Fig. 1. Overview of our analyses.

A We generated multi-antigen images for skin tissue samples from CTCL, AD, and PSO patients. Subsequently, images were pre-processed via cell segmentation, cell-level protein abundance quantification, and cell-type assignment. B We then computed spatial graph representations for all samples, which we analyzed using Squidpy, Leibovici’s entropy, modularity, as well as different heterogeneity scores implemented in our Python package SHouT: local and global entropy, local and global homophily, and egophily.