Dual-Excitation PARS Microscopy Enables High-Fidelity Virtual Multi-Stain Histology Without Tissue Destruction

Key Takeaways

• Dual-excitation PARS microscopy enables label-free generation of multiple virtual histological stains from a single tissue section, preserving scarce biopsy material.
• Combining 266 nm and 355 nm ultraviolet excitation improves visualization of key features—including nuclei, collagen, red blood cells, and melanin—enhancing virtual stain accuracy.
• In a limited masked evaluation, pathologists rated virtual stains comparably to conventional staining and could not reliably distinguish between them.

A new study published in npj Imaging introduces a significant advance in digital pathology: a dual-excitation photon absorption remote sensing (PARS) microscopy platform capable of generating multiple virtual histological stains without physically altering tissue samples.

Histochemical staining remains the cornerstone of tissue-based diagnosis, with hematoxylin and eosin (H&E) serving as the clinical standard. However, conventional staining is inherently destructive, consumes finite tissue, and often requires additional sections for specialized stains such as Masson’s trichrome or periodic acid–Schiff (PAS). These limitations can be particularly consequential in small biopsies, where tissue conservation is critical.

The PARS approach offers a fundamentally different paradigm. Rather than applying chemical dyes, it captures endogenous optical signals—specifically radiative and non-radiative absorption—using pulsed ultraviolet excitation. In this study, investigators combined two excitation wavelengths (266 nm and 355 nm) to expand the range of detectable biochemical contrasts. The shorter wavelength provides strong nuclear contrast via nucleic acid absorption, while the longer wavelength enhances visualization of hemoglobin, melanin, and extracellular matrix components such as collagen and elastin.

As illustrated in the composite imaging examples on page 3, this dual-excitation strategy yields four complementary contrast channels in a single scan, enabling detailed visualization of tumor architecture, stromal features, and pigmentation patterns. These multimodal inputs are then translated into familiar histological stains using a deep learning framework (RegGAN), allowing the generation of H&E as well as specialized stains including Masson’s trichrome, PAS, and Jones methenamine silver.

Across a range of human and murine tissues—including renal cell carcinoma, melanoma, and fungal infections—the virtual stains closely resembled their chemical counterparts. As shown in representative results on page 5, key diagnostic features such as tumor nests, collagen deposition, fungal hyphae, and basement membranes were preserved with high fidelity. Notably, the system enabled simultaneous generation of multiple stains from a single image field, offering complementary diagnostic perspectives without additional tissue sectioning.

Quantitative analyses supported these observations. Models incorporating both excitation wavelengths outperformed single-wavelength approaches across structural and perceptual similarity metrics, reflecting improved reconstruction of clinically relevant features. For example, collagen detection in trichrome stains and nuclear identification in H&E were both significantly enhanced by dual-excitation input.

The study also included a masked evaluation by three pathologists. Although limited in scale, the results were notable: diagnostic quality ratings for virtual images were comparable to those of conventional stains, and pathologists were unable to reliably distinguish between the two, with classification accuracy near chance levels. These findings suggest that virtual staining may achieve a level of visual fidelity sufficient for histopathologic review, though larger validation studies are needed.

Despite its promise, the technology faces practical challenges. Acquisition speed remains a major limitation; current scan times are substantially longer than those of standard whole-slide scanners. Additionally, the study was conducted within a single imaging and staining environment, and broader validation across institutions and workflows will be necessary to establish generalizability.

Still, the implications are considerable. By enabling nondestructive, multi-stain visualization from a single tissue section, dual-excitation PARS microscopy could reshape workflows in both clinical pathology and research. In settings where tissue is limited—or where multiple stains are required for diagnosis—this approach offers a compelling, resource-efficient alternative that preserves both material and time.

As digital pathology continues to evolve, label-free virtual staining may emerge not as a replacement for traditional methods, but as a powerful adjunct—extending diagnostic insight while minimizing the trade-offs inherent to conventional histology.