Radiomics tools

Note: Here is a full system for lung cancer screening radiomics. https://github.com/taznux/LungCancerScreeningRadiomics

Image processing tools and ruffus based pipeline for radiomics feature analysis

Super build

Just run super-build.sh

./super-build.sh

Install software

Python 3.7

• Required
• Script engine and useful modules
  • http://conda.pydata.org/miniconda.html - !recommend!
  • https://www.python.org/downloads/
• Required modules: pandas, ruffus, SimpleITK, scipy, numpy, ipython, matplotlib
• install_modules.sh or install_modules.cmd is available to install these modules using conda.

Slicer 4.10

• Recommended
• Image viewer, contour editor, simple image processing tool
  http://download.slicer.org/

Build

• gcc or visual studio
• cmake
• ITK 4.13.2

  ./build.sh

Tools

1. DICOMTools

1. DICOMTagReader - Display entire DICOM tags

   DICOMTagReader [DICOM directory]

2. DICOM2NRRDConverter - DICOM to nrrd (Slicer file format)
   Simple recursive converting for single patient data

   DICOM2NRRDConverter [DICOM directory] [nrrd directory]

   For large data

   python DICOM2NRRDConverter.py [DICOM directory] [nrrd directory]

3. DICOM-RT2NRRDConverter - DICOM-RT to nrrd

2. ContourTools

1. STAPLEComparison - variation comparison on multiple contours
2. ExtractBoundary
3. GTVs2ITV
4. HoleGenerator
5. ROIGenerator
6. ROI2BinImage
7. ROICropImage

3. GrowCutSegmentation

NoduleSegmentation - Segment small nodular objects for solid nodule and GGO

NoduleSegmentation [InputImageFile] [SeedPoint_x] [SeedPoint_y] [SeedPoint_z] \
                   [NoduleSize_long] [NoduleSize_short] [OutputImageFile]

4. Feature Extraction

FeatureExtraction - Extract image features from the nodule segmentation

FeatureExtraction [InputImage] [LabelImage] [FeatureFile] [Label={1}]

5. Python Tools

1. metadata.py - for handling metadata in csv or xls
2. organize_features.py - for collecting feature data into a single csv file

6. MATLAB Tools

1. NRRD4Matlab - for handing nrrd format in MATLAB
2. PET2SUV - for converting raw PET image to standardized uptake value(SUV)

7. ETC

1. RegistrationSITK - simple registration code, required SimpleITK module for python
2. SlicerPythonExtensions - simple extensions for Slicer
   1. InterpolateROIsEffect.py
   2. LineProfile.py

6. LASSO-SVM

TBD - modeling code for radiomics features

Usage

Radiomics feature extraction pipeline example for LUNGx dataset

1. Download DICOM images
   https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge

   Download all DICOM images to ‘DATA’
   You can use the included metadata files for LUNGx (TrainingSet.csv and TestSet.csv)

2. Environmental parameters
   Set your parameters in script/run_lungx.py (recommend default setting).

   experiment_set = 'TrainingSet'  
   # experiment_set = 'TestSet'  
   output_path = 'output'  
   data_path = 'DATA'  
   dicom_path = data_path + '/DOI'  
   image_path = data_path + '/' + experiment_set  
   nodule_info_path = './' + experiment_set + '.csv'

3. Run radiomics pipeline

   $ python script/run_lungx.py or script/run_lungx.py

4. Analysis feature data
   output files (intermediate images and feature data) will be generated in ‘output’ directory

• TrainingSet: feature_list_TrainingSet.csv
• TestSet: feature_list_TestSet.csv

Wookjin Choi wchoi1022@gmail.com