Increasingly, highly multiplexed tissue imaging methods are used to profile protein expression at the single-cell level. However, a critical limitation is the lack of robust cell segmentation tools for tissue sections. We present Multiplexed Image Resegmentation of Internal Aberrant Membranes (MIRIAM) that combines (a) a pipeline for cell segmentation and quantification that incorporates machine learning-based pixel classification to define cellular compartments, (b) a novel method for extending incomplete cell membranes, and (c) a deep learning-based cell shape descriptor. Using human colonic adenomas as an example, we show that MIRIAM is superior to widely utilized segmentation methods and provides a pipeline that is broadly applicable to different imaging platforms and tissue types.

Colorectal cancer exhibits dynamic cellular and genetic heterogeneity during progression from precursor lesions toward malignancy. Leveraging spatial molecular information to construct a phylogeographic map of tumor evolution can reveal individualized growth trajectories with diagnostic and therapeutic potential. Integrative analysis of spatial multi-omic data from 31 colorectal specimens revealed simultaneous microenvironmental and clonal alterations as a function of progression. Copy number variation served to re-stratify microsatellite stable and unstable tumors into chromosomally unstable (CIN+) and hypermutated (HM) classes. Phylogeographical maps classified tumors by their evolutionary dynamics, and clonal regions were placed along a global pseudotemporal progression trajectory. Cell-state discovery from a single-cell cohort revealed recurring epithelial gene signatures and infiltrating immune states in spatial transcriptomics. Charting these states along progression pseudotime, we observed a transition to immune exclusion in CIN+ tumors as characterized by a novel gene expression signature comprised of DDR1, TGFBI, PAK4, and DPEP1. We demonstrated how these genes and their protein products are key regulators of extracellular matrix components, are associated with lower cytotoxic immune infiltration, and show prog- nostic value in external cohorts. Through high-dimensional data integration, this atlas provides insights into co-evolution of tumors and their microenvironments, serving as a resource for stratification and targeted treatment of CRC.