Medical Image Analysis

- Development of an AI method that enables efficient large-scale model training while protecting personal information with only a single model transmission - Overcomes previous limitations of computational cost and overfitting… expected to be applied in medical image analysis - Research results published in Medical Image Analysis, October 2025, a top-tier journal in the field of medical image analysis

Perimetry (visual field testing) quantifies a patient’s retinal sensitivity to light and clarifies a deviation from normal retinal sensitivity. Visual field tests generally require high patient concentration, which can be exhausted. We constructed a framework for deep reinforcement learning to train ViFT (Visual Field Transformer), which controls all processes of visual field testing. ViFT achieves the same or higher accuracy than the state-of-the-art strategies, with less than half the test time of the state-of-the-art strategies.

- Professor Sanghyun Park’s research team in the Department of Robotics and Mechatronics Engineering successfully developed a novel AI model that can effectively utilize medical images from multiple healthcare institutions with the knowledge distillation technique based on federated learning - Findings were published in Medical Image Analysis (MedIA), one of the top journals in medical AI

- DGIST Professor Park Sang-Hyun's research team developed a deep learning technology that accurately detects cancer sites with little information through joint research with a research team led by Professor Nam Su-jeong and Professor Ko Hyeon-Jeong of Seoul Asan Medical Center - Expected to make significant contribution to enhancing the efficiency of deep learning models, which needed an accurately drawn dataset to detect cancer sites.

Scientists develop new algorithm for rapid, computerized diagnosis of COVID-19, overcoming the limitations of reverse transcription polymerase chain reaction