Investigating the dose dependence of median pixel value in CT lung images of patients undergoing stereotactic body radiation therapy

We investigated the relationship between the local dose delivered and median pixel-value change following radiation therapy (RT) by comparing anatomically matched regions of interest (ROIs) in pre- and post-RT computed tomography (CT) scans. Six patients' clinical pre-treatment baseline CT scans, follow-up CT scans, treatment planning CT scans, and dose maps were collected. The lungs were extracted using an automated segmentation algorithm, and demons deformable registration was used to register each patient's follow-up scan and treatment planning scan to their baseline scan. Median pixel values were calculated in anatomically matched ROIs in the baseline and deformed followup CT scans, and mean dose delivered to the same ROIs was determined from the deformed dose map. Pearson's correlation coefficients, rank correlation coefficients, and linear modeling were utilized to quantify the relationship between median pixel-value change and mean dose delivered. Pearson's correlation coefficients for the six patients ranged from -0.13-0.67. Rank correlation coefficients ranged from -0.12-0.80. Linear regression analysis on the six patients' combined data yielded a slope of 2.62 (p < 0.001) and R-squared value of 0.24. General positive trends were observed between radiation dose and median pixel value change, but no two patients had the same relationship between these variables, indicating it may not be possible to generalize patients' reactions to varying dose levels of radiation. Thus, an individualized method for evaluating normal lung tissue damage based on changes in each patient's CT scan following radiation treatment may be required to assess radiation-induced lung damage.