pwoznicki/radiomics-benchmark

from the publication in Frontiers in Radiology

Code for the analysis used in the paper "AutoRadiomics: A Framework for Reproducible Radiomics Research" Woznicki et al. 2022

Requirements

git clone https://github.com/pwoznicki/AutoRadiomics
cd AutoRadiomics
git checkout 8d35988cd475af84573f11993542724912485825 # commit from 25.04.2022
pip install -e .

Run experiments

Select your base directory for all experiments and update BASE_DIR in worc/config.py.

WORC datasets

1. Download data for all six experiments from https://xnat.bmia.nl/data/projects/worc and put all of them in <your_base_dir>/worc/data/.

2. Download table with labels and clinical data from the XNAT repository (Scroll down to Subjects -> Options -> Spreadsheet). Save it as <your_base_dir>/worc/tables/clinical.csv

3. Run the scripts:

cd worc
python preprocess.py
python feature_extraction.py
python training.py

The results will be saved in <your_base_dir>/worc/results/.

Prostate datasets (Prostate-UCLA and ProstateX)

Data Download

1. Prostate-UCLA:

• Download DICOM images from https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=68550661 (takes a while...) \ Save them in a folder <your_base_dir>/prostate/data/prostate-ucla/dicom/.

• Using the same link, download STL Files (ZIP). Unzip and put all the .STL files in the folder <your_base_dir>/prostate/data/prostate-ucla/masks/STL/.

• From the same link, download Biopsy data (Excel file). Save it as <your_base_dir>/prostate/tables/prostate-ucla/biopsy.xlsx.

2. PROSTATEx:

git clone https://github.com/rcuocolo/PROSTATEx_masks
cd PROSTATEx_masks
git checkout 43b55e454410d78831fd184d8010f23af91e5144
cp -r Files/lesions/* <your_base_dir>/prostate/data/prostatex/lesion
cp -r Files/prostate <your_base_dir>/prostate/data/prostatex/
cp -r Files/lesions/*.csv <your_base_dir>/prostate/tables/prostatex/

Preprocessing:

1. Prostate-UCLA requires conversion to Nifti:

• Convert dicom images to nifti (will be saved in <your_base_dir>/prostate/data/prostate-ucla/nifti/)):

cd <your_base_dir>/prostate/data/prostate-ucla/dicom/
dcm2niix -z y -f %i/%j -z y -o nifti dicom

• Convert segmentations from STL into nifti. For that, run a docker container with 3D Slicer Notebook environment from this directory. In there, run the notebook work/preprocessing/slicer_stl_to_nifti.ipynb (set the paths accordingly) with Slicer kernel.

cd prostate
docker run -p 8888:8888 -p 49053:49053 -v <your_base_dir>/prostate/data/prostate-ucla/:/data -v "$PWD":/home/sliceruser/work --rm -ti lassoan/slicer-notebook:latest

python create_path_df_ucla.py

2. PROSTATEx:

python create_path_df_prostatex.py

Feature extraction:

python feature_extraction.py

Training:

python training.py

radiomics-benchmark's Issues

TypeError: 'type' object is not subscriptable

Hi,
I got an error when running the script. Also to mention, I think the WORC package might got updated. Some samples did not fit into the configuration ( "GIST-018", "Lipo-073") I first added them to the EXCLUDED_SUBJECTS in config.py

This is the error I get:
from autorad.utils import extraction_utils, io, splitting
File "/worc/autorad/utils/splitting.py", line20, in
) -> dict[str, dict[str, Any]]:
TypeError: 'type' object is not subscriptable