Data & Evidence | Pictor Labs

Frequently Asked Questions

What is Virtual Staining?

Virtual staining is a computational method that generates stain-equivalent images from digitized tissue scans.

Instead of applying chemical dyes, tissue is scanned using autofluorescence or brightfield imaging, and a trained AI model produces a virtual representation of stains such as H&E or IHC.

The original tissue is not chemically altered, allowing multiple stain representations from a single section.

Learn more: What Is Virtual Staining

Does Virtual Staining replace pathologists?

No.

Virtual staining is adjunctive in nature and generates images that are visually comparable to chemical stains. It does not interpret findings, assign diagnoses, or make clinical decisions.

Pathologists remain responsible for reviewing images and determining diagnostic conclusions. The system is designed to support visualization and workflow flexibility, not to replace medical expertise.

Is Virtual Staining approved for diagnostic use?

Pictor Virtual Staining is currently offered for research-use only (RUO). It is not cleared or approved for clinical diagnostic use.

It should not be used as a standalone basis for medical diagnosis.

Regulatory status may evolve pending appropriate validation and regulatory review.

How is the AI system governed?

Governance includes structured performance evaluation, expert review, and defined deployment controls.

Model performance is evaluated using quantitative image metrics and blinded pathologist review. Diagnostic concordance studies are conducted for evaluation purposes only.

Data processing and storage depend on the selected deployment model (On-Prem or Cloud) and are governed by service agreements.

Learn more: How Safe AI Works in Pathology

How are pathologists involved?

Pathologists are involved in:

Dataset development
Blinded evaluation studies
Image quality assessment
Concordance review

Human interpretation serves as the reference standard during validation. AI outputs are reviewed by qualified pathologists and are not accepted autonomously.

How does Virtual Staining work?

The process involves three stages:

Capture of unstained tissue using autofluorescence or brightfield imaging
AI-based registration and mapping using paired training data
Generation of stain-equivalent images through learned transformations

The system relies on supervised deep learning trained on co-registered images from the same tissue section.

Learn more: What Is Virtual Staining

Can multiple stains be generated from one tissue sample?

Yes.

Because the tissue is not chemically altered during virtual staining, multiple computational stain representations can be generated from a single section.

This may be beneficial when tissue samples are limited or when multiple analytical approaches are required.

How is Virtual Staining evaluated?

Performance evaluation combines quantitative technical assessment with expert clinical review.

During development, the model is evaluated using objective image-based measures such as PSNR, SSIM, and feature-level quantitative analyses. These metrics facilitate assessment of image quality, structural integrity, and color accuracy compared to reference stains.

Additionally, pathologists review to verify that the virtually stained images meet standards for visual quality and diagnostic interpretability.

Additional details are available on the Safe AI page.

Where does the system run?

Pictor Virtual Staining supports two deployment models:

On-Premises: Processing occurs within the customer’s internal infrastructure.
Cloud: Processing occurs in a managed cloud environment.

Deployment configuration determines where data is processed and stored.

See: Deployment Options

What hardware and software is required?

Hardware requirements depend on deployment model:

On-Prem‍

GPU-enabled server
Internal IT management
Network integration with scanners and storage systems

On Cloud

Secure internet connection
Image upload or API integration
No local GPU required

Technical specifications are provided during onboarding.

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

Virtual Staining 101

What Is Virtual Staining?

Virtual Vs. Chemical Staining

How Safe AI Works in Pathology

Frequently Asked Questions