Digital pathology and artificial intelligence-based approaches to characterize the complex interactions between cellular components of the tumor microenvironment and their spatial distribution

Digital pathology coupled with artificial intelligence (AI)-based approaches is receiving great attention in the field of oncoimmunology, as this can improve current diagnostic workflows and potentiate the analytic outputs. The aim of this work consists of combining different histopathological approaches and high-throughput computational imaging tools to analyse and dissect the tumor microenvironment (TME). Firstly, we focused on tumor tissue and structure, by computational analysis of H&E-stained slides. So, we trained a deep learning algorithm to identify tumor cells and we explored their spatial distribution by deploying the Ripley’s K-function. The resulting K-score value classifies each tumor spot as diffuse, poorly or highly clustered. To better understand the complex interactions between TME cellular components and their spatial distribution, we performed a deeper investigation of the immune contexture, taking advantage of the HyperionTM Imaging System. By preserving tissue architecture and cell morphology information, it allows the simultaneous investigation of 23 protein markers related to tumor cells, tissue architecture and immune cells. We made multiparametric computational analysis of the IMC images to firstly distinguish between tumor and stromal tissues, and then to evaluate the frequency of immune cell populations in the tumor nests versus fibrotic stroma and their interactions. Finally, in poorly and highly clustered samples, we better investigated the tumor heterogeneity in terms of interactions between immune cells and tumor cell distribution within the tissue. The results of our analysis are expected to provide the opportunity to investigate spatial patterns and cell interactions at single-cell level, leading to the identification of tumor patient profiles with clinical relevance.