The use of CT radiomics to predict immune infiltrate in pancreatic ductal adenocarcinoma

Background: Prognostic and/or predictive biomarkers for patients with pancreatic adenocarcinoma are limited. Radiomics is a non-invasive method to quantitatively analyze tumors with imaging. Previous data have demonstrated that CT radiomics is predictive of overall survival in PDAC. Based on evidence that enhanced intratumoral adaptive immunity is associated with improved survival in PDAC, we sought to build an immune infiltration prediction model with CT radiomics. Methods: A tissue microarray (TMA) was constructed with tumor arranged in triplicate from short-term and long-term survivors of PDAC who underwent resection between 2005 and 2010. Multiplexed immunohistochemistry (IHC) was performed for intratumoral mature dendritic cells (DC-LAMP+), regulatory T cells (CD3+FoxP3+), CD8+ T cells, (CD3+CD8+), activated cytolytic CD8+ T cells (CD3+CD8+GrB+), and macrophages (CD68+). The pancreatic tumor in the portal venous phase on preoperative CT was manually delineated, and 255 radiomic features were extracted and analyzed for significance using univariate linear regression. A multivariate linear regression analysis was then performed to correlate radiomic features with immune infiltrate for each of the immune phenotypes. Results: There were 22 patients included in the analysis, 7 of whom were short-term survivors (median survival of 9 months) and 15 were long-term survivors (median survival of 74 months). Feature selection resulted in 17 significant radiomic features for mature dendritic cells, 7 for regulatory T cells, 11 for CD8+ T cells, 16 for cytolytic CD8+ T cells and 12 features for macrophages. Five regression models were then constructed using the significant radiomic features to predict intratumoral immune infiltrate. There was a strong association between radiomic features and mature dendritic cells (R2=0.83, 95% CI 0.761-0.906, p Conclusion: A strong association was found between imaging phenotypes and both dendritic cells and activated cytolytic T cells, demonstrating that non-invasive imaging techniques may help identify tumors with active adaptive immunity. Further analysis is underway in a larger cohort of patients to validate these findings. Citation Format: Caitlin A. McIntyre, Jayasree Chakraborty, Maura Koszalka, Jared Bassmann, Yuting Chou, Mithat Gonen, Peter J. Allen, T. Peter Kingham, Michael I. D9Angelica, Jeffrey A. Drebin, William R. Jarnagin, Richard K. Do, Vinod P. Balachandran, Amber L. Simpson. The use of CT radiomics to predict immune infiltrate in pancreatic ductal adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2444.