Skip to content

README.md

Latest commit

 

History

History
142 lines (112 loc) · 5.69 KB

README.md

File metadata and controls

142 lines (112 loc) · 5.69 KB
Raw

DirLab-4DCT

Download

Manual download required - this dataset is not fetched by monai-physio-download-data; there is no automatic downloader because DIR-Lab distributes each case individually and may require registration.

  1. Visit the DIR-Lab 4D-CT page and request/download the case archives: https://med.emory.edu/departments/radiation-oncology/research-laboratories/deformable-image-registration/downloads-and-reference-data/4dct.html
  2. Extract each case's raw images into data/DirLab-4DCT/downloaded_data/ (see layout below). The .mhd headers already committed in that directory point at those raw files, so no header-writing step is required - see "About the Committed .mhd Files" below.
  3. Run data/DirLab-4DCT/fix_downloaded_data.py. DIR-Lab's raw volumes are not in Hounsfield units (see "Fixing Raw Intensities" below); this script reads every .mhd header in downloaded_data/ with its backing pixel data present, corrects the intensities, and writes one compressed .mha volume per phase directly into data/DirLab-4DCT/. Tutorials and experiments read these top-level .mha files, not the raw .mhd ones.

Once populated, check the layout with:

from monai_physio import DownloadData

assert DownloadData.VerifyDirLab4DCTData("data/DirLab-4DCT")

Directory structure after download and fixing:

data/DirLab-4DCT/
├── downloaded_data/
│   ├── Case1Pack/
│   │   ├── Images/              # T00-T90 phase images
│   │   ├── ExtremePhases/       # T00 and T50 (max inhale/exhale)
│   │   └── Sampled4D/           # Sampled time points
│   ├── Case1Pack_T00.mhd        # Already-committed headers, per phase
│   ├── Case1Pack_T10.mhd
│   ...
│   ├── Case10Pack/
│   └── Convert4DCTToMHD.py      # Documents how the .mhd headers were generated
├── Case1Pack_T00.mha             # Written by fix_downloaded_data.py
├── Case1Pack_T10.mha
...
├── fix_downloaded_data.py
└── README.md (this file)

About the Committed .mhd Files

The Case*Pack_T*.mhd files in downloaded_data/ are already committed to the repository, but they are only MetaImage headers (a few hundred bytes each) - for example:

ObjectType = Image
NDims = 3
DimSize = 256 256 94
...
ElementDataFile = Case1Pack/Images/case1_T00_s.img

Each header's ElementDataFile points at the raw image data inside the corresponding Case*Pack//Case*Deploy/ subdirectory (.gitignored because those raw volumes are large). These .mhd files will not load until you complete the manual download above and the referenced Case*Pack/Images/*.img files exist alongside them.

Convert4DCTToMHD.py is what originally generated these headers from the raw DIR-Lab archives. It is included for provenance/documentation only - you do not need to run it; the .mhd files are already committed.

Fixing Raw Intensities

DIR-Lab's raw .mhd/.img volumes store intensities offset by +1024 from Hounsfield units, not real HU. fix_downloaded_data.py calls DownloadData.FixDirLab4DCTData, which subtracts 1024 and clips the result to [-1024, 1024] for every downloaded phase, writing one compressed .mha file per phase into data/DirLab-4DCT/ (.gitignored, like the raw data). Run it once after downloading; tutorials and experiments assume the .mha files it produces already exist.

Overview

Benchmark dataset for 4D CT respiratory motion analysis. Contains 10 cases of lung CT scans at different respiratory phases with annotated landmark points for registration validation.

Dataset Details

  • Format: .mhd headers + .img raw volumes (MetaImage format)
  • Cases: 10 patient cases (Case 1-10)
  • Phases: 10 respiratory phases per case (T00-T90)
  • Content: Non-contrast lung CT
  • Anatomy: Lungs, airways, thoracic structures
  • Landmarks: 300+ annotated points per case for registration validation

Acknowledgement

Data provided by the DIR-Lab at MD Anderson Cancer Center / Emory University: https://med.emory.edu/departments/radiation-oncology/research-laboratories/deformable-image-registration/

Citation

Dataset: https://med.emory.edu/departments/radiation-oncology/research-laboratories/deformable-image-registration/downloads-and-reference-data/4dct.html

If you use this dataset, please cite:

  • Case numbers 4DCT1-4DCT5: Castillo R, Castillo E, Guerra R, Johnson VE, McPhail T, Garg AK, Guerrero T. 2009. "A framework for evaluation of deformable image registration spatial accuracy using large landmark point sets." Phys Med Biol 54:1849-1870.
  • Case numbers 4DCT6-4DCT10: Castillo E, Castillo R, Martinez J, Shenoy M, Guerrero T. 2009. "Four-dimensional deformable image registration using trajectory modeling." Phys Med Biol 55:305-327.

Using This Dataset

  • Primary dataset for experiments/Lung-GatedCT_To_USD/
  • Registration algorithm validation
  • Respiratory motion analysis
  • Benchmark for deformable registration accuracy

Files in This Directory

  • fix_downloaded_data.py - converts downloaded_data/*.mhd to HU-corrected Case*Pack_T*.mha files in this directory (see "Fixing Raw Intensities" above)
  • downloaded_data/Convert4DCTToMHD.py - documents how the committed .mhd headers were generated from the raw DIR-Lab archives; not needed to run
  • downloaded_data/Case*Pack_T*.mhd - MetaImage headers for each case/phase (see above)
  • Case*Pack_T*.mha - HU-corrected volumes written by fix_downloaded_data.py; what tutorials and experiments actually read

Additional Resources