Auto CT Pelvimetry

Automated CT-based pelvimetry pipeline for measuring pelvic bony landmarks, validated against manual 3D Slicer measurements.

Overview

This pipeline extracts 6 pelvic morphometric measurements from CT scans using AI-based segmentation (TotalSegmentator):

Metric	Description
ISD	Interspinous Distance (mm)
Inlet AP	Pelvic Inlet Antero-Posterior Diameter (mm)
Outlet AP	Pelvic Outlet Antero-Posterior Diameter (mm)
Outlet Transverse	Intertuberous Distance (mm)
Sacral Length	Promontory → Coccygeal Apex (mm)
Sacral Depth	Maximum Anterior Sacral Concavity (mm)

Additionally, Outlet Area (cm²) is derived as an ellipse approximation (π/4 × Outlet AP × Outlet Transverse).

Repository Structure

├── Pipeline/
│   ├── Combined_Pelvimetry_Pipeline_v6.3.py   # Main pipeline
│   ├── Error_Handling_System.py                # Metric-level error tracking (required)
│   ├── Body_Composition_Analysis.py            # Body composition analysis (see note below)
│   └── requirements.txt                        # Pipeline-specific dependencies
├── Notebooks/
│   ├── 1_Data_Cleaning.ipynb                   # Cohort construction & data merging
│   ├── 2_Validation.ipynb                      # ICC, Bland-Altman analysis
│   └── 3_Clinical_Association.ipynb            # Clinical outcome association
├── README.md
└── .gitignore

Note: Error_Handling_System.py must be in the same directory as the main pipeline. It provides MetricResult, ErrorLogger, and safe_execute for per-metric error isolation and structured failure logging.

Body Composition Analysis

Body_Composition_Analysis.py is a standalone module for computing body composition metrics (VAT, SAT, V/S ratio, skeletal muscle area) at the L3 vertebral level and mid-pelvis (ISD level) from TotalSegmentator tissue-type segmentations. It depends on the pelvimetry pipeline's CSV output for ISD-level slice localization. This module was not used in the current study but is included for reference and potential future analyses.

Requirements

Python Dependencies

pip install -r requirements.txt

External Tools

Tool	Purpose	Installation
dcm2niix	DICOM → NIfTI conversion	brew install dcm2niix (macOS)
TotalSegmentator	AI anatomical segmentation	pip install TotalSegmentator

Usage

Single Patient

cd Pipeline/Main_Pipeline

python Combined_Pelvimetry_Pipeline_v6.3.py \
    --mode single \
    --patient-id Patient_001 \
    --dicom-dir /path/to/Patient_001/DICOM \
    --output-dir /path/to/output

Batch Processing

python Combined_Pelvimetry_Pipeline_v6.3.py \
    --mode batch \
    --input-dir /path/to/all_patients \
    --output-dir /path/to/output

Pre-segmented Data (Pelvimetry Only)

If segmentations already exist:

python Combined_Pelvimetry_Pipeline_v6.3.py \
    --mode measure \
    --nifti /path/to/patient.nii.gz \
    --seg-dir /path/to/segmentation_folder \
    --output-dir /path/to/output

Output

• CSV: One row per patient with all metric values, anatomical landmark coordinates (World RAS), and pipeline status
• QC Figures: Sagittal combined and extended axial QC images (per patient)
• Error Log: Structured error log with per-metric failure taxonomy

Status Codes

Status	Meaning
Success	All 6 metrics computed
Partial_N/6	N of 6 metrics computed; see Error_Log for details
Failure	No metrics computed

License

This project is provided for academic and research purposes.