Jeonbuk National University (JBNU) College of Medicine student Sehyun Lee, a first-year medical student, received an Outstanding Presentation Award at the 2026 Spring Conference hosted by the Korean Society of Medical Informatics (KOSMI). Since 2024, Lee has achieved the unusual feat of winning awards five times at major conferences in the fields of medical informatics and medical artificial intelligence.

The Korean Society of Medical Informatics is the leading academic society in Korea for medical informatics, and its spring and fall conferences are major venues where medical AI researchers present their findings. It is considered rare in the Korean medical community for a first-year medical student who has not yet begun clinical rotations to win consecutive awards at nationwide academic conferences.

The awarded study proposed an artificial intelligence model that identifies patients at high risk of lung metastasis using only preoperative FDG-PET and MRI images for the rare cancer soft-tissue sarcoma. The research was conducted under the supervision of Professor Youngjae Moon of the Department of Orthopedics at Jeonbuk National University Hospital and Professor Hyunho Kim of the Department of Pediatrics at Seoul St. Mary’s Hospital, The Catholic University of Korea.

Soft-tissue sarcoma is a disease in which the presence of lung metastasis greatly affects patient prognosis. If lung metastasis occurs, the five-year survival rate can drop from the 90% range to around 25%, but there are currently insufficient tools to objectively identify patients at high risk of metastasis using only preoperative imaging.

The research team integrated and analyzed FDG-PET and MRI scans using a large-scale foundation model trained on medical images. The analysis showed that the metabolic signals of tumors captured by PET scans were the main drivers of lung metastasis risk prediction, and the regions highlighted by the AI model corresponded to areas of metabolic tumor activity.

This finding is significant because it suggests that AI can do more than provide risk scores; it can potentially link its decision basis to the biological characteristics of tumors and provide explanations. Notably, the model passed statistical validation even with a small dataset, and the actual survival difference between the model-classified high-risk and low-risk groups was approximately 18-fold.

The team aims to further develop the tool into a prognostic prediction instrument that clinicians can trust, to assess individual patient risk before surgery and guide appropriate treatment interventions. They also plan external validation using a cohort from Jeonbuk National University Hospital and follow-up studies combining pathology and omics data.

Sehyun Lee said, "I am sincerely grateful to Professor Youngjae Moon and Professor Hyunho Kim for believing in and guiding me. I will become a physician-scientist who develops AI that clinicians can trust and that can change patient prognoses in clinical practice."

Professor Youngjae Moon said, "It is very rare for an undergraduate student to lead and complete a multimodal AI study handling multicenter imaging data." He added, "Since lung metastasis largely determines the prognosis of soft-tissue sarcoma and it is difficult to estimate risk before surgery, this study could contribute to future prognosis prediction and treatment decisions for patients."

Professor Hyunho Kim also said, "For imaging AI to be used in actual clinical practice, it must be able not only to perform well but also to explain the basis for its decisions." He added, "This study is significant in that it shows the direction that explainable medical AI should take."

Meanwhile, Sehyun Lee has continued research in neuro- and pediatric imaging AI since undergraduate studies and has published successive first-author papers in SCI journals. This year, Lee has presented research results in journals such as Annals of Biomedical Engineering, an international journal in biomedical engineering, gaining recognition as a medical AI researcher.