An Evaluation and Feasibility Study for the Need of New Dosimetric Tools and Metrics for Lung Cancer Patients Receiving Radiotherapy

Abstract

Evaluation of potential toxicity to healthy lung tissue in radiotherapy of lung cancer is assessed via the percentage of the healthy lung volume receiving doses between 5 to 30 Gy. Respiration during treatment changes lung density, affecting the radiological path length. This results in differences in the absorbed dose and variations in volume metrics taken from the static DVH, making the V5, for example, dynamic. This retrospective analysis of 10 lung cancer patients demonstrates the percent change (%Δ) of V5, V10 and V20 within the same patient when the clinical dose calculated on the average-intensity-projection CT (avgCT) is recomputed on 10 4DCT phases. We also compute and propose alternative mass-based metrics generated from the same CT data. We assess if they are demonstrably more stable and statically significant in their differences from the volumetric data when compared to the avgCT. Lung contours were redrawn on each patient’s avgCT using auto segmentation and then propagated to all the breathing phases (0%-90%) using deformable registration. The mean Hounsfield units (HU) and the volume of the lungs was extracted. The data sets were pushed to the treatment planning system were the clinically administered plan was recomputed on each phase and the V5, V10 and V20 were tabulated for all 110 CT sets. The 5, 10 and 20 Gy isodose lines (IDL) on each CT were converted to contours and their volumes acquired. Using the HU as a surrogate for density, the mass of each patient’s lung, along with the mass of contours of 5, 10 and 20 Gy IDLs were computed on all breathing phases and compared to the avgCT, yielding the mass of lung receiving 5, 10 and 20Gy (M5, M10, M20). For consistency, data from avgCT was the benchmark for %Δ of the all the mass and volume metrics. We observed a mean standard deviation (SD) of lung mass %Δ of 0.94± 0.4%, versus a mean SD of lung volume %Δ of 4.32 ± 0.95%. Comparisons of patients’ M values and V values to the avgCT values yielded mean SD of %Δ of V5 vs. M5 as 2.64 ± 1.03% vs. 1.99 ± 0.48% respectively, V10 vs. M10 as 2.61 ± 0.94% vs. 1.75 ± 0.49%, and V20 vs. M20 as 2.81 ± 1.38% vs. 2.16 ± 0.67%, respectively. An Analysis of Variance (ANOVA) between the %Δ of mass data and the reciprocal volumetric data, all yielded p-values of < 0.02.One-sample, two tail t-tests of the %Δ in mass metrics from a hypothesized 0% variance from the avgCT also yield p-values of <0.01, showing that there is a low probability that the change observed from 0% variation is due to chance. The low p-values of the ANOVA test show that although the mass metrics are derived from the volume data, the %Δ between corresponding data points is independent of each other. The lower SD and low p-values of the %Δ in mass metrics in all the breathing phases may point to the mass of lung irradiated to be more reliable way to assess dose to healthy lung. This would allow for larger retrospective study where the association could be determined between radiation induced pneumonitis and mass of lung irradiated by specific doses.