Code to automatically assemble phantoms from DICOM CT scans and RT STRUCT files using machine learning models.
If you find this code helpful, please consider citing our work:
@article{virgolin2020machine,
title={Machine learning for the prediction of pseudorealistic pediatric abdominal phantoms for radiation dose reconstruction},
author={Virgolin, Marco and Wang, Ziyuan and Alderliesten, Tanja and Bosman, Peter AN},
journal={Journal of Medical Imaging},
volume={7},
number={4},
pages={046501},
year={2020},
publisher={SPIE}
}
This code is an anonymized version of the one actually used to generate phantoms for [1]. In the lookup table, information on predictions of machine learning algorithms is pre-collected to assemble the phantoms. For a practical use, lookup tables must be replaced by actual predictions of pre-trained models, so models need to be integrated and called-upon when needed, to generate predictions on the fly given the features of the input patient.