- [Resource Hub](/) - [Publication](/?resource-type=publication#library) - Cross-Domain Image Synthesis: Generating H&E from Multiplex Biomarker Imaging # Cross-Domain Image Synthesis: Generating H&E from Multiplex Biomarker Imaging Saurav JR, Nasr MS, Luber JM. arXiv . Posted August 5, 2025. arXiv [2508.04734](https://arxiv.org/abs/2508.04734). How to cite ### AMA Saurav JR, Nasr MS, Luber JM. Cross-Domain Image Synthesis: Generating H&E from Multiplex Biomarker Imaging. arXiv . Posted August 5, 2025. arXiv:2508.04734 ### APA Saurav, J. R., Nasr, M. S., & Luber, J. M. (2025). Cross-Domain Image Synthesis: Generating H&E from Multiplex Biomarker Imaging. arXiv . https://arxiv.org/abs/2508.04734 ### BibTeX @misc{saurav2025crossdomain, title = {Cross-Domain Image Synthesis: Generating H&E from Multiplex Biomarker Imaging}, author = {Saurav, Jillur Rahman and Nasr, Mohammad Sadegh and Luber, Jacob M.}, year = {2025}, eprint = {2508.04734}, archivePrefix = {arXiv}, primaryClass = {q-bio.QM} } Multiplex immunofluorescence can read dozens of proteins across a tissue section, but most of the analysis software pathologists rely on was built for ordinary hematoxylin-and-eosin stains. That mismatch keeps rich molecular images from reaching established tools. This group trained image-generation models to turn multiplex immunofluorescence into a matching H&E view of the same tissue. They tested the models on two public colorectal-cancer datasets, one of them the 19-channel Orion set, and asked not only whether the pictures looked right but whether downstream software read them correctly. A two-level generative model produced the most useful synthetic stains on the Orion data, holding up under automated tissue classification and nuclei segmentation. [Read publication at arXiv](https://arxiv.org/abs/2508.04734) ## Key findings - The Orion colorectal-cancer dataset supplied the 19-channel arm of the benchmark. Its paired multiplex immunofluorescence and H&E images were split into 47,824 training, 10,800 validation and 10,800 test patches (Table I). - On the Orion data, a two-level vector-quantized network gave the best image reconstruction of the three models tested. It reached the lowest L1 error (0.1491) and the highest structural-similarity and peak-signal-to-noise scores (SSIM 0.5221, PSNR 20.38 dB), ahead of a single-level network and a conditional-GAN baseline. - The same model's Orion outputs led on both downstream checks. A pre-trained tissue classifier agreed with ground-truth H&E on 69.9% of Orion images, and watershed nuclei segmentation reached a mean overlap of 0.6800, each the top score on the Orion set. ## How this study used Orion data “Orion CRC Dataset: This dataset, featuring 19-channel” — Saurav et al., arXiv (2025), Methods, “Datasets” Provenance: this study used the publicly available Orion CRC dataset (Lin et al., Nature Cancer 2023); the authors did not run a RareCyte instrument. Disclosure: RareCyte is listed as an author affiliation on the publication cited above. Disclosure: RareCyte is named in the competing-interests statement of the publication cited above. ## Why it matters for Orion users If you are weighing the Orion platform for a tissue study, look at what its data was asked to carry here. This group ran no instrument of their own; they reached for a public benchmark, and the Orion colorectal-cancer set was one of the two they chose. What made it usable for this problem is the same thing that makes it useful at the bench. Orion captures high-plex immunofluorescence and a hematoxylin-and-eosin view of the same tissue section, spatially registered, in a single staining round. Because the fluorescence and the morphology come from one piece of tissue rather than two serial cuts, a model can learn the mapping between them cell for cell; the 19-channel Orion arm supplied 47,824 aligned training patches to learn from. For your own work, that registration is the payoff. When your molecular readout and your morphology share the same coordinates, you can hand either one to tools built for the other, instead of stitching together two separately stained slides after the fact. [Back to all resources](/#library) ## Related resources [Article](https://www.selectscience.net/article/advanced-spatial-biology-sheds-new-light-on-therapeutic-research) Orion SelectScience ### Advanced spatial biology sheds new light on therapeutic research [Read Article](https://www.selectscience.net/article/advanced-spatial-biology-sheds-new-light-on-therapeutic-research) [Article](https://www.selectscience.net/editorial-articles/bringing-spatial-biology-to-the-clinic-a-new-lens-on-cancer-biology/?&artID=58147) Orion SelectScience ### Bringing spatial biology to the clinic: “A new lens on cancer biology” [Read Article](https://www.selectscience.net/editorial-articles/bringing-spatial-biology-to-the-clinic-a-new-lens-on-cancer-biology/?&artID=58147) [Article](https://www.selectscience.net/article/how-to-unlock-the-complexity-of-tumor-microenvironments) Orion SelectScience ### How to unlock the complexity of tumor microenvironments [Read Article](https://www.selectscience.net/article/how-to-unlock-the-complexity-of-tumor-microenvironments) [Brochure](https://go.pardot.com/l/455452/2020-07-15/8p7453/455452/1606951069teslfZQe/BR21_101_Orion_Brochure_201202.pdf) Orion ### Orion Spatial Proteomics Platform [View Brochure](https://go.pardot.com/l/455452/2020-07-15/8p7453/455452/1606951069teslfZQe/BR21_101_Orion_Brochure_201202.pdf) [Spec Sheet](https://pardot.rarecyte.com/l/455452/2026-02-24/965v3w/455452/1771949067AeuR6syb/Orion_HT_spec_sheet.pdf) Orion ### Orion HT [View Spec Sheet](https://pardot.rarecyte.com/l/455452/2026-02-24/965v3w/455452/1771949067AeuR6syb/Orion_HT_spec_sheet.pdf) [Spec Sheet](https://pardot.rarecyte.com/l/455452/2025-10-31/961nyb/455452/1761933957WEJWJyeN/2025.07.14___SS___Orion__LE_.pdf) Orion ### Orion LE [View Spec Sheet](https://pardot.rarecyte.com/l/455452/2025-10-31/961nyb/455452/1761933957WEJWJyeN/2025.07.14___SS___Orion__LE_.pdf) [Spec Sheet](https://go.pardot.com/l/455452/2020-07-15/8p742r/455452/1607104572nfrlJouR/DF_24_101_201203_Orion_Datafile.pdf) Orion ### Orion [View Spec Sheet](https://go.pardot.com/l/455452/2020-07-15/8p742r/455452/1607104572nfrlJouR/DF_24_101_201203_Orion_Datafile.pdf) [Research Spotlight](/resources/spatial-profiling-immune-correlates-mss-crc/) Orion 2026 ### Spatial Profiling Uncovers Immune Correlates of Durable Response in Phase 1/2 MSS Colorectal Cancer Trial [Read Spotlight](/resources/spatial-profiling-immune-correlates-mss-crc/) [Publication](/resources/copanlisib-nivolumab-microsatellite-stable-colorectal/) Orion Nature Communications 2026 ### Copanlisib in combination with nivolumab for microsatellite stable colorectal cancer: a phase 1/2 trial [Read More](/resources/copanlisib-nivolumab-microsatellite-stable-colorectal/) [Publication](https://doi.org/10.1158/2159-8290.CD-26-0171) Orion Cancer Discovery 2026 ### Spatial Integration of Protein and Chromosomal States Reveals Early Copy-Number Changes and Genotype-Associated Immune Neighborhoods in Serous Ovarian Cancer Evolution [View Publication](https://doi.org/10.1158/2159-8290.CD-26-0171) [Publication](https://www.nature.com/articles/s41467-026-74933-w) Orion Nature Communications 2026 ### Multiple human transgenes prolong survival of triple-carbohydrate knockout porcine kidney xenografts in nonhuman primates [View Publication](https://www.nature.com/articles/s41467-026-74933-w) [Video](https://www.youtube.com/watch?v=KOiToZIGN0g) Orion 2026 ### Dr. Tanjina Kader on mapping spatial proteomics and chromosomal changes on the same tissue section [Watch Video](https://www.youtube.com/watch?v=KOiToZIGN0g)