CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Quantification of spatial interaction between cancer and immune cells Authors:  Inna Chervoneva - Thomas Jefferson University (United States) [presenting]
Abstract: Advanced analysis systems for pathology allow capturing spatial coordinates of all cells in immunohistochemistry images of the tumor microenvironment, but there are no established methods for objective quantification of spatial interaction between cancer and immune cells. The motivation comes from studies of cancer biomarkers in a tissue microarray of surgical specimens from a large cohort of ER+ breast cancer patients. We developed an objective thresholding algorithm for immune cell type classification and novel metrics of spatial interactions between cancer and immune cells based on distributions of the nearest neighbor distances. The spatial localization of CD8+ T cells and cancer cells was used to generate distributions of the nearest neighbor distances (NND) between cancer and CD8+ cells. The summary statistics of these distributions were considered as predictors of progression-free survival (PFS). The larger quantiles of NND were associated with shorter PFS, which is consistent with published results based on manual counts of tumor-infiltrating CD8+ lymphocytes. The median NND from cancer to CD8+ cells remained a significant predictor of PFS in the multivariable Cox model adjusted for known clinicopathological prognostic factors. The distributions of NND between cancer and various immune cells have the potential to provide novel cancer biomarkers.