ESTRO 2024 - Abstract Book

S4577

Physics - Machine learning models and clinical applications

ESTRO 2024

descending order for V20, MLD, and V5 sequentially (Fig. 1(c)). These results demonstrated that the clinically adopted predictors V20 and MLD were proved to be correlated with RP but independent of the lung volume definitions. Multivariate logistic regression analysis showed similar predictive performance among three methods (Fig. 2(a) for the ROC curves) with the AUC values of ~0.58. PCA analysis based on 70 bins of DVH data yielded the first two principal components with variance ratios of 67% and 22% respectively for all three methods. Logistic regression with the first five principal components from PCA led to similar ROC curves (Fig. 2 (b)) and AUC values: 0.597 (TL), 0.599 (TL-CTV), and 0.603 (TL-PTV), indicating insignificant differences among three lung volume definitions. The AUC values additionally confirmed that DVH-based predictors might not be good discriminators in terms of data classification.

Conclusion:

The common dosimetric variables V20 and MLD used in clinical practice have been found to be statistically significant in data classification through both the Mann-Whitney U test and the univariate logistic regression analysis. Multivariate logistic regression analysis using both the clinical dosimetric variables and the PCA selected predictors out of the 70 binned DVH data yielded statistically insignificant difference among the target volume margin choices. The DVH-based predictors from logistic regression classification revealed poor radiation pneumonitis prediction performance. The findings from this study slightly differ from prior results [3] and suggest further research to identify more discriminative predictors, such as those incorporating spatial information of dose variation within the lungs.

Keywords: radiation pneumonitis, lung cancer

References:

1. Bradley, J.D., et al., Standard-dose versus high-dose conformal radiotherapy with concurrent and consolidation carboplatin plus paclitaxel with or without cetuximab for patients with stage IIIA or IIIB non-small-cell lung cancer (RTOG 0617): a randomised, two-by-two factorial phase 3 study. Lancet Oncol, 2015. 16(2): p. 187-99. 2. Movsas, B., et al., Quality of Life Analysis of a Radiation Dose-Escalation Study of Patients With Non-Small-Cell Lung Cancer: A Secondary Analysis of the Radiation Therapy Oncology Group 0617 Randomized Clinical Trial. JAMA Oncol, 2016. 2(3): p. 359-67.

3. Meng, Y., et al., Excluding PTV from lung volume may better predict radiation pneumonitis for intensity modulated radiation therapy in lung cancer patients. Radiat Oncol, 2019. 14(1): p. 7.

3256

Digital Poster

Surface Mapping with Machine Learning for Resimulation Decision Support in Prostate Proton Therapy

Stephen Shang 1 , Grant Evans 1 , Maxwell Kassel 1 , Matthew Kligerman 1 , Michael Shang 2 , Timothy R Williams 3 , Charles Shang 1