Room: Exhibit Hall | Forum 2
Purpose: Due to very large fraction dose, small tumor-margin the ability to access treatment response during the course of lung SBRT is highly desirable for adaptive radiotherapy. Daily CBCT imaging offers an opportunity of early treatment adaptation using the changes of imaging featuresâ€“delta radiomics. Radiation-induced changes in images features as a function of SBRT fraction dose were quantified via daily CBCT.
Methods: Daily CBCT images co-registered with the planning CT images for image-guided SBRT treatment of 12 early stage-I NSCLC patients who received 50Gy in 5 fractions every other day were analyzed using the open-source IBEX software. Five radiomic features (Sum Variance, Sum Average, Sum Entropy, Auto Correlation, Homogeneity) were selected based on linear correlation strength for two regions of interest (ROI): planning target volume (PTV) and normal lung V20 for each fraction for all patients. CBCT images prior to the 1st SBRT fraction dose were the baseline scan. CBCT acquisitions parameters were the same for all scans.
Results: ROIs typically produced values that have a linear correlation value of >Â±0.5, notable exceptions being the PTV of patients 7 and 9. The features that exhibit the largest directional influence were homogeneity and sum entropy, as the variance between directions was greatest. The remaining features were largely unaffected by directionality. Additionally, the sign of correlation for both structures typically remained the same across most patients, exceptions being patients 5, 8, and 11. Radiomic features analysis of CBCT images for 5 fractions SBRT provided feature values that have a strong linear response to treatment fractions. For features that exhibit most stability, this linear response can be potentially used to predict changes in regional-texture and voxel-intensity.
Conclusion: These quantitative features obtained from daily CBCT images could provide information to contribute to early SBRT treatment adaptations. Future research involves analyzing gross tumor volume features.