Academic Research
Clinically grounded AI for burn care, craniofacial and reconstructive surgery.
My research sits at the intersection of plastic & reconstructive surgery and applied AI. Most projects start from a problem I meet in the burn unit or operating room, and aim for tools that are genuinely usable in clinical workflow rather than benchmark-only models.
Clinical Research
Clinical-question-driven studies grounded in real practice and large-scale hospital data.
Quantitative rhinoplasty. Photogrammetric, 3D-based frameworks for measuring and predicting nasal tip projection, rotation, and stability after cartilage grafting.
Burn outcomes from the Chang Gung Research Database (CGRD). Large-scale clinical outcome studies — transfusion, infection and antimicrobial resistance, and length of stay — across decades of burn admissions.
Craniofacial & orthognathic measurement. 3D photogrammetry and CBCT-based assessment for orthognathic surgery planning, including Bad Split risk.
AI & Computational
Deep-learning, computer-vision, and language methods built from the bedside outward.
Burn wound segmentation & TBSA estimation. Automatic segmentation and total-body-surface-area estimation from clinical photographs.
Inhalation-injury prediction. Machine-learning and multimodal models for early risk stratification of inhalation injury.
Craniofacial deep learning. Landmark detection and segmentation on CBCT and 3D imaging — with a focus on the systematic errors that off-the-shelf Western models make on Asian Class III anatomy.
Multimodal clinical AI. Vision–language and NLP methods for extracting structure from clinical narratives, imaging, and structured records.
AI demos
Live, interactive demos of tools coming out of the lab — hosted on Hugging Face Spaces. The full collection lives on the VDI Lab site.
- Burn wound segmentation — Deep Supervision UNet++ (IoU 0.85) → demo
- Flap perfusion prediction — temperature + color features → demo
Note: a 2022 research prototype with known patient-level data leakage — shown for transparency, not as a clinical tool.
For the lab, team, and complete project list, visit lab.cjhuang.com →