Auto-ISD: Automated CT-Based Pelvimetry Pipeline

Source code for the automated pelvimetry pipeline described in:

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. DOI: 10.1007/s11548-026-03606-2

Research artifact	Evidence
Publication	Peer-reviewed article in IJCARS
Reproduction	Core Python implementation, demo CT, segmentation masks, and QC notebook
Interactive explanation	Live scrollytelling demo
Reusable package	Maintained separately as ctpelvimetry

Overview

The pipeline automates the extraction of key anatomical metrics from routine staging CT scans:

1. Interspinous Distance (ISD): The narrowest transverse distance at the ischial spine level, identified via valley detection or plateau fallback.
2. Posterior Pelvic Triangle: Area, depth, and shape index — a geometric representation of the mid-pelvic surgical workspace.
3. Soft Tissue Occupancy: Bowel area, posterior pelvic fat area (pPFA), occupancy ratios, and residual working space.

Prerequisites

1. Python 3.8+

2. Input Data Requirements: To run the analysis code, you must first process your CT scans to obtain:

   • Original CT Image: In NIfTI format (.nii.gz).
   • Segmentation Masks: Generated via TotalSegmentator. The following masks are required:
     • femur_left.nii.gz, femur_right.nii.gz
     • hip_left.nii.gz, hip_right.nii.gz
     • sacrum.nii.gz
     • colon.nii.gz
     • torso_fat.nii.gz (for body composition)

   Citation for TotalSegmentator: Wasserthal, J., Breit, H. C., Meyer, M. T., Pradella, M., Hinck, D., Sauter, A. W., Heye, T., Boll, D. T., Cyriac, J., Yang, S., Bach, M., & Segeroth, M. (2023). TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images. Radiology. Artificial intelligence, 5(5), e230024. DOI

Installation

1. Clone this repository or download the source code.
2. Install Python dependencies:

pip install -r requirements.txt

3. Ensure TotalSegmentator is installed if you need to generate masks from scratch:

pip install TotalSegmentator

TotalSegmentator License

Bone-based metrics (ISD, triangle geometry) work with the free version of TotalSegmentator — no license needed.

For soft tissue metrics (pPFA, working space), the tissue_types task requires an academic license:

# 1. Apply for an academic license at:
#    https://github.com/wasserth/TotalSegmentator#license

# 2. Set the license as an environment variable:
export TOTALSEG_LICENSE='your_license_key_here'

# 3. Register the license:
totalseg_set_license -l $TOTALSEG_LICENSE

Note: If you only need bone-derived pelvimetry (ISD, triangle area/depth), you can skip the license step entirely. The pipeline will still compute all bone-based metrics successfully.

Usage

1. Demo Data

A Demo folder is included in this package containing a sample patient:

• Demo/Patient_CT.nii.gz: Original CT.
• Demo/*.nii.gz: Pre-computed segmentation masks. You can use this data to verify the code immediately.

2. Running the Code

The provided code performs two distinct tasks:

A. Metric Calculation (Pelvimetry_Demo.ipynb)

Calculates all anatomical metrics (ISD, APD, Triangle Area, pPFA, Working Space, etc.) and outputs them numerically.

• Input: NIfTI file + Segmentation Masks folder.
• Output: Printed metrics and coordinate points.

B. QC Figure Generation (Generate_QC_Plot.ipynb)

Generates a publication-quality Quality Control (QC) figure showing the anatomical landmarks and the ISD search profile curve.

• Input: NIfTI file + Segmentation Masks folder.
• Output: A high-resolution PNG file (Demo_QC_Plot.png) visualizing the analysis.

3. Library Usage

The core logic is encapsulated in pelvimetry_core.py. You can import AutomatedPelvimetry to analyze your own NIfTI files.

from pelvimetry_core import AutomatedPelvimetry

# Initialize
pipeline = AutomatedPelvimetry()

# Run Analysis
# Note: Ensure segmentation masks are present in 'seg_output_dir'
results = pipeline.pipeline_single_case(
    nifti_path="path/to/patient_ct.nii.gz",
    seg_output_dir="path/to/output_folder_with_masks"
)

print(f"ISD: {results['ISD_mm']} mm")

Citation

If you use this code in your research, please cite:

@article{huang2026autoiSD,
  title={A fully automated CT-based pelvimetry pipeline for quantifying mid-pelvic surgical workspace in rectal cancer},
  author={Huang, Shih-Feng and Tseng, Hsin-Ping and Hsu, Chao-Wen},
  journal={International Journal of Computer Assisted Radiology and Surgery},
  year={2026},
  doi={10.1007/s11548-026-03606-2}
}

License

MIT License — Academic use encouraged. See LICENSE for details.

Machine-readable citation metadata are available in CITATION.cff.