POSTECH: POSTECH-ETRI-AMC Win 2nd Place at World’s Top Medical AI Competition
A Korean research team recently placed second among 18 teams from around the world in the Anatomical Tracings of Lesions After Stroke (ATLAS) Challenge. This challenge was hosted by the Brain Lesion Workshop and sponsored by Amazon Web Services (AWS) and Medical Image Computing and Computer Assisted Intervention (MICCAI).
The joint team of researchers are: PhD candidate Hyunsu Jeong from the Graduate School of Artificial Intelligence and Chiho Yoon from the Department of Electrical Engineering (in the Master’s program) at POSTECH; Dr. Kwang-Ju Kim from the Electronics and Telecommunications Research Institute (ETRI); and Professor Bum Joon Kim and Soo Jeong (fellow) from the Department of Neurology, and Won-jung Chung (researcher, advisor Namkug Kim) at the Asan Medical Center (AMC).
The competition challenged the participants to develop new automated stroke lesion segmentation algorithms. The condition was to use only T1-weighted (T1w) MRIs, which are commonly used with stroke patients in rehabilitation research. It is not easy to conduct accurate lesion segmentation with the limited information from T1w MRI images alone.
The team focused on pre- and post-interventions by making the most of the T1w stroke MRIs with the help from the most basic model, U-Net, in deep learning segmentation, instead of the latest deep learning models. When it comes to medical data, the key is to design pre- and post-intervention methods to read all data set types, considering that various imaging machines are used in hospitals overseas. Going forward, the researchers plan to take one step further by improving the deep learning model based on the medical imaging technology they built during this challenge.
MICCAI is a globally renowned conference in the field of medical artificial intelligence, serving as a venue for researchers and doctors worldwide. MICCAI 2022, which was held in Singapore, attracted leading scientists, engineers, and businesses from around the world, including NVIDIA (the US), University of Toronto (Canada), King’s College London (the UK), Catholic University of Leuven (KU Leuven, Belgium), and National University of Singapore (Singapore). The top three prizes were awarded to King’s College London (1st place), POSTECH (2nd place), and KU Leuven (3rd place).