Computed Tomography-Based Delta Radiomics of Tumor Core-Edge Combination for Systemic Treatment Response Evaluation in Pancreatic Cancer

Abstract

Introduction: Systemic treatment is routinely used in pancreatic cancer (PC) patients who are ineligible for surgery. A subgroup of patients can likely undergo surgery after downstaging. Hence, it is important to provide surgeons with effective approaches to evaluate the response during the treatment. Methods: A total of 71 PC patients were enrolled. According to the pathologic response, patients were classified into the responder and non-responder groups. Based on the results of admission and post-treatment CT, the regions of interest (ROI) were specified. The tumor ROI was divided into the core and edge areas, and the delta radiomics features of the whole tumor, core, edge, and peritumoral area were extracted. Finally, we compared the accuracy of models based on radiomics features, changes of tumor markers, and radiologic response. Results: The numbers of selected features from the whole tumor, core, edge and peritumoral area were 4, 8, 6, and 14 respectively. The model based on core (AUC=0.841) and edge features (AUC=0.860) showed much better performance than that based on whole tumor (AUC=0.828) or peritumoral area (AUC=0.814). Moreover, the tumor core-edge combination model (AUC=0.873) could increase confidence in treatment response as compared to using either of them alone. The accuracy of models based on changes of tumor markers and radiologic response was relatively poorer than the radiomics model. Conclusions: The tumor core-edge combination delta radiomics model is an effective approach to evaluate pathologic response in PC patients with systemic treatment.