What problem does it solve? Mapping cells from 10x Xenium mouse brain sections to standardized Allen Brain CCFv3 regions requires cross-modal diffeomorphic alignment that is fragile: AP slice drift, axis flips, and EM loss collapse routinely produce wrong region assignments. This Skill encodes a validated pipeline with known pitfalls, tuned parameters, and a source patch so each cell gets a correct Allen ontology annotation. ## Core Features & Use Cases - End-to-end registration pipeline: Load Xenium cell coordinates, rasterize sections, match the best CCF coronal slice via edge-based Dice/NCC scoring, run GPU-accelerated LDDMM 3D-to-slice alignment, and warp atlas annotations into Xenium space. - Per-cell brain region annotation: Map every cell to Allen regions with ontology hierarchy traversal, depth-based rollups, and official Allen colors, exported as CSV/JSON plus overlay, parcellation, scatter, and bar-chart visualizations. - Battle-tested troubleshooting guidance: Documents the freeze_ap source patch, negative scale_x DV-flip fix, sigmaM tuning, and atlas brain-mask construction to avoid common failure modes. - Use Case: A researcher with a fresh-frozen mouse brain Xenium run needs per-cell brain region labels for a hippocampal study; this Skill produces the aligned annotation table and publication-ready figures on a GPU server via Pixi and Quarto. ## Quick Start Ask the agent to register your Xenium cells.csv.gz to the Allen CCFv3 atlas and generate a brain region annotation report using the provided Pixi environment and Quarto template.
