DAEJEON, July 28 (Korea Bizwire) — The Korea Advanced Institute of Science and Technology (KAIST) said Wednesday it had developed new artificial intelligence (AI) technology that can automatically decode chest X-ray images, identifying illnesses such as tuberculosis, pneumothorax and COVID-19.
In collaboration with a domestic hospital, the research team developed a new AI algorithm featuring self-supervised and self-train learning (DISTL).
The algorithm involves self-learning and teacher-student knowledge transfer techniques to mimic the process in which radiologists learn to read images at hospitals.
If the initial model is trained with only a small quantity of label data, the model can improve its performance autonomously with only label-free data that accumulates over time.
The research team said that it’s an original technology to be able to use countless chest X-ray images collected through annual health checkups.
Most currently used medical AI techniques are based on the supervised learning method, which served as a stumbling block preventing the development of medical AI due to the problem of continuously obtaining expert data labels that requires a huge amount of cost and time.
M. H. Lee (email@example.com)