Preprocessing scripts to convert Dicoms to PNGs and extract relevant metadata.

This repo contains data preprocessing scripts for our Mammography projects. For details on model development, see OncoNet. To use the models, see OncoServe.

Required python pip packages are listed in requirements.txt. All required pip packages and command line tools can be installed by running ./requirements.sh.

DICOMs can be converted to PNGs using the script dicom_to_png.py located in the scripts/dicom_to_png folder. Conversion can use either the dcmj2pnm tool from the dcmtk package or the Matlab dicomread tool.

DICOM header metadata can be extracted and saved either as a JSON file or to a SQL table. Both scripts are located in the scripts/dicom_metadata folder. To save as a JSON file, use dicom_metadata_to_json.py. To save to a SQL table, use dicom_metadata_to_sql.py. To examine dicom metadata in the SQL table, use dicom_metadata_from_sql.py and replace the query with your own query. DICOM metadata in JSON format can be summarized and plotted using plot.py and summarize.py.

A directory can be copied in parallel using copy_dir_parallel.py in the scripts/utils folder.