Competing with 2,200 AI experts from all over the world, Nguyen Hong Dang, an expert at Viettel Cyberspace, a member of Viettel Group, won first prize in a contest on detecting breast cancer through mammography screening.
The contest was launched among the world’s largest community of AI experts and data science and organized by the Radiological Society of North America (RNSA) on the Kaggle platform.
Kaggle serves as a place for experts to share knowledge, and is a platform that connects businesses, organizations and scientists aiming to seek and develop solutions utilizing AI to solve questions in real life.
The winning AI solution developed by Dang can recognize the signs of breast cancer from a very small area in a mammogram. The solution can be integrated into a computer-aided diagnosis, CADx systems, in order to improve accuracy in diagnosis and treatment.
According to the jury, Dang’s solution can solve the question of data imbalance in AI model training. In the data file provided at the contest, the labeled proportion of malignant cancer samples just accounted for 2 percent of total samples, while the remaining number of samples labeled as noncancerous or normal accounted for 98 percent.
The data imbalance may cause AI to give a wrong diagnosis.
Nguyen Manh Quy, director of Viettel Cyberspace, said: “We are working on turning the solution into a service on Viettel AI Platform. Hospitals, medical organizations and individuals can access and use the service in their work."
“The solution will optimize the time and costs compared with the self-studying AI model,” he said.
In order to win the contest, Dang worked hard with support from Viettel Cyberspace and used the knowledge about AI discovered by the center.
He also used the super-computer infrastructure belonging to the 500 most powerful computers in the world.
The super-computer helped find solutions, optimize algorithms with the calculation performance of 20 Petaflops, or 20 million of billions of calculations per second.
According to the World Health Organization, breast cancer is the most common cancer in the globe. Screening breast cancer can help reduce the mortality rate by 40 percent. AI tools can help the screening process become more valuable and effective.
Hai Dang