Auto-ISD · Interactive Demo

Interactive companion to Huang S-F, Tseng H-P, Hsu C-W. A fully automated CT-based pelvimetry pipeline for quantifying mid-pelvic surgical workspace in rectal cancer. Int J Comput Assist Radiol Surg. 2026.

Read the paper (DOI: 10.1007/s11548-026-03606-2) Pipeline source code Live demo

What this is

A scrollytelling web demo of the Auto-ISD pelvimetry pipeline, built as a CARS / ASCRS audience-facing companion to the IJCARS paper. Walks through:

1. The problem of inter-rater variability in manual pelvimetry
2. The pipeline as a deterministic terminal flow
3. The valley-detection algorithm, animated step by step
4. 3D anatomical reconstructions across 5 representative cases
5. The posterior pelvic triangle and soft-tissue occupancy metrics
6. Validation summary (full Bland–Altman / ICC tables in the paper)

Stack

• Vanilla HTML / CSS / JS, single file
• Three.js r128 for 3D bone meshes
• GSAP ScrollTrigger for scroll-driven act transitions
• Lenis for smooth scrolling
• SVG-based algorithm visualization with auto-loop animation
• Hosted on Vercel

Asset structure

auto-isd-site/
├── index.html              # everything: HTML, CSS, JS in one file
├── og-image.png            # 1200x630 OG share card
├── vercel.json             # cache headers (mesh = immutable, JSON = revalidate)
└── assets/
    ├── logo*.png
    ├── cases/
    │   ├── manifest.json          # case list with summary stats
    │   ├── Patient_005/           # five representative cases
    │   ├── Patient_011/
    │   ├── Patient_017/
    │   ├── Patient_044/
    │   └── Patient_086/
    │       ├── results.json       # ISD, triangle, soft-tissue metrics + trace
    │       ├── landmarks.json     # 3D landmark world coordinates
    │       └── meshes/            # bone meshes, decimated to ~7500 verts
    │           ├── hip_L.json
    │           ├── hip_R.json
    │           ├── sacrum.json
    │           ├── femur_L.json
    │           └── femur_R.json

Local dev

python3 -m http.server 8000
open http://localhost:8000

No build step.

License

Demo code: MIT. Patient imaging data is not redistributable per institutional review board terms; the cases shown here are anonymized derived geometries (mesh-decimated bone surfaces from segmentation masks, no raw CT data).

Citation

If this work is useful, please cite the paper:

Huang S-F, Tseng H-P, Hsu C-W. A fully automated CT-based pelvimetry pipeline for quantifying mid-pelvic surgical workspace in rectal cancer. Int J Comput Assist Radiol Surg. 2026. https://doi.org/10.1007/s11548-026-03606-2

Author

Shih-Feng Huang, MD — Division of Colorectal Surgery, Kaohsiung Veterans General Hospital