Regional variability in radiation-induced lung damage can be predicted by baseline CT numbers.

Abstract

Lung volumes are functionally heterogeneous but typically considered uniformly during radiotherapy planning. The present study aims to predict regional differences in radiation-induced lung damage based on pre-treatment CT information. For 42 lung cancer patients (including 15 from an external validation set), two 200cc lung subvolumes (low-density (LD) and high-density (HD)) were auto-segmented in the ipsilateral lung of the planning CT_0. After non-rigid registration of 3month follow-up CT scans, sigmoidal dose-density change (ΔHU=HU_3M-HU₀) response curves were determined for all subvolumes. Predictive factors for the sigmoidal response parameters D₅₀ and saturation level ΔHU_max were analyzed. The baseline density difference between LD (mostly in the upper lobe) and HD (mostly in the lower lobe) was on average 102HU. The saturation level ΔHU_max,LD was significantly smaller than ΔHU_max,HD (p=0.03). Expressed as mass density increase relative to the baseline density, saturation levels were 20.7% on average irrespective of baseline density, and they could be predicted in LD and HD subvolumes (AUC=0.70-0.78). Intra-lung differences in D₅₀ were significantly smaller than inter-patient differences. Limited amount of damage was observed in LD subvolumes, while the relative density increase of all subvolumes was well predictable. This could allow dose redistribution preferentially targeting low-density lung regions.