STimage dataset with BreastCancer_Xenium and SkinVisium files

This dataset contains the diverse spatial transcriptomics (ST), proteomics, and clinical datasets used to develop and validate STimage, a comprehensive deep learning suite for predicting gene expression and classifying cell types directly from haematoxylin and eosin (H&E) stained histopathology images. Spanning three cancer types and one chronic disease, the collection features data from multiple platforms, including standard ST, single-cell resolution Xenium, and proteomics from PhenoCycler Fusion, linking tissue morphology with high-dimensional molecular profiles. The data is curated to train and benchmark models on regression and classification tasks, enabling researchers to develop novel computational pathology tools with a focus on model robustness, interpretability, and the creation of prognostic biomarkers for patient stratification.  A data record was published later with supplementary raw data added - see the above link.