AI Researcher & Ph.D. Candidate, Yonsei UniversityMedical Informatics Multimodal AI Federated Learning
My research aims to build medical AI systems that learn from diverse medical data and address real-world healthcare needs. I focus on multimodal AI to integrate information across clinical modalities, and federated learning to enable collaboration across institutions where high-value medical data cannot be easily shared.
BiographyI am currently a Ph.D. candidate in the Department of Biomedical Systems Informatics at Yonsei University College of Medicine, supervised by Prof. Yu Rang Park. I also work as an AI Researcher at the Digital Healthcare Laboratory (DHLab) in Yonsei University College of Medicine and Severance Hospital in Seoul, Republic of Korea. Previously, I received my B.B.A. in Big Data Management and Statistics from Kookmin University in Seoul, Republic of Korea. From July to September in 2021, I worked as a Data Scientist at DACON, which is the largest data science competition platform in Korea.

Hong JaeSeong, Minkyu Kim, Yu Rang Park
Information Fusion 2026 Medical Dataset Multimodal
This study addresses the challenge by training modality-specific encoders independently while aligning them to a shared Equiangular Tight Frame (ETF) decision geometry, then composing only the observed modalities through metadata-conditioned fusion.
Hong JaeSeong, Minkyu Kim, Yu Rang Park
Information Fusion 2026 Medical Dataset Multimodal
This study addresses the challenge by training modality-specific encoders independently while aligning them to a shared Equiangular Tight Frame (ETF) decision geometry, then composing only the observed modalities through metadata-conditioned fusion.

Hangnyoung Choi*, Hong JaeSeong*, Hyun Goo Kang*, Min-Hyeon Park, Sungji Ha, Junghan Lee, Sangchul Yoon, Daeseong Kim, Yu Rang Park, Keun-Ah Cheon (* equal contribution)
npj Digital Medicine 2025 Medical Dataset Deep+Machine Learning
This study demonstrated that machine learning analysis of retinal fundus photographs can serve as a noninvasive biomarker for screening attention-deficit/hyperactivity disorder (ADHD) and stratifying executive function deficits, achieving up to 96.9% AUROC and showing strong performance particularly in the visual attention domain.
Hangnyoung Choi*, Hong JaeSeong*, Hyun Goo Kang*, Min-Hyeon Park, Sungji Ha, Junghan Lee, Sangchul Yoon, Daeseong Kim, Yu Rang Park, Keun-Ah Cheon (* equal contribution)
npj Digital Medicine 2025 Medical Dataset Deep+Machine Learning
This study demonstrated that machine learning analysis of retinal fundus photographs can serve as a noninvasive biomarker for screening attention-deficit/hyperactivity disorder (ADHD) and stratifying executive function deficits, achieving up to 96.9% AUROC and showing strong performance particularly in the visual attention domain.

Hong JaeSeong*, Sangchul Yoon*, Kyu Won Shim, Yu Rang Park (* equal contribution)
Stroke 2024 Medical Dataset Deep Learning
This study developed a deep learning model using retinal fundus photographs to screen for and stage Moyamoya disease (MMD), achieving high diagnostic performance (AUROC = 94.6%) and identifying retinal vascular regions as key features, suggesting that retinal imaging may serve as a noninvasive biomarker for MMD detection and progression assessment.
Hong JaeSeong*, Sangchul Yoon*, Kyu Won Shim, Yu Rang Park (* equal contribution)
Stroke 2024 Medical Dataset Deep Learning
This study developed a deep learning model using retinal fundus photographs to screen for and stage Moyamoya disease (MMD), achieving high diagnostic performance (AUROC = 94.6%) and identifying retinal vascular regions as key features, suggesting that retinal imaging may serve as a noninvasive biomarker for MMD detection and progression assessment.

Jae Han Kim*, Hong JaeSeong*, Hangnyoung Choi*, Hyun Goo Kang*, Sangchul Yoon, Jung Yeon Hwang, Yu Rang Park, Keun-Ah Cheon (* equal contribution)
JAMA Network Open. 2023 Medical Dataset Deep Ensemble
This study developed deep learning models using retinal fundus photographs to screen for autism spectrum disorder (ASD) and assess symptom severity, achieving perfect diagnostic accuracy for ASD detection and demonstrating the optic disc as a key biomarker for objective, accessible ASD screening.
Jae Han Kim*, Hong JaeSeong*, Hangnyoung Choi*, Hyun Goo Kang*, Sangchul Yoon, Jung Yeon Hwang, Yu Rang Park, Keun-Ah Cheon (* equal contribution)
JAMA Network Open. 2023 Medical Dataset Deep Ensemble
This study developed deep learning models using retinal fundus photographs to screen for autism spectrum disorder (ASD) and assess symptom severity, achieving perfect diagnostic accuracy for ASD detection and demonstrating the optic disc as a key biomarker for objective, accessible ASD screening.