Reproduce “GESTALT: Fast, Generalizable Histopathology Image Synthesis from Gene Expression with Pretrained Pathology Foundation Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Felix Ho, Alexandros Graikos, Dimitris Samaras
GESTALT predicts a pathology-image embedding from spatial gene measurements, then uses a pretrained decoder to render a tissue image. On Xenium C1, it reports FID 15.75, SSIM 0.350, and FeatDist 12.66, improving on GeneFlow’s reported scores. The random tile split may include tiles from the same slide, and transfer degrades under harder cohort shifts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.