Colon & Code
clinical problems → structured data → deployable systems
Building real-world AI systems for surgical workflows
I build an end-to-end surgical data pipeline that connects:
CT imaging → intraoperative workflow → postoperative outcomes
From raw clinical data to reproducible analysis and deployable systems.
高雄榮總大腸直腸外科主治醫師
專注於將臨床問題轉化為可計算、可重現、可部署的資料科學系統
CT Imaging → Feature Extraction → Difficulty Modeling ↓ Intraoperative Video / Workflow Analysis ↓ Postoperative Outcomes / PRO / Prediction Models
- Automated CT-based pelvimetry
- Surgical difficulty modeling (FREDRIC framework)
- Learning curve modeling (RA-CUSUM)
- Video-based workflow analysis
- Outcome prediction (ML / survival analysis)
- Digital follow-up & PRO systems
ctpelvimetry
→ Fully automated CT-based pelvimetry
→ Published & validated (IJCARS)
→ Packaged and distributed via PyPI
RiSSA_ML_Learning_Curve
→ ML-based surgical safety profiling
→ Published in Journal of Robotic Surgery
Stage_III_Colon_EDR
→ Early recurrence prediction
→ Multi-center validation
Hemorrhoids_PostOp
→ Digital postoperative monitoring system
→ Deployed in IRB-approved clinical study
Focus: surgical AI, learning curves, and outcome modeling
- Machine learning–based learning curve analysis — J Robotic Surg. 2026
doi - Video-based RA-CUSUM proficiency assessment — Int J Colorectal Dis. 2026
doi - Robotic single-stapling vs double-stapling anastomosis — J Robotic Surg. 2025
doi
Full list → ORCID
- Data Science: pandas, scikit-learn, lifelines, PyTorch
- Imaging: CT processing, TotalSegmentator, 3D Slicer
- Causal Inference: overlap weighting, RMST, survival modeling
- Systems: Next.js, TypeScript, PostgreSQL
Building the infrastructure of Surgical Data Science:
- From clinical intuition → quantitative modeling
- From retrospective data → real-time systems
- From isolated studies → reproducible pipelines
- Next: surgical video analytics — automated phase recognition, workflow decomposition, and AI-assisted intraoperative feedback