Jung-Oh Lee, MD, MSc · Clinical Assistant Professor of Radiology, NYU Langone Health
Advancing radiology AI for real clinical impact.
Board-certified radiologist and AI researcher developing clinically grounded AI for medical imaging, with a focus on 3D imaging, foundation models, agentic systems, and rigorous clinical evaluation.
Future Directions
Agentic Data Curation
AI agents that autonomously curate and annotate large-scale imaging archives — turning raw clinical data into scalable, high-quality training sets.
Real-World Radiology Benchmarks
International, multicenter benchmarks grounded in real clinical practice — spanning cancer detection and differential diagnosis.
3D Foundation Models & Agents
Foundation models and agentic systems that reason across entire 3D studies and longitudinal exams, not isolated slices.
Recent Research
Bone Metastasis Detection on CT
An expert-comparable 3D model, validated across multiple centers against CT, MRI, and PET/CT reference standards — and released publicly with open model weights.
Multimodal Generative AI for 3D Imaging
A perspective on how generative models should interpret volumetric scans and clinical videos — laying out the core capabilities, opportunities, and open challenges for 3D and video-based medical AI.
Evaluation Systems for Radiology AI
Benchmarks and frameworks for evaluating the reliability of radiology AI models — spanning visual question answering, entity-level clinical safety, structured diagnostic reasoning, and LLM-based systematic analysis.
AI for Cancer Imaging & Prognosis
A 3D deep learning model that extracts prognostic features from preoperative brain MRI, adding prognostic value on top of established clinical and molecular markers in adult-type diffuse glioma.
Selected Publications
Radiology: Artificial Intelligence · 2026 — A multicenter, externally validated 3D AI model for detecting bone metastases on CT, evaluated against expert radiologists and released publicly.
npj Digital Medicine · 2025 — A perspective on generative models for interpreting volumetric medical images and clinical videos — capabilities, opportunities, and open challenges.
Neuro-Oncology · 2024 — 3D deep learning features from preoperative MRI add independent prognostic value beyond clinical and molecular markers in diffuse glioma.
MICCAI · 2025 — An entity-level evaluation framework for chest X-ray report generation focused on clinical safety rather than text overlap.
NeurIPS (PhysioNet) · 2025 — A benchmark evaluating whether AI models follow clinically valid diagnostic reasoning steps on chest X-rays.
Full list on Google Scholar.
Vision
“AI is the new electricity.”
I believe this — and radiology is where that current reaches patients. I am always looking to meet people who believe it too, so we can advance radiology AI together.
Collaborate
Let's build clinically impactful AI together. Whether you bring a sharp clinical question, high-quality imaging data, or a bold idea, I'd love to hear from you.