(Tuesday, 7/16/2019) 1:15 PM - 1:45 PM
Room: Exhibit Hall | Forum 2
Purpose: Radiomics features extracted with different regions of interest (ROIs) may perform differently for predicting radiation pneumonitis. We investigated the potential of radiomics extracted from the planning CT images with five different ROIs in predicting radiation pneumonitis.
Methods: We studied 31 patients that accepted fractionated radiotherapy to the thorax, and divided them into two groups: 16 with and 15 without radiation pneumonitis. The planning CT image datasets with five ROIs were exported for analysis: Gross Tumor Volume (GTV), Planning Target Volume (PTV), PTV-GTV, Total Lung-GTV and Total Lung-PTV. A total of 1085 radiomics features were extracted from the CT image datasets with each ROI. We calculated the differences between two groups of patients using Wilcoxon rank sum test, where P value less than 0.05 was statistically significant. The correlation between each pair of the significant features was quantified by the spearman correlation coefficient (rs). If the rs was > 0.85, the one with smaller P value was selected.
Results: Radiomics features extracted with five ROIs can all predict the radiation pneumonitis, with the predictive ability ranging from the 0.001 to 0.010. Among five ROIs, PTV had 16 significant radiomics features (P: 0.002-0.010), exceeding the others. In GTV, only 4 features are statistically significant (P: 0.005-0.009). In PTV-GTV, there are 8 features that are significantly different (P: 0.004-0.010). A total of 14 features in Total Lung-GTV and 10 features in Total Lung-PTV were associated with radiation pneumonitis (P: 0.001-0.010; 0.001-0.002).
Conclusion: This study found that radiomics features extracted with five different ROIs were all able to predict radiation pneumonitis. Although there are no many differences in the predictive ability, the definition of the ROIs can influence the selection of significant features in radiomics study. This study needs to be further investigated.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by National Key Research and Development Program of China (2016YFC0103400), Key Research and Development Program of Shandong Province (2017GSF218075), Jianfeng Q. was supported by the Taishan Scholars Program of Shandong Province (TS201712065).
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
