Abstract 855: Spatial analytics of the tumor microenvironment on double stained immunohistochemistry images using deep learning

Abstract

Abstract Spatial locations of immune cells in the tumor microenvironment (TME) have been shown to correlate with clinical outcome in different cancers. A worse patient outcome has been reported in oral squamous cancer for individuals with an increased number of Tregs within 30 µm of CD8+ cells [1]. Likewise, the spatial relationship between CD8+ and PD-L1+ cells has become an area of interest as a possible indication for the response to PD-L1-inhibition. It was recently shown in a retrospective study that the combined assessment of CD8 and PD-L1 in NSCLC tumors outperformed CD8 or PD-L1 alone as prognostic markers for predicting treatment with immune checkpoint inhibitors [2]. As reported in these cases, quantifying spatial relationships between two biomarkers can provide clinical and/or biological insights. Spatial analytics of the TME requires accurate cell segmentation and classification. To that end, NeoGenomics have developed an IHC assay for combined CD8 and 22C3/PD-L1 staining, as well as a deep learning pipeline to automatically identify, segment, and classify CD8+ and PD-L1+ cells from whole slide IHC images. We see a 95% concordance of the number of CD8+ and PD-L1+ cells detected in our double stained IHC assay with serial sections stained either for CD8 or PD-L1 alone. We integrated our previously-reported cell segmentation and classification workflow used for MultiOmyx data (multiplexed IF images) [3] with Indica HALO to analyze whole slide IHC images double stained using two distinct chromogens. In this study, we performed IHC double staining and cell classification analysis on a CD8 - PD-L1 assay in NSCLC tumors. In addition to cell segmentation and classification image outputs, we also generate cell-level and slide-level tables with various cell morphological information, phenotype counts and densities, and biomarker intensity values that can be used to automatically define H-score measures for each biomarker. Additionally, advanced spatial analytics is performed to calculate the average distance between cells of various phenotypes, and spatial clustering patterns of different phenotypes in the TME (i.e. CD8+ and PD-L1+ cells). These analyses enable investigation of numerous complex cell interactions in TMEs. NeoGenomics quantitative double stained IHC assay is compatible with any two biomarkers of interest even if they are expressed on the same cell as long as the sub-cellular localization of the markers is different. The combination of double staining for CD8 and PD-L1, and quantifying spatial relationships between the two biomarkers is capable of providing the needed context to guide treatment decisions.