MO-AB-BRA-01: A Novel Method to Objectively Quantify Normal Tissue Toxicity in the Esophagus




Inconsistency between NTCP models for esophagitis prediction leads to reduced treatment efficacy, and arises from using subjective endpoints. We investigate esophageal swelling as an objective endpoint for NTCP modeling.


A novel voxel-based method was developed to quantify expansion in the esophagus. Utilizing deformable image registration (DIR) between a plan and serial CT image set, this technique calculates esophageal volume change from the Jacobian Map. This voxel-volume change quantifies esophageal swelling. Next, a technique was developed to reduce uncertainty in esophageal volume change calculations due to physiological variation (i.e. swallowing, esophageal air pockets, etc.). The technique's effectiveness was evaluated by comparing expansion calculations on week01 CTs with and without corrections for 65 patients. This methodology was tested on sixty-five patients with NSCLC that had esophagitis scoring (CTCAE3.0, 15 Grade3, 31 Grade2, 19 Grade0) and 6–8 weekly 4DCTs during treatment (IMRT). Esophageal swelling was analyzed for the treatment week of highest response. Dose was accumulated weekly using DIR. Logistic regression models were used to examine the relationship between expansion and grade3 esophagitis, and used to determine optimal binary endpoints of expansion. The relationship between dose and esophageal expansion as the endpoint was compared with dose to grade NTCP models. Performance of models was quantified with area under the curve (AUC) from ROC analysis.


Uncertainty in esophageal expansion calculations were reduced from 14.3% to 6.9%. The following expansion metrics were the most significantly correlated to grade3 esophagitis: maximum expansion (AUC=0.87,pval<0.001) and length of esophagus expanding>20% (LExp20%) (AUC=0.88,pval<0.01). The optimal expansion endpoint metrics were: Maximum Expansion>1.4 and LExp20%>50mm. Multivariate NTCP model with expansion endpoints outperformed NTCP model with grade endpoints: dose-to-expansion endpoint AUC=0.91, dose-to-grade endpoint AUC=0.81.


Esophageal swelling is a continuous, objective, and in-vivo metric of esophageal toxicity. Esophageal swelling endpoints can increase the performance of NTCP models.