CT characteristics allow identification of patient-specific susceptibility for radiation-induced lung damage

Background and purpose: There is a huge difference in radiosensitivity of lungs between patients. The present study aims to identify and quantify patient-specific radiosensitivity based on a single pre-treatment CT scan. Materials and methods: 130 lung cancer patients were studied: 60 stereotactic ablative 'radiotherapy (SABR) treatments and 70 conventional treatments (20 and 30 patients from external datasets, respectively). A 3 month-follow-up scan (CT3M) was non-rigidly registered to the planning CT scan (CT0). Changes in Hounsfield Units (Delta HU = HU3M - HU0) inside lung subvolumes were analyzed per dose bin of 5 Gy. Delta HU was modeled as a function of local dose using linear and sigmoidal fits. Sigmoidal fit parameters Delta HUmax (saturation level) and D-50 (dose corresponding to 50% of Delta HUmax) were collected for all patients. Results: Sigmoidal fits outperformed linear fits in the SABR groups for the majority of patients. Sigmoidal dose-responses were also observed in both conventional groups but to a lesser extent. Distributions of D-50 and Delta HUmax showed a large variation between patients in all datasets. Higher baseline lung density (p <0.001) was prognostic for higher Delta HUmax in one SABR group. No prognostic factors were found for D-50. Conclusions: Baseline CT characteristics are prognostic for radiation-induced lung damage susceptibility.