Inform v 2.4 image analysis software

Immunofluorescence images were acquired with the Vectra V.3.0.5 multispectral imaging microscope (PerkinElmer) at 20× magnification and exposure times were set to avoid spectral overlap. Immune cells in the TME were automatically phenotyped and counted with inForm V.2.4 image analysis software (PerkinElmer), after manual training and validation of procedures. The software was trained to segment epithelial and stromal fractions, segment DAPI +nucleated cells, and to assign a phenotype to each cell (online supplementary file 3a). All images were visually inspected on accurateness in segmenting tissue and phenotyping cells, and if errors were detected, the training was further optimized. Given the multitude of possible coexpressed markers, each seven color panel was divided into multiple subanalyses (online supplementary file 3b) to preclude the exclusion of relevant phenotypes, after which the phenotypes of the different subanalyses were merged per cell based on X, Y-positions to obtain the full seven marker expression profile of each cell. Immune cell counts were normalized for tissue size (cells/mm² epithelium and cells/mm² stroma). After merging all subanalyses, a threshold of a median cell count ≥10 cells/mm² was applied to enable the analysis of relevant phenotypes.