fedorov / prostate_mri_us_biopsy_dcm_conversion

Examples for converting prostate MRI-US-Biopsy STL prostate and lesion surfaces to DICOM SEG and DICOM STL.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Prostate_mri_us_biopsy_dcm_conversion

Provides examples for converting prostate MRI-US-Biopsy collection STL prostate and lesion surfaces to DICOM SEG and DICOM STL.

Conversion to DICOM process

Conversion to DICOM SEG

  1. STL to labelmap -- Utils folder
  2. labelmap to SEG DICOM -- see DEMO_Conversion_STL_TO_DICOM.ipynb

Conversion to DICOM STL

  1. STL to DICOM STL -- see DEMO_Conversion_STL_TO_DICOM.ipynb

Repository organization

Data samples folder

The folder contains STL and labelmaps samples for prostate and lesion surfaces, used for conversion showcase. We include the already computed labelmaps from Slicer conversion, since we cannot showcase it inside Google Colab.

metadata folder

Contains segments metadata.json files used as input for conversion to DICOM SEG, for both lesion and prostate surfaces.

Utils folder

Contains example of slicer script used to convert STL files into labelmaps with reference T2W images.

Notebook

DEMO_Conversion_STL_TO_DICOM.ipynb notebook gives an overview of the conversion steps and how to reproduce the results.

Conversion tools used

  • STL To labelmap : Slicer3D
  • labelmap to SEG DICOM : dcmqi 1.2.4
  • STL to DICOM STL : dicom3tools

Collection information : Prostate-MRI-US-Biopsy

Please see TCIA website and Imaging Data Commons for additional details/visualization of the T2W MRI images considered for annotations conversion.

Where to download the DICOM SEG/STL objects

A zenodo publication will follow shortly, where 1017 prostate annotations and 1311 lesions were converted from STL to DICOM SEG/STL and publicly available. A final step will be the publication of these annotations in the Imaging Data Commons platform.

About

Examples for converting prostate MRI-US-Biopsy STL prostate and lesion surfaces to DICOM SEG and DICOM STL.


Languages

Language:Jupyter Notebook 99.5%Language:Python 0.5%