Background: Pancreatic tumours exhibit substantial epithelial, stromal, vascular and immune diversity, complicating efforts to identify shared microenvironmental programmes across precursor and invasive lesions. We profiled 55 samples spanning pancreatic ductal adenocarcinoma (PDAC), intraductal papillary mucinous neoplasms (IPMN), pancreatic neuroendocrine tumours (PET) and ampulla of Vater carcinoma (AVC) using Imaging Mass Cytometry (IMC). Data were asinh-transformed, batch-corrected with fastMNN, clustered using Rphenograph and visualised using UMAP; phenotypes were interpreted with asinh-scaled dot plots and heatmaps. Clustering stability (k = 30, 45, 60) and neighbourhood structure were assessed via ARI/NMI and kNN contact patterns. Across entities, we resolved both conserved and disease-biased cellular states. PDAC showed a strongly stromal-dominant programme (Collagen-I/αSMA/Vimentin), with myeloid-rich and mixed T-cell niches positioned at tumour–stroma interfaces and variable epithelial pockets (panCK/EpCAM). Distinct phenotypic subsets, including cluster-22, were preferentially expanded in PDAC. IPMN displayed organisation (panCK/EpCAM/β-catenin) with emerging stromal and myeloid co-enrichment, and focal CAIX-positive hypoxic micro-niches suggesting drift towards PDAC-like features. PET exhibited segregated epithelial, stromal and immune territories, with clear enrichment of clusters 29/25 and 27/3 and restricted immune activation. AVC was skewed towards immune- and stromal-rich states, with a pronounced cluster-14 population and broad Vimentin/αSMA-positive regions.
Conclusions: Ongoing analyses will extend these comparisons to the full 746 sample dataset and integrate clinical metadata to identify spatial biomarkers with translational relevance for risk stratification and early intervention.