Computational methods for the analysis of whole slide images in tumor tissues

The tumor microenvironment (TME) is highly populat-ed by immune cells like tumor associated macrophages (TAMs) and lymphocytes. The study of the TME in tissues, in terms of cellular composition and spatial or-ganization, could provide an insight of the role of the immune system within the TME, and provide correla-tions with clinical progression. To do this, we exploit-ed several imaging techniques and digital tools, ena-bling us to obtain complementary information. Main point of novelty of our work is the attempt to process the whole slide (WSI) of the tissue sections and not only small rectangular ROIs, giving us the op-portunity to compare cellular composition of different regions. Then, through computational analysis of slides stained by classical immunohistochemistry techniques (such as H&E and Ferangi), we studied the tumor structure and the distribution of specific macrophage markers. Moreover, the OpalTM Multiplex Immunohistochemis-try allowed us to identify and quantify cell subpopula-tions, cell morphology and localization. Finally, anoth-er promising technique is the HyperionTM technology, which allows us to concomitantly investigate up to 37 markers, avoiding auto fluorescence issues. To perform the computational analysis of the WSIs, we took advantage of specific open source platforms based on artificial neural networks such as QuPath[1] and CellProfiler[2], to segment cells and classify them in a supervised manner, while software like Cytomap was mainly used to perform spatial analyses like dis-tance analysis, colocalization and region interactions[3]. The potential of these computational tools is impres-sive and the results of the cell segmentation and classi-fication are very accurate, so we expect it will be of interest for researchers in the field.