Histopathology is the gold standard for disease diagnosis through microscopic examination of tissue. Conventional methods depend on physical sectioning and two-dimensional (2D) analysis, which can introduce artifacts and obscure three-dimensional (3D) tissue architecture. Although computational reconstruction from serial sections is possible, it is labour-intensive and susceptible to registration errors. X-ray-based virtual histology offers a complementary, non-destructive alternative by enabling 3D imaging of intact tissue. Soft tissues exhibit weak X-ray attenuation contrast, and phase-contrast methods substantially improve visibility. However, they lack the tissue specificity provided by conventional histological stains. Recent developments in X-ray compatible stains like hematein-lead complexes create new opportunities, but quantitative interpretation is hindered by edge artifacts and the difficulty of separating stain and tissue signals accurately. This work addresses these challenges by introducing a modulation-based X-ray imaging framework capable of quantitative material decomposition at the micrometer resolution.
The aim of this study was to develop and validate a virtual histology method that quantitatively maps stain concentrations while simultaneously providing high-contrast 3D tissue morphology at a scale relevant to pathology. It seeks to overcome limitations of modulation-based X-ray imaging for quantitative attenuation measurements, enabling reliable material decomposition to differentiate tissue and stain contributions and to establish correspondence between X-ray virtual histology and conventional optical histology.
The method combines modulation-based X-ray phase-contrast tomography with physics-informed signal correction and material decomposition. Initial system calibration and performance assessment were conducted using a multi-material phantom with known X-ray properties to evaluate the accuracy of the retrieved electron number density and attenuation coefficients. Biological validation used mouse and rat kidney samples, either unstained or stained with hematein lead complexes that selectively bind nuclear DNA and provide contrast in both visible light and X-rays. Imaging was performed at synchrotron beamlines using Talbot array modulators and highly coherent monochromatic X-ray beams. Phanse and attenuation signals were retrieved by unified modulated pattern analysis, followed by tomographic reconstruction.
A modified Paganin-based correction removed Laplacian phase effects that induce boundary-related edge artifacts, which corrupt attenuation measurements under coherent light. Quantitative material decomposition was performed by expressing each voxel’s attenuation and electron density as a linear combination of soft tissue and lead. Independent validation of lead concentration was obtained using K-edge subtraction imaging. Virtual histology volumes were registered to conventional optical histology sections for direct comparison.
Phantom experiments demonstrated that modulation-based imaging accurately retrieves electron number density and attenuation coefficient in homogeneous regions, with strong agreement between measured and theoretical values. Pronounced edge artifacts at material boundaries were observed in uncorrected attenuation images, which explain why earlier methods failed in biological tissue dominated by interfaces. The proposed correction successfully removed these artifacts, improved noise characteristics, and slightly improved spatial resolution. The technique produced 3D maps of stain concentration at micrometer resolution, clearly separating lead-based stain distribution from soft tissue characteristics in stained mouse kidney samples.
Cell nuclei were selectively highlighted in false-color images that closely resembled traditional histology. Beyond providing better spatial resolution and eliminating the need for multi-energy collection, quantitative comparison with K-edge subtraction imaging revealed excellent agreement, high correlation, and minimal bias. Virtual histology slices demonstrated great visual concordance with optical microscopy in rat kidney samples, faithfully representing renal structures, including glomeruli and arteries. X-ray virtual histology was the only method that offered quantitative, non-destructive 3D data, although optical histology still resolves finer details.
This study introduces the first modulation-based X-ray imaging method capable of qualitatively extracting stain concentrations while preserving tissue morphology at histopathology-relevant scales. The technique bridges the long-standing gap between phase-contrast X-ray imaging and tissue specificity of traditional histology by eliminating edge artifacts and allowing for robust material decomposition. Validation against K-edge imaging and direct comparison with optical sections confirm quantitative accuracy and clinical relevance. This method complements traditional pathology by providing a whole sample 3D context to guide targeted sectioning and analysis. The framework opens new opportunities for 3D quantitative pathology, X-ray stain development, and creation of high-quality datasets for AI-driven virtual staining and automated tissue analysis, though challenges remain in achieving optical-level resolution, optimizing stain uniformity, and expanding decomposition to more complex tissues.
Reference: John D, Paganin DM, Zdora M-C, et al. Quantitative stain mapping in X-ray virtual histology. Adv Sci. 2026;19783. doi:10.1002/advs.202519783
