Warning: Undefined variable $sessionDesc in D:\inetpub\aapmnfe\meetings\2019AM\programInfo\programAbs.php on line 485
(Sunday, 7/14/2019) 1:00 PM - 2:00 PM
Room: Stars at Night Ballroom 2-3
Purpose: To evaluate the prognostic power of radiomics (CT) and dosiomics (TPS Dose Image) features using segmentations of V₂₀ and V₅ peritumoral dose distributions to predict radiation pneumonitis (RP) in a large cohort of 701 locally advanced non-small cell lung cancer patients.
Methods: Planning CT and TPS thoracic dose data was used to subtract the tumor volume and generate dose-level (V₂₀ and V₅) based peritumoral volumes registered to CT pulmonary volumes. Only patients that completed a curative treatment course were included (median dose = 64 Gy). The endpoint was CTCAE grade>=2 RP (n=77). Radiomics and dosiomics features were computed for the ipsilateral V₂₀ and V₅ CT and dose volumes using the open-source PyRadiomics toolkit. Feature selection was performed for each data set using the Minimum Redundancy Maximum Relevance (mRMR) algorithm and classification performance was evaluated for top 15 features using the AUC statistic. For comparison, conventional DVH parameters including lung V₂₀ and V₅ mean lung dose (MLD), and pack years were incorporated in the univariate analysis.
Results: We find that the performance of radiomics and dosiomics V₂₀ & V₅ peritumoral features for the ipsilateral lung is comparable to MLD in predicting RP. MLD was statistically significant at predicting RP with AUC = 0.674. The AUC statistic for the statistically significant (p < 0.05) V₂₀ radiomics predictors ranged from 0.589 to 0.644 and dosiomics predictors ranged from 0.582 to 0.628. The AUC statistic for statistically significant (p < 0.05) V₅ radiomics predictors ranged from 0.593 to 0.647. Only one of the V₅ dosiomic features was statistically significant with AUC = 0.611.
Conclusion: Univariately, radiomics and dosiomics features from the peritumoral regions are comparably predictive to clinical predictors i.e MLD. A multivariate model combining both radiomics & dosiomics predictors to further enhance the predictive power for RP must be explored.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
