“At the age of 31, a turning point in my scientific research process came to me. That was when I attended the American Gastroenterology Week and the European Gastroenterology Week. The most interesting scientific session was the application of AI (artificial intelligence) in gastrointestinal endoscopy. At that time, in Vietnam, this issue was very new and had never been mentioned in the field of gastrointestinal endoscopy. A series of questions arose, I wrote a letter to my teacher in Japan, worried that “Will AI replace endoscopists?”. His answer made me feel very interesting and excited: AI cannot replace but is a support tool. With AI, the time, effort, and resources for diagnosing each normal case will be shortened; endoscopists will focus on diagnosing and intervening in difficult cases. Thus, AI will help redistribute the resource structure. Looking at the reality in Vietnam, I think that AI can be applied in two roles: Supporting training to standardize doctors' capacity and post-checking to help the endoscopy process ensure time, image quality, and minimize missed lesions, which will be very valuable. That was the motivation for me to start implementing the idea of ​​researching the application of AI in digestive endoscopy, although at first, my colleagues and I felt it was difficult and concerned about the possibility of application." That was the opening story of Associate Professor, Dr. Dao Viet Hang, Deputy Director of the Endoscopy Center of Hanoi Medical University Hospital. Then came the arduous journey of applying AI to digestive endoscopy techniques in Vietnam by her and her colleagues.

180 DAYS AND NIGHTS OF "TRAVELING AND PROBING THE WAY"

Vietnam is a country with a high population of digestive diseases, but the majority of patients, especially in rural areas, do not have access to specialized technologies for examining digestive disorders. The ability to diagnose and detect early digestive tract lesions of medical facilities is also limited. According to medical literature, in developed countries, the rate of gastrointestinal cancer lesions (stomach, esophagus) missed on endoscopy is 11%, while the figure for colon polyps is 26%. In our country, although there are no official statistics, the reality shows that the unevenness of equipment and experience of doctors and nurses is potentially causing the risk of missing lesions when diagnosing patients, especially in local hospitals. To overcome this problem, most doctors agree that technological assistance is needed. However, they face difficulties due to lack of tools. At the time Dr. Hang and his colleagues started to research AI in gastrointestinal endoscopy, many famous equipment manufacturers had integrated AI software into endoscopes, but the cost was very expensive and only compatible with the company's modern equipment. Meanwhile, the resources for medical equipment in Vietnam were limited, especially provincial and district hospitals could not afford such expensive systems. The problem of economic cost-effectiveness that had to be solved was to develop a Vietnamese-specific AI algorithm, develop a system that could integrate many types of endoscopes, especially in localities, thereby helping doctors at the grassroots level increase the ability to detect lesions. The research team expected that based on a large image data set recording specific lesions of Vietnamese people, along with assessments by Vietnamese endoscopy experts, a "Make in Vietnam" AI algorithm with accuracy equivalent to reports in the world would be born. According to Dr. Hang, after a small study in 2019 showed positive results, from 2020, more than 20 experienced gastrointestinal endoscopists in Hanoi, Ho Chi Minh City and Hue were "gathered online" and worked with colleagues from Hanoi University of Science and Technology to conduct research and develop the product.

Associate Professor, Dr. Dao Viet Hang. Photo: Hoang Ha

The biggest difficulty in the process of creating this algorithm was building an image dataset. Associate Professor Hang said: “With AI, the most important thing is that the image dataset is large enough, the number can be up to millions of images, and it must also be diverse in morphology and accurately labeled and localized. However, the process of collecting images is not simple because the endoscopy system in Vietnam is not uniform. In the world, to have reports published in 2018, international research groups had to start 5-7 years ago”. Recalling the journey of the first days, she shared: “I always remember the first 6 months of “walking and exploring the way”, we had to find a common voice among specialists, agree on the localization of lesions on endoscopic images, and call the lesions by the correct name; and had to find a way to interpret the language with the group of information technology engineers. That period of time taught us unprecedented perseverance and patience. If it were not for the bigger picture, for the common good of the Vietnamese digestive endoscopy industry and for the patients, doctors with a busy work schedule, patients waiting in long lines, could not spend time day and night, taking the effort to draw each image in detail for discussion". With the "dizzying" work schedule of a doctor, lecturer, scientific instructor, manager, Dr. Hang never misses any calls from colleagues, even at 3-4am. At that time, she immediately turns on the lights, opens the zoom to analyze each endoscopic image collected with colleagues. It could be a conversation with a doctor who has just completed an emergency endoscopic intervention and the discussion must end before 6am to continue the new day with an early medical examination schedule. Inspired by her colleagues, Dr. Hang and her colleagues were also empowered by the curiosity, excitement, and expectations of patients for a tool that could help increase the ability to detect gastrointestinal lesions.

INITIAL ACHIEVEMENTS

Up to now, the initial effectiveness of the artificial intelligence software implemented by Dr. Hang and his colleagues shows that the rate of detecting colon polyps and classifying benign and malignant lesions in the lower digestive tract is 98 - 99%. With the upper digestive tract including esophageal cancer, stomach cancer, the algorithms for detecting lesions have an accuracy of 80 - 85%. With lesions in the lower digestive tract, especially colon polyps, the doctors have completed the development of the polyp detection algorithm and in step 2, classifying the lesions as benign or malignant so that doctors can intervene immediately during the endoscopy process. "We hope that in the future, the product will not only be a machine used at medical facilities but also a rich image database source for training and building an e-learning system so that lower-level doctors can improve their skills and knowledge no matter where they work," Dr. Hang expressed. In addition, the research team also launched two smartphone applications that directly serve patients, including: a specialized application to support patients preparing for colonoscopy and an application to help patients manage gastroesophageal reflux. It can be affirmed that more and more doctors are recognizing the superiority of AI. However, according to Associate Professor Hang, to take advantage of the strengths of this technology in the medical field in general, Vietnam needs to continue to build core technology solutions at reasonable costs. In addition, we also need to solve another important problem for AI to truly become a powerful "assistant" for doctors. That is the interaction between doctors (humans) and AI (machine systems). "Although the data published so far has shown quite positive results, will AI and doctors agree and harmonize in their assessments between an expert who always raises questions and doubts, and a machine system that is regularly upgraded and trained? That is a common problem that many large research groups in the world are also facing," Associate Professor Hang commented.
Associate Professor, Dr. Dao Viet Hang (born in 1987) graduated with honors from Hanoi Medical University, successfully defended her PhD thesis at the age of 29, and was recognized as Associate Professor 6 years later, becoming one of the youngest female associate professors in our country. Dr. Hang is currently the Deputy Director of the Endoscopy Center of Hanoi Medical University Hospital, Deputy Secretary General of the Vietnam Gastroenterology Association, and a pioneer in Vietnam in applying AI in gastrointestinal endoscopy. At the age of 34, she received the Golden Globe Award in the field of medical technology and the Promising Young Face of Vietnam. Currently, the female doctor is the author of more than 20 international articles and 60 domestic articles and is also the President of the Global Network of Young Vietnamese Intellectuals.

Vietnamnet.vn

Source