Tissue is one of oncology’s most valuable resources – and often one of its most limited. Traditional staining workflows can consume precious material with each additional test, while also introducing variability between sections. Virtual staining offers a way to preserve scarce tissue while supporting integrated analysis across pathology and downstream molecular workflows.
We spoke with Megan Rothney, VP of Product at Pictor Labs, about newly published data presented at the DP&AI USA Congress 2026.
What clinical or laboratory workflow challenges led to this research into virtual staining and integrated tissue analysis?
Virtual staining addresses a growing challenge in pathology: the amount of tissue available from a patient sample has not changed, but the demands placed on that tissue have increased significantly. Expanded immunohistochemistry (IHC) panels, routine next-generation sequencing (NGS), and emerging spatial biology applications all compete for the same limited material.
For small biopsies, particularly needle biopsies, those demands can quickly exceed what the tissue block can provide. When molecular testing returns a quantity not sufficient (QNS) result, clinicians may have to make treatment decisions without critical information, and some patients require repeat biopsies.
Pathologists and oncologists need as much information as possible from every specimen. Virtual staining is a logical place to start because every patient already has an H&E slide generated as part of routine care. If additional information can be derived from that slide, tissue can be preserved for downstream testing while expanding the insights available for patient management.
How does virtual staining compare with conventional H&E and special stains from a diagnostic perspective?
The process starts with an unstained tissue section. We capture an image using autofluorescence – the natural signal produced when tissue is exposed to fluorescent light – and a deep learning model translates that image into the desired stain, whether H&E, an IHC marker, or a special stain.
The model is trained on paired data, where the same tissue section is imaged unstained and then chemically stained. Once trained and locked, it can generate a diagnostic image without the section ever going through a stainer. The pathologist reviews the result in the same whole-slide viewer they already use.
The key question is whether pathologists reach the same interpretation from a virtual stain as they do from the chemically stained version of the same tissue. Across our work so far, the answer has consistently been yes. The images look familiar, and the diagnostic features pathologists rely on are preserved.
Virtual staining is not replacing the underlying biology. It provides a digital version of stains pathologists already trust while preserving tissue for any additional testing the case may require.
Could you explain how virtual staining works and how it compares with conventional H&E and special stains from a diagnostic perspective?
The process starts with an unstained tissue section. We capture an image using autofluorescence – the natural signal produced when tissue is exposed to fluorescent light – and a deep learning model translates that image into the desired stain, whether H&E, an IHC marker, or a special stain.
The model is trained on paired data, where the same tissue section is imaged unstained and then chemically stained. Once trained and locked, the model can generate the diagnostic image without the section ever going through a stainer. The pathologist then reviews the virtual stain in the same whole-slide viewer they already use.
The key question is whether pathologists reach the same interpretation from a virtual stain as they do from the chemically stained version of the same tissue. Across our work so far, the answer has consistently been yes. The images look familiar, and the diagnostic features pathologists rely on are preserved.
This approach is not intended to replace the underlying biology. It provides a faithful digital version of stains pathologists already trust, while preserving the original tissue block for whatever testing the case may require next.
How important is tissue preservation becoming in modern pathology workflows?
Tissue preservation is becoming one of the defining challenges in modern pathology. Technologies such as spatial transcriptomics, spatial proteomics, and high-plex imaging often require multiple high-quality serial sections from precisely selected regions of tissue. Those analyses can only be performed if sufficient tissue remains available.
Historically, pathology workflows were built around H&E and IHC, with molecular testing using whatever tissue remained. Today, molecular findings frequently drive treatment selection, trial eligibility, and patient management. Virtual staining offers a way to rethink that workflow by preserving tissue for the analyses that truly require it.
What did your study show regarding the quality, consistency, and diagnostic reliability of virtual stains?
Across studies conducted by us and our collaborators, virtual stains have shown high concordance with their chemically stained counterparts when reviewed by trained pathologists. These evaluations involve side-by-side assessment of virtual and conventional stains from the same tissue, with any discrepancies analyzed at both the case and feature level.
One of the key advantages of virtual staining is consistency. Conventional staining can vary because of reagent lots, instrument performance, laboratory workflows, and stain maintenance. Once a virtual staining model is trained, validated, and locked, it generates images consistently across cases.
For pathologists, this creates a more uniform reading experience. For computational pathology applications, it also reduces variability in the input images themselves.
How important is calibration and standardization when implementing virtual staining across different laboratories and workflows?
Calibration and standardization are critical for broader clinical adoption. Pathology has good reason to be cautious about technologies that perform well in development but fail to generalize across laboratories.
That means clearly defining model inputs, validating performance across different scanner platforms and workflows, and maintaining strict version control. Robust validation is essential to ensure reliable and reproducible performance regardless of where the technology is deployed.
What are the main barriers to wider adoption of virtual staining in clinical practice and research settings?
The barriers are largely operational rather than technical. Virtual staining depends on digital pathology infrastructure, so adoption is closely tied to the broader transition to digital pathology, which remains uneven across institutions.
Any technology that influences diagnostic interpretation must also undergo local validation. A practical approach is to run virtual staining alongside conventional staining during implementation. This allows laboratories to assess concordance using their own tissue, scanners, and pathologists before gradually transitioning selected case types.
Looking ahead, how do you see virtual staining shaping the future of digital pathology and molecular diagnostics?
I expect virtual staining to become a standard component of the digital pathology workflow rather than a separate technology category. Pathologists may eventually think about it the same way they think about an H&E stain today – as a reliable image source rather than a distinct process.
Its greatest impact will likely come from enabling more efficient use of limited tissue. By generating routine diagnostic stains digitally, tissue can be reserved for analyses that require physical material, including NGS, spatial transcriptomics, proteomics, and multiplexed immunofluorescence.
Ultimately, that means a single biopsy could provide a more complete picture of disease, helping clinicians make better-informed treatment decisions while reducing the need for additional procedures.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
