SheafStain: Sheaf-Theoretic Schrodinger Bridge for Spatially and Biologically Coherent Virtual Staining

arXiv:2606.11846v1 Announce Type: new Abstract: Current virtual staining approaches offer the potential for time- and cost-efficient biomarker quantification in cancer diagnostics and prognostics. However, patch-wise inference for gigapixel whole slide images (WSIs) fails to maintain spatial continuity, yielding artifacts that cause catastrophic mismatches with ground-truth images. Although pathology Vision Foundation Models (VFMs) offer rich representations, their self-attention causes varying global contexts to produce inconsistent embeddings for the same physical region. We formalize and validate this ``context contamination'' as a sheaf-theoretic problem where these embeddings form a presheaf that violates the gluing axiom. To address this, we propose SheafStain, a new approach that reinterprets VFM features as sheaf-like sections for spatially and biologically coherent virtual staining. Specifically, SheafStain integrates class and patch tokens into a Schrodinger Bridge framework as sheaf-like sections. While the class token anchors biological consistency, patch tokens form a per-position spatial map. A backbone co-pretrained on Hematoxylin & Eosin (H&E) and Immunohistochemistry (IHC) yields non-degenerate cross-stain stalks, so a single VFM feature space supervises both input conditioning and output stain alignment. Departing from prior work that evaluates on isolated 256 imes 256 patches and either random-crops or resizes the 1024 imes 1024 ground truth, we translate at 256 imes 256 and evaluate on the stitched 1024 imes 1024 outputs across HER2, ER, PR, and Ki-67. SheafStain demonstrates promising results against six prior methods while mitigating patch-boundary stitching artifacts. Code will soon be released.

Source: arXiv cs.CV | 2026-06-11