Pictor Labs Data & Edvidence

Frequently Asked Questions

What is Virtual Staining?

Virtual staining simply takes the traditional histochemical process and digitizes it. While the process involves biopsy, tissue sectioning, and slide preparation, we diverge from the traditional approach. Utilizing a standard fluorescence scanner, we capture autofluorescence images, which are then digitized. We also utilize brightfield scanners and capture brightfield images (ReStain). Through machine learning algorithms, these images are transformed into virtual stains. This innovative method enables pathologists to obtain multiple stains from a single tissue section.

How does Virtual Staining work?

The core of virtual staining lies in our supervised deep learning workflow. A significant focus of our efforts is on precise image registration at the pixel, and even sub-pixel, level. Our approach maps autofluorescence images of unstained tissue sections to their corresponding stained versions, such as IHC or H&E stains.

Is Virtual Staining approved for clinical use or research-use only?

We are engaged in projects that are assessing our technology on a clinical level; however, our technology is currently for research-use only.

How much tissue is needed to create a virtual stain model?

Typically we can see a proof of concept for a marker converge within 20 slides. We see good generalization in ~100 slides. We build robustness by augmenting pre-trained models with additional diseases. Our models converge quickly because we train them against staining results from the same slide as the AF image. In this way we get pixel perfect matching of the AF signal and staining patterns for the network to learn.

How long does it take to train a new virtual stain model?

For custom biomarker services, we can typically readout within 90 days from the completion of scanning training data.

How do you validate your virtual staining into my workflow?

We have the capability to evaluate specific regions and address particular challenges, if present, within those areas. Currently, our approach closely aligns with the established validation standards for digital pathology scanner validation, which means, establishing concordance between the virtual and chemical stains following a washout period between the two reads. Concordance studies remain central to model validation, while we continue to explore alternative evaluation methods to strengthen our overall approach.

How do I incorporate virtual staining into my workflow?

For labs, the workflow remains consistent with slide preparation, but with the added advantage of a comprehensive, one-stop-shop solution for stain work and review. Lab managers no longer need to concern themselves with maintaining stock levels of staining reagents or validating such processes. Our technologies streamline the histochemical process, creating a simpler, more efficient workflow for the entire team.

What hardware or software is compatible with virtual staining?

Our virtual staining technology is designed to work symbiotically with AI-based histopathology. The virtual stains and the cloud-based infrastructure seamlessly integrate with digital pathology and computer-aided-diagnostic platforms. Additionally, we offer smooth integration with 3rd-party software to provide data storage, collaboration tools, and cloud-based platform access.

Scientific Publications

Multi-modal spatial analysis of classic Hodgkin lymphoma microenvironment utilizing multiplex immunofluorescence and virtual staining

Combination analysis of tumor-associated collagen frameworks and tumor immune phenotype of lung carcinomas using virtual staining

Assessment of AI Computational H&E Staining Versus Chemical H&E Staining For Primary Diagnosis in Lymphomas

Frequently Asked Questions