Implementation of plan feature extraction in the paper "Predicting gamma passing rates for portal dosimetry based IMRT QA using machine learning", Medical Physics, 2019. Following features are calculated:
- BA Beam aperture area weighted by MU
- BI Beam irregularity
- BM Fraction of BA normalized by UAA
- UAA Union area of aperture (UAA)
- MFAS2,5,10,20 Mean of fraction of aperture smaller (MFAS) than 2, 5, 10, 20 mm
- MaxFAS2,5,10,20 Max of fraction of aperture smaller (MaxFAS) than 2, 5, 10, 20 mm
- MAA Mean aperture area
- MAD Maximum distance of the mid‐point between any open leaf‐pair in a beam
- MUCP Mean of MUs per control point in a beam
- MLO1,2,3,4,5 Moment order of 1, 2, 3, 4, 5 of leaf openings
- minAP_h Minimum aperture perimeter in horizontal direction
- maxAP_h Maximum aperture perimeter in horizontal direction
- minAP_v Minimum aperture perimeter in vertical direction
- maxAP_v Maximum aperture perimeter in vertical direction
- maxRegs Maximum number of regions in the beam
- AAJA Ratio of the average area of an aperture over the area defined by jaws
- MAXJ Maximum of x‐y jaw positions
- MCS Modulation complexity score
- EM Edge metric: ratio of MLC side‐length to aperture area
python plan_complexity -i RP.dcm -o output.csv
The program takes a plan as input and outputs a csv file in which each line is extracted features for each beam in the plan
I recommend Anaconda as Python package manager. It comes with numpy. The program needs pydicom package which can be installed by pip: pip install pydicom. Please use Python3 as Python2 is no longer supported.
Please cite our paper if you find it useful for your research.
@article{lam2019predicting,
title={Predicting gamma passing rates for portal dosimetry based IMRT QA using machine learning},
author={Lam, Dao and Zhang, Xizhe and Li, Harold and Yang, Deshan and Schott, Brayden and Zhao, Tianyu and Zhang, Weixiong and Mutic, Sasa and Sun, Baozhou},
journal={Medical physics},
year={2019},
publisher={Wiley Online Library}
}