349 - Predicting Radiation Pneumonitis with Robust 4DCT-Ventilation and 4DCT-Perfusion Imaging Using Prospective Lung Cancer Clinical Trial Data

T. Neupane¹, E. Castillo², Y. Chen³, S. H. Pahlavian⁴, R. Castillo⁵, Y. Vinogradskiy³, and W. Choi⁶; ¹Thomas Jefferson University, Philadelphia, PA, ²Department of Biomedical Engineering, The University of Texas at Austin, Austin, TX, ³Department of Radiation Oncology, Sidney Kimmel Medical College at Thomas Jefferson University, Philadelphia, PA, ⁴MIM Software Inc., Beachwood, OH, ⁵Department of Radiation Oncology, Winship Cancer Institute of Emory University, Atlanta, GA, ⁶Memorial Sloan Kettering Cancer Center, New York, NY

Purpose/Objective(s): Methods have been developed that apply image processing to 4DCTs to generate lung ventilation maps (4DCT-ventilation). 4DCT-ventilation provides functional information without burdening the patient with an extra imaging procedure. Traditional 4DCT-ventilation methods that rely on Hounsfield-Unit (HU) density-change lack reproducibility and do not provide 4DCT-perfusion data. Novel 4DCT-ventilation/perfusion methods have been developed that are robust and provide the perfusion information. The purpose of this study was to use prospective clinical trial data to evaluate the most effective 4DCT-based lung function imaging method for predicting radiation pneumonitis (RP). Materials/

Methods: 63 advanced-stage lung cancer patients enrolled in a multi-institutional, phase 2 clinical trial on 4DCT-based functional avoidance radiation therapy were used. 4DCT- were used to generate 4 lung function images: 1) 4DCT-ventilation using the traditional HU (‘4DCT-vent-HU’) approach, and 3 methods using the novel statistically robust methods: 2) 4DCT-ventilation based on the Mass Conserving Volume Change (‘4DCT-vent-MCVC’), 3) 4DCT-ventilation using the Integrated Jacobian Formulation (‘4DCT-vent-IJF’) and 4) 4DCT-perfusion. For each 4DCT-ventilation/perfusion image, dose-function metrics were calculated including mean functional lung dose (fMLD), and percentage of functional lung volume receiving = 5Gy (fV5), and = 20Gy (fV20). The ability of standard lung dose metrics (MLD and V20) and dose-function metrics derived from the 4DCT-ventilation/perfusion imaging modalities to predict for = grade 2 RP was assessed using univariate logistic regression and machine learning. Model performance was evaluated using the area under the curve (AUC) and validated through 10-fold cross-validation (CV).
Results: 10/63 (15.9%) patients developed grade =2 RP. Logistic regression analysis yielded mean AUCs of 0.70±0.02 (p=0.04), 0.64±0.04 (p=0.13), 0.60±0.03 (p=0.27), 0.63±0.03 (p=0.20), and 0.64±0.08 (p=0.38) for 4DCT-vent-MCVC, 4DCT-perfusion, 4DCT-vent-IJF, 4DCT-vent-HU, and standard lung metrics, respectively. Machine learning modeling with 10-fold CV resulted in AUC values of 0.86±0.07, 0.83±0.08, 0.78±0.07, 0.75±0.08, and 0.78±0.08 for the 4DCT-vent-MCVC, 4DCT-perfusion, 4DCT-vent-IJF, 4DCT-vent-HU, and standard lung metrics, respectively. fMLD and fV20 exhibited the highest feature overlap in predicting RP.
Conclusion: This is the first study to comprehensively evaluate 4DCT-perfusion and robust 4DCT-ventilation in predicting RP outcomes. Our data identified 4DCT-ventilation based on the MCVC formulation along with fMLD and fV20 as the best predictors of RP; outperforming both standard lung metrics and traditional 4DCT-ventilation formulations. These findings will inform the design of phase III clinical trials and the optimal incorporation of 4DCT-ventilation/perfusion into thoracic radiotherapy clinical practice for reduced toxicity.