Characterization of Radiographic Changes Following Lung Stereotactic Body Radiotherapy – A Dose-Based Radiomic Analysis

Abstract

<h3>Purpose/Objective(s)</h3> Evaluation of diagnostic imaging following lung stereotactic body radiotherapy (SBRT) can be challenging. We hypothesize that radiation dose maps can be used to characterize post-SBRT changes utilizing radiomics. <h3>Materials/Methods</h3> We retrospectively identified patients with early stage (T1-T2, N0) non-small cell lung cancer treated at a single center with SBRT to a dose of 48-55 Gy in 3-5 fractions between 2017 and 2020. Pre-SBRT and the 1-year post-SBRT diagnostic CT scans with IV contrast and 1mm slice thickness were imported into MIM (MIM Software Inc. Cleveland OH). A semi-automated workflow consisted of biologically effective dose (BED) maps generation, rigid fusions of planning CT with each diagnostic scan prioritizing alignment of the gross tumor volume, contouring the lung, and lung volumes of interest (VOI) generation corresponding to high-dose (BED10 >= 100 Gy), intermediate-dose (BED10>=24.48 Gy), and low-dose (BED10<24.48 Gy). Using open-source software and a bin width of 25 Hounsfield units, dose-based VOI masks were used for radiomic analysis of pre- and post-SBRT diagnostic scans. Changes in radiomic features were evaluated with paired 2-sample t-tests and p<0.05 indicating significance Comparing pre- and post-SBRT scans, feature differences were considered treatment related if a significant difference was detected in the intermediate or high dose VOI but not the low dose VOI. <h3>Results</h3> Among 28 patients, 86% were stage T1a-T1c, with the most common dose of 50 Gy in 5 fractions. Diagnostic CT scans were an average of 3.6 months before SBRT and follow up scans were an average of 13.1 months after completion of SBRT. Of 106 radiomic features extracted from each image set, 8 were significant in both intermediate and high dose VOI and 5 were significant in the high dose VOI only (see Table). <h3>Conclusion</h3> We demonstrate the feasibility of a semi-automated workflow for dose related radiomic analysis of diagnostic CT scans in patients treated with SBRT and identify possible radiomic features pertinent to treatment related changes. The workflow can be leveraged to investigate predictive capabilities and assist interpretation of post-SBRT radiologic changes.