Anatomical pathology
In 95% of cases, cancer diagnosis relies on anatomic pathology, a key medical field which focuses on tissues and cells analysis.
As a central part of modern medicine, pathologists are facing growing pressure to accelerate diagnostic workflows and improve clinical accuracy.
With its infrared imaging plateform and deep-learning solutions, ADMIR ambitions to tackle both issues at once.
Introducing MirStain
Infrared imaging for enhanced digital staining
Staining-free
Fast
Reliable
Anotomical pathology relies on tissue sampling, usually by biopsy. These samplings are then cut into slices and deposited on microscopy slides. However, the transparency of thin human tissues prevent direct observation, which forces biologists to stain samples, using usually toxic chemicals. Only then, can pathologists apply their expertise.
With MirStain, we offer a stain-free solution for tissue analysis. It combines high-resolution visible imaging and spectral infrared imaging to analyze unstained tissue slides. The resulting images are detailed enough to digitally stain the original sample, thus skipping the need for chemical staining steps. This dramatically reduces diagnostic time and laboratory costs.
100% diagnostic time reduction
30% laboratory costs reduction
30% less water and chemicals
95% analysis accuracy
Technique and assets
Spectral imaging
Our infrared scanner incorporates several discrete quantum cascade lasers (QCLs), each tuned to emit at an absorption wavelength of a major biomolecule class (nucleic acids, proteins, carbohydrates…). This spectral mapping allows to fully characterize a tissue.
Biomodal system
We developed a bimodal imaging platform that combines infrared lens-free optics, with conventional brightfield visible-light microscopy. This combination preserve the subcellular spatial resolution of conventional microscopy, while adding the chemical data of an infrared stack.
Digital staining – stable, precise, reliable
MIRSTAIN optical staining ensures consistent slide results by eliminating chemical process variability, maintaining uniform color and contrast in cellular and nuclear structures across labs, operators, and time.
A demonstration study already collected data from 40 tissues slides. So far, we achieve state of the art performances in digital staining (PSNR: 22.8, MSSIM: 0.82). New accquisitions will soon allow to expand beyond this thershold.
Special digital staining & tissue segmentation
MIRDIAG generates whole slide infrared images that reveal biomarker-specific wavelengths for tissue segmentation to support diagnosis and prognosis of various pathologies in a novel way. Infrared spectral features provide greater specificity and reliability in diagnostics, enabling the investigation of cellular disorders and tumorigenesis mechanisms.
Workflow – faster, smarter, cheaper
The MIRSTAIN scanner creates digitally stained whole slide images from translucent, unstained slides—eliminating the need for chemical stains or reagents. This dramatically simplifies inner worflows and decrease time-to-result.
Generative Deep Learning
A deep learning model processes features from both brightfield and infrared image stacks. It integrates morphological and chemical information to enable optical H&E staining. Infrared modality provides spectral information, acting as a reference palette, while the brightfield image serves as the structural template that the model can colorize accordingly.
Further reading
Duraffourg, L., Borges, H., Fernandes, M., Beurrier-Bousquet, M., Baraillon, J., Vianey, K., et al. (2025). Pre-print.
In this work, we introduce a fast, large-field bimodal imaging platform that combines conventional brightfield microscopy with a lensless infrared imaging scanner capable of acquiring whole-slide IR image stacks within minutes. Leveraging a dedicated deep learning model, we implement…
White paper
In a few pages, get a full overview of the context of our work and the ambitions of ADMIR regarding Anatomic Pathology.