malignancy, affecting more than 10,000 people each year in Japan alone, with approximately 3,000 people losing their lives. At the Graduate School of Medicine at Kyoto University, the Department of Gynecology and Obstetrics is conducting research on AI support systems for the prevention and early diagnosis of cervical cancer.
Challenges Before AI Introduction
Prior to implementation, there were two major challenges that emerged in the field.
- There was an overwhelming shortage of medical specialists
- AI models need to digitize the know-how and industry knowledge of medical professionals
In Japan, the incidence of cervical cancer has been rapidly increasing among women in their late 20s and 30s. However, it is easy to see that the cervical cancer screening (primary screening) uptake rate in Japan is lower than those of Western countries.
At the same time, the number of facilities in Japan that can perform a thorough examination (secondary checkup), for cases where an abnormality was previously detected, is also decreasing due to a shortage of gynecologists.
To work toward solving this problem, Kyoto University, together with HACARUS, decided to introduce AI to assist in the diagnosis of cervical cancer. To help compensate for the shortage of physicians, the AI project aimed to provide highly accurate diagnostic assistance in facilities that do not have a specialist on staff.
The AI needed to learn the characteristics of the affected area from the video data taken by the endoscope from the entrance to the exit, which required a different technology than still images. The difficulty of this research was the fact that the affected area had to be analyzed using moving video data.
When building an AI that identifies areas of risk using videos, it is important for the AI to learn how expert physicians make decisions regarding affected areas. For the specialists, finding affected areas was done manually by observing the endoscopic images. As a first step, the specialists were interviewed to better understand their thought processes.
AI Programs Incorporating Medical Know-how
Using this data, our data scientists, who routinely use sparse modeling to perform their work, programmed the AI while incorporating medical know-how.
This is a case in which we successfully utilized our knowledge and experience, which are not limited to one method alone. The AI we constructed this way was also highly evaluated by the previously interviewed physicians.
We mentioned to these doctors that a virtual doctor was being created using their knowledge and expertise. In the end, even if they leave the medical field or retire, their knowledge will live on through these AI programs. I hope that everyone involved in this project can be happy knowing that there are lives being saved by our AI project.
Furthermore, the results of this study suggest the possibility of applying this technology to train physicians. As mentioned earlier, there is a major shortage of skilled physicians who can diagnose cervical cancer. By utilizing AI, we are building the groundwork for using AI diagnostic results to train new physicians.
In the next blog, I will introduce another case of AI implementation at Osaka Gas, a Japanese utility company. For your updated knowledge and insight about AI technology, subscribe to our newsletter or visit HACARUS website https://hacarus.com.
