A digital pathology platform to quantify the spatial heterogeneity and patterns of immune markers in pathology specimens from triple-negative breast cancer

Overwhelming evidence has shown the significant role of the tumor microenvironment (TME) in governing the triple-negative breast cancer (TNBC) progression. Digital pathology can provide key information about the spatial heterogeneity within the TME using image analysis and spatial statistics. These analyses have been applied to CD8+ T cells, but quantitative analyses of other important markers and their correlations are limited. In this study, we formulate a workflow for characterizing the spatial distributions of five immune markers (CD3, CD4, CD8, CD20, and FoxP3) in specimens from patients with TNBC. The workflow is initiated by image processing to extract and colocalize immune marker-labeled cells and then convert this information to point patterns. A pixel-based algorithm is developed to infer invasive front (IF), central tumor (CT), and normal tissue (N). For each region, we examine the intra-tumoral heterogeneity. Our workflow is then repeated for all specimens to capture inter-tumoral heterogeneity. In this study, both intra- and inter-tumoral heterogeneities were observed for all five markers across all specimens. Among all regions, IF tends to have higher densities of immune cells and overall larger variations in spatial model fitting parameters and higher density in cell clusters and correlation hotspots compared to CT and N. In conclusion, a comprehensive workflow was developed to characterize the spatial heterogeneity and patterns of different immune markers. These results suggest a distinct role of IF in the tumor immuno-architecture; we quantitatively characterize both intra- and inter-tumoral spatial distributions and their variations. Our workflow is robust, as the results are in agreement with the previously published findings on the same specimens. In future studies, these analyses could be linked to treatment outcomes and computational immuno-oncology models to guide patient stratification and identify predictive biomarkers for responder/non-responder characterizations