AI-based system can forecast onset of diabetes-related eye complication
Local medics teamed with experts on computer and artificial intelligence successfully developed an AI-assisted system to precisely predict the risk and progression of people with diabetic retinopathy, providing guidance and support on DR screening, prevention, and control.
Diabetic retinopathy, a major eye complication of diabetes, is also the major cause of preventable blindness. Every diabetic faces a risk of DR and there are no symptoms in the early stages. Previously, AI has been used to identify DR, while how to forecast the risk of DR based on eye-ground images remains a challenge worldwide.
Because diabetics don't usually see their doctors for months at a time, it can be difficult to know the exact time of onset or development status of their DR. The new system impacts the establishment of an accurate model to follow DR development and provide appropriate forecasts on individual's DR onset and development.
The research was published by world-leading journal Nature Medicine.
To solve the problem, Dr Jia Weiping from Shanghai 6th People's Hospital cooperated with experts from Shanghai Jiao Tong University, and Tsinghua University to create an innovative AI-based modeling approach to give precise warnings of the risk and onset time of DR.
Doctors used the system in real clinical processes in China and India, confirming its effects even while reducing wide-scale screening and public health costs. The system can help achieve individualized screening and patient management, doctors said.
"Early screening and intervention are extremely important for DR prevention and management," said Dr Jia Weiping, a leading expert in the research. "We suggest diabetics with no or light DR should receive checks each year to identify problems and conduct intervention in time. But it is difficult to achieve this due to the limitation of medical resources or economic reasons in many middle- and low-income countries.
"So we developed this AI-based system, which can precisely forecast the individualized risk and development of DR in the coming five years. It also can give precise guidance on schedule for follow-up screening based on patients' individual risks, and provide management policy."
The research was published by world-leading journal Nature Medicine.
