Visitor publication

Prof. Liang Jianhong: Application of Artificial Intelligence in ROP Prevention and Control | WOC 2022

Posting time:2023-01-31 14:29:43

Prof. Liang Jianhong: Application of Artificial Intelligence in ROP Prevention and Control | WOC 2022

Editor's Note Artificial intelligence (AI) has been widely used in many fields, providing convenience for all aspects of people's lives. AI has also been well applied in the field of ophthalmology. In 2021, Professor Liang Jianhong of Peking University People's Hospital won the invention patent (ZL201810354164.2) of China's first ophthalmic retinopathy of prematurity (ROP) artificial intelligence diagnosis and treatment assistance system, and It was presented at the 38th World Ophthalmology Congress (WOC 2022). "International Ophthalmology News" specially invited Professor Liang Jianhong to conduct an exclusive interview on this content. Professor Liang introduced in detail the current application of AI in the prevention and treatment of ROP and the problems to be solved. I believe it will bring some enlightenment to the majority of doctors. AI for ROP prevention and control has a good development soil in my country. As we all know, in the world, ROP is the first blinding eye disease in infants and young children. The prevention and control of ROP is of great significance. There are more than one million premature infants in my country every year. According to the screening guidelines for premature infants in my country, most premature infants need to be screened. On average, the consistency of disease classification among doctors is poor, the screening standards and modes are relatively single, and there are many challenges in the screening process. Therefore, the existing screening modes are far from meeting the needs of preterm infant screening. The Department of Ophthalmology of Peking University People's Hospital has been engaged in the screening and diagnosis and treatment of ROP since the late 1990s. It may be the earliest medical institution in China to carry out this work, and it has accumulated very rich experience in this field. Because of this, we want to use AI to widely promote this experience, hoping to share these experiences with the majority of ophthalmology colleagues, and then benefit patients. The working principle of the AI ​​system and the performance of the image reading accuracy This system was jointly developed by us and the artificial intelligence team of the national high-tech enterprise. Combined with the large amount of data accumulated by our ROP prevention and control work for more than 20 years, based on the deep convolutional neural network technology, Let the AI ​​algorithm model repeatedly learn the lesion characteristics of the labeled data, so as to have the ability to identify lesions. This AI system can not only characterize ROP, but also quantitatively stage ROP, and can provide preliminary diagnosis and treatment advice, which is very helpful to many doctors. At present, our experimental evaluation results show that compared with general experts in the ROP field, the accuracy of AI image reading is significantly higher, and the current accuracy rate is about 95%. my country has unique conditions to apply AI to ROP prevention and control. The use of AI to control ROP has some incomparable advantages in my country. First of all, in terms of hardware, my country has established a very complete communication network system, which covers every country in my country. On the other hand, it is also a more important point. After years of development, my country has established a very complete maternal and child health care system, which covers the health care of every newborn. Although the medical staff in the maternal and child health care system are not specialists in the prevention and treatment of ROP, they can use the AI ​​system to conduct preliminary screening and diagnosis of ROP, creating conditions for timely treatment of ROP. What are the current problems in the clinical application of AI in the prevention and treatment of ROP? We and the Bethune Public Welfare Foundation launched a three-year large-scale public welfare project on the prevention and treatment of child blindness in Tibet. In this project, we integrated the AI ​​system to screen and diagnose blindness in infants and young children in Tibet. During the application process, we found some problems: ① At this stage, this AI system cannot work autonomously, and we still need to manually monitor its image reading and diagnosis in the background. In the future, through continuous improvement and more extensive and in-depth learning, this system may work autonomously and conduct self-supervision without requiring expert intervention in most cases, or the opportunity for expert intervention will gradually decrease, but this is not yet possible. ;2 AI is a new technology product for clinical use, how to integrate it with some processes in our existing work, we do not have much experience at present and are still exploring; 3 To develop medical AI and apply AI to medical treatment, it is necessary to It is led by senior doctors in related fields, supported by medical institutions, assisted by medical staff, and participated by professional and technical personnel such as algorithms, software, and engineering, and has various resources such as hardware equipment, computing power, storage, and clinical data. Funding is also required. support, and involves the issue of marketization; ④ The more critical and complex issue is that the clinical application of AI still lacks corresponding mature regulations and policy support, and the environment is still developing, which requires a lot of work. How can the diagnosis and treatment opinions of the AI ​​system be recognized from the perspectives of medical profession, law, and social ethics? If there is a medical dispute in the application process, how to judge? These issues need to be defined by corresponding regulations. What are the application prospects of AI systems in the medical field? Technically, this system has opened up the path of ROP diagnosis and treatment, so from the principle point of view, it can also be applied to the diagnosis and treatment of other eye diseases. There is no technical problem. I personally think that the application of AI system in the medical field The prospects are very broad. Expert Profile Professor Liang Jianhong, MD, Chief Physician, Eye Center, People's Hospital of Beijing Medical University, Deputy Director, Eye Center Member of the Standing Committee of the Chinese Journal of Ophthalmology, Editorial Board of the Chinese Journal of Ophthalmology, Chinese Medical Association and Ophthalmic Pathology Group of the Chinese Medical Association. Member of Beijing Medical Malpractice Appraisal Committee. Referral specialist from the Hong Kong Hospital Authority (responsible for the diagnosis and treatment of intraocular tumor patients in Hong Kong) graduated from Beijing Medical University in 1990. From 1990.9 to 1994.9, he worked in the Ophthalmology Department of Beijing Friendship Hospital as a resident physician. From 1994.10 to 1997.10, he worked in the People's Hospital Affiliated to Beijing Medical University as an attending physician, specializing in fundus diseases. In 1997, he was admitted as an on-the-job doctoral student by Professor Li Xiaoxin. In the same year, he was admitted to the German Academic Exchange Center Scholarship (DAAD). From October 1997 to February 1998, he studied German at the Gothic Language Institute in Göttinging, Germany. From 1998.3 to 2000.12, he studied and worked in the Ophthalmology Center of the University of Essen, Germany, specializing in fundus diseases and intraocular tumors. During this period, he studied under Professor Norbert Bornfeld, director of the Eye Center and an expert in fundus diseases and eye tumor. From 2001.1 to 2002.9, he worked as a postdoctoral fellow at the Ophthalmology Center of Columbia University in the United States. Since 2002.10, he has been the deputy chief physician of the Department of Ophthalmology, Peking University People's Hospital; in 2008, he has been the chief physician, specializing in fundus diseases, intraocular tumors, pediatric eye diseases and retinopathy of prematurity. He has been working in the clinical front line for many years and has accumulated rich clinical experience. He is good at various vitreoretinal surgeries, the prevention and treatment of pediatric fundus diseases, especially retinopathy of prematurity, as well as retinoblastoma, choroidal melanoma, choroidal metastases, retinal and Modern eye-saving treatment for various intraocular benign and malignant tumors such as choroidal hemangioma. Was rated as the national outstanding ophthalmologist.

Top ranking