The five calculations described in the Material and Methods Section C above were exported in DICOM format. The dose and volume data for PTV and organs at risk (OAR) listed in Table
2 were extracted from the DICOM files for the five plans, respectively, using the in‐house MATLAB (The MathWorks, Natick, MA) code and the computational environment for radiotherapy research (CERR)
(20) software.
Based on parameters listed in Table
2, the following metrics were derived to evaluate the differences and relations among the five calculations:
R(Dx)2,1: Ratio of Dx of PTV between two calculations defined as Dxplan2Dxplan1, where Dx represents the dose to x percent of the PTV and x was 1%, 95%, or 99%. We compared the quantity [R(Dx)2,1−1] which represented the percentage difference between two calculations;
Percentage of cases in which the difference in D95 of PTV between two calculations was more than 7%. The 7% dose difference was chosen since it might be detectable from clinical outcomes;(21) and
Correlation of mean lung dose (MLD) and V20 of lungs between two calculations.
Comparisons between MCHete
* and PBHomo were of great interest since the calculations with MC and heterogeneity correction in MCHete
* gave the actual planned doses, which may have been different from the dose given by PB calculations without heterogeneity correction.
To study the tumor location dependence of the dose differences between MC and PB calculations, all lesions were separately grouped based on the distance between the GTV contours to the chest wall. A lesion was considered as an edge case if the distance was smaller than 1 cm; otherwise, it was considered as an island lesion, as shown in Fig.
2. The comparisons were conducted separately for the two types of lesions.
†