Research | DHLab

Research Topics at DHL

DHLab builds the algorithms, data, and clinical tools that turn healthcare signals into decisions. Our research spans four connected pillars: digital phenotyping, medical AI algorithms, multi-omics, and quantum computing for medicine.

Digital Phenotyping — Autism, Panic Disorder, Sleep, and Beyond

We derive digital phenotypes from smartphones, wearables, interaction videos, and sleep sensors, and translate them into screening tools and digital therapeutics for autism, panic disorder, and neurodegenerative disease. Our work moves from data collection to prospective clinical validation and regulatory-track development.

Selected work

Kim T*, Jung D*, Kim KM, …, Lee PH†, Park YR†. "Sleep Disorders and Sleep Behaviors as Predictors of Neurodegenerative Diseases." Alzheimer's & Dementia 22(3):e71301 (2026).
Jang S, Sun TH, …, Park YR, Cho CH. "A digital phenotyping dataset for impending panic symptoms: a prospective longitudinal study." Scientific Data 11(1):1264 (2024).
Chun S, Jang S, Kim J, Ko C, Lee J, Hong J, Park Y. Comprehensive Assessment and Early Prediction of Gross Motor Performance in Toddlers With Graph Convolutional Networks–Based Deep Learning: Development and Validation Study. JMIR Form Res 2024;8:e51996

Medical AI Algorithm Development

We design AI algorithms for the real constraints of clinical data — incomplete and asynchronous inputs, distributed cohorts, evolving distributions, and privacy. Our recent work spans multimodal information fusion, patch-specialized foundation models, continual learning, and federated collaboration, and appears at top medical and AI venues including CVPR, Information Fusion, and npj Digital Medicine.

Selected work

Park JY, Seo JY, Kang MJ, Park YR. "MoECLIP: Patch-Specialized Experts for Zero-shot Anomaly Detection." CVPR 2026, p. 35534–35544 — the #1 AI venue by Google Scholar Metrics and CSRankings.
Hong JS, Kim M, Park YR. "Geometry-guided unimodal-to-multimodal composition for incomplete and asynchronous clinical information fusion." Information Fusion 104725 (2026).

Multi-omics for Disease Mechanisms and Biomarkers

We integrate spatial transcriptomics, label-free 3D cell morphology, and clinical genomics to uncover mechanisms of infection, inflammation, and cancer — and to translate those signals into AI-based biomarkers for sepsis, CNS infection, and pediatric inflammatory bowel disease.

Selected work

Jang S, Lee EJ, …, Park YR. "Spatial host-microbiome profiling demonstrates bacterial-associated host transcriptional alterations in pediatric ileal Crohn's disease." Microbiome 13(1):189 (2025).
Sung M*, Kim JH*, Min HS*, …, Chung KS†, Park YR†. "Three-dimensional label-free morphology of CD8+ T cells as a sepsis biomarker." Light: Science & Applications 12:265 (2023).

Quantum Computing for Medicine

As one of the earliest Korean groups exploring quantum computing for medicine, we investigate how quantum sampling and quantum-classical hybrid models can help medical AI where data are scarce or noisy — from rare-cancer survival to probabilistic inference over psychiatric time-series data.

Selected work

Yu JY, Sim WS, Jung JY, Park SH, Kim HS, Park YR. "Evaluation of Conventional and Quantum Computing for Predicting Mortality Based on Small Early-Onset Colorectal Cancer Data." Applied Soft Computing 111781 (2024).
Lee JY, Sim WS, Park YR. "Empirical Validation of Medical Data-Based Bayesian Network Inference Using Quantum Sampling." KSAIM 2025 Fall.