Most SHG research to date has relied on forward-detected signals acquired from ultra-thin tissue sections. Because the forward signal is coherent, the resulting images provide a clear and detailed representation of collagen in the sample under study. However, the requirement for the sensor to sit behind the sample generally limits forward-SHG to ex vivo experiments on thin sections. Backward-detected SHG, by contrast, uses the portion of the signal scattered back from the tissue and therefore holds real potential for translation to in vivo imaging contexts. The problem is that the origin of the backward signal is not fully understood. It represents complex optical interactions, and interpretation of backward images remains challenging, which has led to repeated misinterpretations of the collagen structure of biological tissues.

A team of researchers from University College London and the University of Arizona has now addressed this puzzle in a study published in Frontiers in Photonics. Using a custom optical chamber that allows bidirectional SHG imaging of the cornea under controlled pressure, they assessed the spatial relationship between features in backward images and the underlying structural features of collagen visible in paired forward images. Their central finding is striking: collagen waviness, known as crimp, is a key structural feature responsible for periodic bands of fluctuating signal intensity in backward SHG images. Structures that numerous previous studies had identified as collagen bundles, and even used to compute predominant orientations, are in fact oriented perpendicularly to the actual lamellar direction and originate from signal contributions from different regions of wavy lamellae.

The cornea was selected as the imaging target for several reasons. It is a transparent, collagenous tissue within which the collagen has a specific depth-dependent, highly ordered arrangement that has been characterized previously using histology, scanning electron microscopy, transmission electron microscopy, and X-ray diffraction techniques. The transparent nature of the tissue enables full-depth imaging in both directions of intact tissue, ensuring the structure closely represents the in vivo state, while the complex collagen arrangement and access to previously acquired structural information facilitate convenient validation and a means to assess the origin of imaging artefacts. Ten fresh porcine eyes were obtained from a local abattoir, and corneas were imaged within seventy-two hours postmortem in an intact, non-processed state to prevent structural modifications induced by sectioning or fixation agents.

Imaging was performed with a Zeiss LSM980 on an upright Examiner.Z1 microscope, with laser excitation provided by a Spectra-Physics Insight DS+ laser tuned to 820 nanometers, linearly polarized in the x-axis of the image plane. Initial imaging was performed with the cornea remaining within the eye globe at normal physiological intraocular pressure, using reflection-mode SHG at two magnifications. The cornea was then dissected from the globe with a three-millimeter rim of sclera and re-imaged to acquire simultaneous forward- and backward-detected SHG signals for direct comparison. The custom-designed chamber allowed the cornea to be pressurized with phosphate buffered saline while a manometer provided real-time pressure monitoring, enabling the team to investigate the effects of pressure variation on collagen bundles and the subsequent effect on the generated images. To the best of the authors’ knowledge, this is the first experimental setup described that enables forward imaging of the cornea in its unaltered, pressurized state, providing layer-by-layer detail and the ability to capture structural changes under different loads.

The results revealed a consistent spatial relationship between the two signals. Backward-detected images of the posterior corneal stroma showed highly organized, vertically aligned bands of high-intensity signal with periodic spacing, structures previously interpreted as collagen bundles, while areas of low signal were believed to be regions absent of collagen, corresponding to voids or interlamellar spacing. Higher magnification resolved speckled, elongated patterns oriented perpendicularly, rather than aligned, to the structures previously identified as bundles. When periodic regions of the backward images were overlaid with the corresponding regions of the forward images, the aligned bands in the backward image corresponded to regions of the wavy bundles observed in the forward images. The dark regions did not correspond to voids but to absent or weak backward signal due to the specific collagen position along the wavy lamellae. The forward-detected bundles consistently ran perpendicular to the backward-detected high-signal bands throughout the entire imaging plane.

To demonstrate this mechanism dynamically, the cornea was mounted in the transparent chamber and imaged throughout its depth, first at low pressure of approximately five millimeters of mercury and then at supraphysiological levels of approximately sixty millimeters of mercury, pressures relevant to disease states such as glaucoma and hypotony. It had been theorized, but not yet proven, that collagen waviness is a structural mechanism governing the corneal response to increasing intraocular pressure. At low pressure, the collagen bundles appeared wavy and crimped, with periodic features clearly evident in the backward images. When the pressure was raised, the forward image revealed straightened bundles, while the backward image showed a more uniform forward-to-backward signal and loss of the periodic bands previously observed. In some imaging planes where the collagen did not uncrimp completely, periodic patterns remained in the backward images in regions directly corresponding to the crimped areas in the forward images. A complete layer-by-layer recording across a thickness of 985 micrometers showed the collagen becoming noticeably uncrimped after the pressure increase, with clear differences in the backward signal between crimped and uncrimped states.

The findings also clarify the theoretical frameworks proposed to explain SHG directionality. Theoretical models based on fibril size suggest that fibrils with diameters on the order of the SHG wavelength generate predominantly forward-directed signals, whereas smaller fibrils produce nearly equal forward and backward signals. However, this theory cannot account for the difficulty in correlating backward and forward signals in the cornea, which has a consistent fibril diameter of approximately thirty-one nanometers in pigs across the depth. A more comprehensive quasi-phase matching model treats the fibril no longer as an isolated SHG generator but considers the collective packing of fibrils within a domain, which influences phase-matching conditions and the continuity of coherent second harmonic generation. The team extended this framework by introducing a tilt parameter, defined as the out-of-plane angle between the collagen bundles and the axial propagation direction. This inclination, inherent to the crimp pattern, induces a change in the effective domain size, affecting phase-matching conditions and thus the forward-to-backward signal distribution, creating a periodic alternation between different forward-to-backward ratios that matches the sides of the crimped patterns in the forward images.

The implications extend well beyond correcting image interpretation. A previous study using backward SHG imaging to evaluate corneal response to intraocular pressure changes had observed a narrowing of dark low-signal bands and attributed it to changes in interlamellar spacing. The new study clarifies that changes to these periodic intensity levels occur due to straightening of the crimped lamellae under increased pressures, meaning a key biomechanical mechanism governing the corneal response to pressure was previously overlooked. Because these structural changes can occur within very short timescales, during each heartbeat when the cornea experiences pulsatile intraocular pressure variations, this dynamic behavior must be considered when analyzing backward signals, particularly if an in vivo diagnostic device or AI-based model is to be developed to distinguish between normal and diseased states. The authors note that a key limitation of the work is the absence of quantitative analysis of the correlation between the two signals, since many confounding variables at both the fibrillar and bundle level are expected to affect both signals, but this remains a future aim of the continued work.

With respect to clinical translation, the potential is vast. Diseases such as keratoconus and ectasia are associated with structural and biomechanical changes but are currently not fully understood and are diagnosed at a late stage due to the lack of diagnostic tools capable of clinical evaluation of structure and biomechanics. In vivo SHG imaging of the cornea has already been demonstrated to be achievable, and with continued advancements such as those presented in this study, widespread clinical translation may become possible. In the shorter term, SHG can be employed ex vivo to understand the structural and biomechanical changes that occur during common ocular surgeries, including refractive procedures, to optimize their application and safety profile. Although further work remains in understanding the correlation between forward and backward signals in full, this research represents an important step forward in the interpretation of backward SHG and the association between specific structural features and their representation in the image.

Subject of Research: Biophotonics

Article Title: Correlation between forward and backward generated SHG images of biological tissue relating to collagen structure and biomechanics

Article References: Boutkhil, M., Moulding, D., Sawyer, T. W., & Wilson, A. (2026). Correlation between forward and backward generated SHG images of biological tissue relating to collagen structure and biomechanics. Frontiers in Photonics, 7, Article 1771907. https://doi.org/10.3389/fphot.2026.1771907

Image Credits: AI Generated

DOI: 10.3389/fphot.2026.1771907

Keywords: SHG microscopy, collagen, cornea, biomechanics, collagen crimp, nonlinear optics, biophotonics, Correlation, between, forward, backward, generated

The review’s central argument is that venous valve biomechanics operates as a strict dependency pipeline, and that a weakness at any stage corrupts everything downstream. Biaxial mechanical testing feeds extracellular matrix characterization; matrix architecture explains mechanical anisotropy; anisotropy determines which constitutive model form is appropriate; and model parameters set the material boundary conditions for fluid-structure interaction, or FSI, simulation. Before 2017, not a single experimentally measured material parameter for venous valve tissue existed. Every computational model of venous valve function published up to that point borrowed linear isotropic stiffness values from arterial tissue or material constants derived from aortic or mitral valve leaflets, structures whose extracellular matrix architecture and loading environments differ from venous valves in ways that no simple scaling factor can correct. The review quantifies just how wrong those borrowed numbers were: fitted constitutive coefficients for venous valve tissue differ from cardiac valve values by factors of 5 to 100 across parameters, meaning the parameter spaces do not even overlap in a way that permits extrapolation.

The experimental foundation of the review rests on biaxial testing, a technique that applies simultaneous tensile loads along two orthogonal axes and captures the coupled mechanical response that uniaxial testing cannot resolve. Venous valve leaflets experience exactly this kind of biaxial loading in vivo: they billow open during forward flow and coapt under retrograde pressure, with the ratio of circumferential to radial stress varying continuously through the cardiac-phasic venous pressure cycle. The only uniaxial study of human tissue, performed in 1985 on femoral vein valve leaflets, established a breaking stress of roughly 11.5 megapascals, about twice that of the adjacent sinus wall, but could say nothing about the coupled deformation behavior that governs leaflet kinematics. Kaul and Huang’s group applied force-controlled biaxial protocols at multiple circumferential-to-radial loading ratios to bovine jugular and saphenous valve leaflets, producing the only systematically characterized nonlinear anisotropic material parameters for venous tissue published to date.

The results revealed a tissue that is strongly anisotropic and highly nonlinear. Jugular valve leaflets reach a peak circumferential tangent modulus of 27 plus or minus 9 megapascals under equibiaxial stretching, versus 9 plus or minus 6 megapascals radially, an anisotropy ratio of approximately 3 to 1 driven by preferential circumferential alignment of collagen fibers. Saphenous leaflets, the tissue directly implicated in lower-limb CVI, are stiffer still, with circumferential peak tangent modulus rising from 62.7 plus or minus 34.9 megapascals proximally to 92.1 plus or minus 27.9 megapascals distally. Strikingly, the two valve types show opposite proximal-to-distal stiffness gradients that track opposing collagen concentration gradients, a crossover that establishes jugular and saphenous valves as mechanically distinct tissues rather than interchangeable surrogates. For comparison, aortic valve leaflets exhibit circumferential moduli of 10 to 15 megapascals with an anisotropy ratio near 10 to 1, ruling out direct cross-tissue material substitution. Intra-valve variability of 30 to 40 percent in tangent modulus further means that single-leaflet sampling systematically underestimates the population-level mechanical range.

Why does venous valve tissue behave this way? The answer lies in its extracellular matrix architecture, which the review treats not as incidental background but as the mechanistic explanation of mechanical behavior. Imaging reveals a clear bi-layer organization: on the parietal, wall-facing side, collagen fibers align preferentially along the circumferential axis with angular deviations of less than 10 degrees in proximal jugular leaflets, resisting the tensile stress that peaks at full valve closure. On the luminal, blood-facing side, a crosslinked radial mesh of elastin dominates, providing the compliance that lets the leaflet billow toward coaptation without tearing. Collagen crimp, the sinusoidal undulation of unstressed fiber bundles, was quantified by second-harmonic generation microscopy at a wavelength of 38.46 plus or minus 8.06 micrometers and an amplitude of 4.51 plus or minus 1.65 micrometers, corresponding to a fiber recruitment threshold of approximately 12 percent true strain. Because physiological circumferential strains during valve closure are estimated at 10 to 20 percent, a substantial fraction of the leaflet’s working range sits precisely within this crimp-recruitment transition, the region where matrix degradation most severely compromises function.

This architecture also explains how disease destroys the valve. Elevated venous pressure triggers an inflammatory cascade that upregulates matrix metalloproteinases, particularly MMP-2 and MMP-9, which selectively degrade fibrillar collagen and elastin. CVI specimens show collagen disorganization and loss of the preferential circumferential alignment that generates the healthy 3-to-1 anisotropy ratio. The consequence is not merely reduced stiffness: the leaflet converts from an anisotropic load-bearing structure into a mechanically isotropic, globally compliant sheet, impairing the directed coaptation force needed for competent closure independent of any change in total collagen content. The authors are careful to note that this interpretation should be read as a plausible proximate mechanical contributor rather than a definitively established causal chain, since direct evidence in human venous valve tissue remains limited, and reported elastin changes in CVI specimens remain contested, likely reflecting differences in disease stage and assay methodology.

On the modeling side, the review identifies a critical failure mode in current constitutive models. The May-Newman and Yin exponential form, fitted to jugular venous data with R-squared values of 0.90 to 0.97 under equibiaxial and near-equibiaxial loading, breaks down systematically under asymmetric conditions. At a 2-to-1 circumferential-to-radial force ratio, the model overpredicts circumferential stiffness and underpredicts radial stiffness, an off-axis prediction error that does not diminish with additional parameter fitting. The mechanism traces to collagen fiber reorientation under off-axis loading, which exponential formulations cannot capture because they lack a structural coupling term to redistribute load across fiber families. The error is most consequential precisely where it matters most: the 2-to-1 ratio at which the model fails coincides with the loading state that dominates physiological valve closure at peak hemodynamic load. The Holzapfel-Gasser-Ogden framework, with its angular fiber dispersion parameter, and the structural constitutive approach of Sacks, which integrates measured fiber orientation distributions directly into the strain energy function, offer principled resolutions, but neither has yet been applied to venous valve tissue.

Three unresolved gaps emerge from the synthesis, each experimentally tractable with existing techniques. First, the off-axis prediction error requires expanded biaxial testing at asymmetric ratios and, ideally, co-registered confocal imaging and mechanical testing on the same specimens, a pairing not yet achieved for venous valves. Second, no failure criterion or damage evolution law exists for venous valve tissue; all biaxial studies to date have operated below the failure threshold, conservatively estimated at 60 to 70 percent true strain, leaving bioprosthetic durability computationally unassessable. Third, no validated material parameters have been published for bioprosthetic candidate materials such as decellularized pericardium, electrospun polyurethane, or silk fibroin scaffolds, even though tissue-engineered venous valves have reached preclinical testing in ovine and porcine models. An isotropic scaffold, the review argues, is mechanically unsuitable as a native surrogate regardless of geometric optimization, because the anisotropy mismatch alone produces divergent closure kinematics and shear distributions.

Perhaps the most sobering caveat is one of species translation: every biaxial dataset currently available derives from bovine tissue, while the clinical target is human. Studies of aortic valves have shown substantial interspecies divergence across humans, pigs, and bovines, and there is no principled reason to assume venous valves behave differently. Human venous valve biaxial characterization remains, in the authors’ framing, the single most consequential missing experiment in the pipeline. The review also flags uncharacterized residual stress, untested viscoelastic effects at physiological loading rates, and an identifiability problem in which multiple parameter sets fit equibiaxial data equally well yet diverge sharply under off-axis prediction. The path forward, the authors conclude, requires no new theoretical frameworks, only disciplined application of established biaxial testing and structural imaging methods to larger and more anatomically diverse specimen sets, ultimately including human tissue. Until then, patient-specific simulation of venous valve function, however sophisticated the software, remains built on a foundation that is only now beginning to be measured.

Subject of Research: Biomechanics, extracellular matrix architecture, and constitutive modeling of venous valve tissue in chronic venous insufficiency

Article Title: From Tissue to Simulation: A Review of Venous Valve Biomechanics, Extracellular Matrix Architecture, and Constitutive Modeling in Chronic Venous Insufficiency

Article References: Kaul, N., & Huang, H.-Y. S. (2026). From Tissue to Simulation: A Review of Venous Valve Biomechanics, Extracellular Matrix Architecture, and Constitutive Modeling in Chronic Venous Insufficiency. Annals of Biomedical Engineering. https://doi.org/10.1007/s10439-026-04412-2

Image Credits: AI Generated

DOI: 10.1007/s10439-026-04412-2

Keywords: chronic venous insufficiency, venous valves, biaxial mechanical testing, extracellular matrix, collagen crimp, elastin, constitutive modeling, fluid-structure interaction, soft tissue biomechanics, tissue-engineered venous valves, matrix metalloproteinases, hyperelasticity