The research, led by Azin Mirzajavadkhan, Christian Chum Him Lai, Margarete K. Akens, Cari M. Whyne, and Chenxi Qian at the University of Toronto and Sunnybrook Research Institute, focused on mixed femoral metastases, a model of the osteolytic and osteoblastic lesions that complicate advanced prostate, breast, and lung cancers. Mixed lesions are particularly treacherous because they combine bone destruction and bone formation within the same anatomical site, producing tissue whose mechanical quality varies dramatically over distances of less than a millimeter. Standard analytical tools, such as bulk Raman spectroscopy or micro-computed tomography, average over those gradients or measure structure without chemistry, leaving the local composition of tumor-adjacent and tumor-distant bone largely unexplored.

Stimulated Raman scattering offers a way around that limitation. Unlike spontaneous Raman spectroscopy, which collects weak inelastic scattering from a laser spot and can require long acquisition times, stimulated Raman scattering uses two synchronized lasers to excite molecular vibrations directly, amplifying the signal by many orders of magnitude. The result is rapid, label-free imaging of specific chemical bonds: amide I and amide III vibrations report on the protein matrix of bone, proline signals trace collagen deposition, phenylalanine marks aromatic amino acid content, and phosphate bands quantify the mineral phase. Because the technique requires no stains, dyes, or decalcification, the same tissue can be interrogated in a near-native state.

To test the approach, the team induced mixed metastases in five- to six-week-old athymic male rats by intracardiac injection of ACE-1 canine prostate cancer cells, a well-characterized preclinical model of osteoblastic-dominant mixed lesions. Twenty-one days after inoculation, the researchers harvested both femora. Sagittal sections of the right femora were processed for histological analysis and stimulated Raman imaging of the organic phase, while undecalcified sections preserved the mineral phase for separate assessment. Regions of interest were then selected within newly formed osteoblastic bone, within bone distant from the tumor, and in distinct samples containing only healthy bone, allowing direct comparison of chemistry across the metastatic landscape.

The measurements revealed a striking biochemical signature in the osteoblastic regions. Compared with healthy bone, the newly formed tumor-associated tissue showed increased signal intensity for amide I, amide III, and proline, indicating enhanced collagen deposition, alongside reduced phosphate and phenylalanine signals, reflecting diminished mineral content and altered protein composition. In practical terms, the tumor-driven bone formation produced a matrix that was collagen-rich but under-mineralized, a combination known to compromise stiffness and fracture resistance. Because mineral-to-matrix ratio is one of the strongest Raman predictors of bone mechanical quality, the finding suggests that even bone that appears to be growing in response to cancer may be structurally inferior to healthy tissue.

Perhaps the more provocative result came from the bone far from the tumor. In regions with no direct tumor involvement, the researchers measured elevated proline and reduced phenylalanine signal intensity, a pattern consistent with systemic perturbation of collagen composition. This supports the study’s central hypothesis: that interactions between tumor and bone influence bone quality beyond the areas of direct tumor involvement. If cancer cells can reprogram the chemistry of bone they have not yet colonized, the implications for fracture risk assessment are significant, because current clinical imaging evaluates lesion geometry rather than the biochemical integrity of the surrounding skeleton.

The findings build on a growing body of Raman-based bone research. Earlier work by several of the same investigators showed that osteolytic and mixed metastases modulate collagen and mineral parameters in rat vertebral bone, and independent studies in mouse models of prostate and breast cancer metastasis demonstrated that tumors alter bone mineral and matrix composition independently of effects on bone architecture. What the new study adds is spatially resolved, quantitative chemical mapping of the organic and mineral phases in mixed femoral lesions, achieved non-destructively and without the exogenous labels that can perturb tissue chemistry.

The technical achievement rests on the maturity of stimulated Raman scattering microscopy itself, which has matured rapidly over the past decade as a biological imaging platform. Recent applications include hyperspectral imaging of microcalcifications in fresh breast biopsies and high-content stimulated Raman histology of human breast cancer, where the technique distinguishes tumor from stroma without section staining. Applying the method to bone required careful handling of the mineral phase, since undecalcified sections preserve hydroxyapatite but are harder to prepare, and the organic phase signals must be disentangled from the strong phosphate background. The Toronto team’s dual-section strategy, using decalcified sections for organic phase imaging and undecalcified sections for mineral assessment, provided a practical route to characterizing both components of the bone matrix.

For clinicians, the study points toward a future where bone quality in metastatic disease could be assessed chemically rather than merely structurally. Skeletal metastases are a leading cause of morbidity in advanced cancer, producing pain, fracture, and spinal cord compression, and treatment decisions, from bisphosphonates to stereotactic body radiation therapy to surgical stabilization, depend on judgments about how much load the compromised bone can bear. A label-free imaging modality that quantifies collagen deposition, mineral content, and protein composition at the microscopic level could eventually inform those decisions, identifying regions of bone that look intact on a CT scan but are biochemically weakened, or vice versa.

The authors caution that the work is preclinical, conducted in a rat model with approval from the institutional animal care committee at University Health Network in Toronto, and that translation to human diagnostics will require further validation. Data supporting the findings are stored at Sunnybrook Research Institute and may be available from the corresponding author under reasonable request. Still, the study establishes stimulated Raman scattering microscopy as a non-destructive tool for detecting spatially resolved biochemical alterations in metastatic bone, and it demonstrates that the chemical footprint of cancer extends beyond the visible margins of the tumor. As the technique matures, the patchwork of metastatic bone, long invisible to conventional imaging, may finally come into focus.

Subject of Research: Stimulated Raman scattering spectromicroscopy reveals biochemical heterogeneity of bone quality in mixed femoral metastases.

Article Title: Bone Quality Heterogeneity in Mixed Femoral Metastases Revealed by Stimulated Raman Scattering Spectromicroscopy

Article References: Mirzajavadkhan, A., Lai, C. C. H., Akens, M. K., Whyne, C. M., & Qian, C. (2026). Bone Quality Heterogeneity in Mixed Femoral Metastases Revealed by Stimulated Raman Scattering Spectromicroscopy. Annals of Biomedical Engineering. https://doi.org/10.1007/s10439-026-04433-x

Image Credits: AI Generated

DOI: 10.1007/s10439-026-04433-x

Keywords: bone metastases, stimulated Raman scattering, bone quality, osteoblastic lesions, collagen, bone mineralization, femur, prostate cancer, Raman spectroscopy, cancer biology, biomedical imaging, Bone

Letian Dou, senior author of the paper and professor of chemistry at Emory University, where he joined the faculty last year as the Paul and Phyllis Fireman Professor of Chemistry after a career at Purdue University, frames the advance in simple terms. Most previous work on perovskite LEDs has yielded quantum efficiency, the measure of how well a device converts electricity into light, within a range of 20 to 30 percent. The new approach brings that figure closer to 40 percent, and the team reports a peak of 37 percent. For a technology whose commercial promise has been constrained by exactly this kind of performance ceiling, the jump is significant.

The discovery began as a laboratory mystery. Sung-Doo Baek, first author of the paper, who carried out the work as a postdoctoral fellow in the Dou lab, noticed something odd while producing perovskite LEDs for experimentation and testing. The basic formulation for making these devices includes an additive, first reported in 2018, that delivers a baseline efficiency of 20 to 25 percent. Yet a batch made in the Dou lab somehow reached around 30 percent. Something in the routine of the lab, rather than the recipe itself, appeared to be responsible.

Baek’s first hypothesis pointed toward the nitrogen glove box, the sealed, clear-walled enclosure filled with purified nitrogen gas in which the ultrathin films are spin-coated onto a substrate, protecting the material from oxygen and moisture while a researcher works through gloves attached to the box. Although oxygen is generally considered harmful to perovskite materials, a handful of studies had suggested it could occasionally be beneficial, so a slow leak seemed a plausible culprit. That turned out to be a false trail, Baek says. Instead, he turned his attention to the hidden chemistry of the solution itself, a stage of the process that, as he points out, no one knew what was happening during.

To make a perovskite LED, metal halide salts along with various additives are heated so that they dissolve into a solvent. The solution is then spin-coated in ultrathin layers onto a substrate, and the resulting film is heated to evaporate the solvent and allow the chemicals to crystallize into the perovskite structure. The team kept tweaking the process of making the solution, producing the LEDs and comparing their efficiencies. After months of painstaking experimentation, a pattern emerged: differences in how long the solutions were heated led to variations in efficiency, and twenty minutes of heating turned out to be the sweet spot.

Comparative analyses then revealed how the chemistry transforms during the solution phase. Optimal heating favors the formation of a linear compound that boosts the efficiency of the LED while simultaneously suppressing the formation of a ring compound that harms it. The researchers went further, synthesizing and purifying the linear compound and developing a method of targeted reaction-pathway control to selectively harness its power. Applying that control raised the efficiency of a perovskite LED to the reported peak of 37 percent. Yu-Ting Yang, a PhD candidate in the Dou lab and co-first author of the paper, says what surprised her most is that even a detail that appears trivial, one you think isn’t really important, can have such a major effect, a lesson she describes as learning the importance of paying close attention.

The advance matters because of where perovskite LEDs sit in the broader landscape of lighting and display technology. Starting in the 2010s, LEDs made from inorganic semiconductors such as gallium nitride began largely replacing incandescent lightbulbs in households. Incandescent bulbs work by heating a metal filament and waste roughly 90 percent of their energy, whereas an inorganic LED converts energy directly into light by passing electric current through a semiconductor to excite electrons, using only 8 to 10 watts to produce the same amount of light as a 60-watt incandescent bulb. Those inorganic devices are efficient, stable and long lasting, but they can be difficult to assemble into large-area displays or to shrink into extremely small pixels, and their rigidity limits their use in flexible devices.

LEDs made with organic molecules, by contrast, are thinner, more flexible and more designable, and they are known for producing brilliant colors in displays, but they remain more expensive to produce and less efficient and durable than their inorganic counterparts. Hybrid LEDs built from metal-halide perovskite, a material that combines organic and inorganic molecules, aim to capture the best of both worlds. The Dou lab focuses on developing lightweight, flexible, low-cost, highly durable hybrid LEDs that are efficient in the near-infrared region, the light found just past what is visible to the human eye.

That near-infrared focus connects the work to a set of applications with real commercial and scientific weight. Dou notes that the facial recognition feature on a cell phone relies on infrared light, which avoids shining a beam of blue light into a user’s eye, and that the same region of the spectrum is used in night-vision cameras, such as those used to monitor a sleeping baby. In biomedicine, the longer wavelengths of infrared light can provide clear, non-invasive imaging of tissues for applications ranging from tracking immune responses to detecting abnormal tissue cells. Improving the efficiency of the emitters that produce this light is a step toward moving perovskite LEDs closer to their potential for commercial use.

The team, whose co-authors include scientists from Emory, Purdue University, the University of Michigan at Ann Arbor and Princeton University, says the work sets the stage for a new chemistry-driven strategy for additive design to drive further improvements. Dou says the lab is working to keep improving both the efficiency and the operational lifetime of perovskite LEDs, and also wants to move toward applications by demonstrating the relevance of the technology in biomedical research, with hopes of forming collaborations within Emory School of Medicine. The research was funded by the U.S. National Science Foundation and the U.S. Department of Energy, and the findings appear in Nature under the title describing how hidden reaction pathways inform additive engineering for perovskite LEDs.

Subject of Research: Reaction-pathway control in solution-processed metal-halide perovskite light-emitting diodes

Article Title: A lightbulb moment: Chemists find way to boost efficiency of hybrid LEDs

Article References: A lightbulb moment: Chemists find way to boost efficiency of hybrid LEDs. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: perovskite LEDs, quantum efficiency, metal-halide perovskite, near-infrared light, additive engineering, reaction pathways, solution processing, biomedical imaging, facial recognition, optoelectronics, Emory University, Nature journal

Hydroxyapatite, with the chemical formula Ca10(PO4)6(OH)2 and a calcium-to-phosphorus ratio of 1.67, is the inorganic scaffold of human bone and dental enamel. Because it is biocompatible, non-toxic, and non-inflammatory, and because its porous structure offers a large surface area, materials scientists have long viewed it as an ideal candidate for bone regeneration and for controlled drug delivery. An efficient delivery system, however, needs more than biocompatibility. It must load therapeutic molecules, transport them to target tissues, and release them predictably, ideally in response to physiological cues such as the acidic pH found in tumor environments, where hydroxyapatite dissolves more readily and can dump its cargo on cue.

The Mexican team added a second function to this already versatile material: light. Europium ions, Eu3+, belong to the rare-earth family and emit sharp, photostable red luminescence when excited, a property that makes them valuable for cellular imaging and diagnostic applications. The catch is that europium can be toxic at high concentrations, and too much of it in a crystal lattice causes luminescence quenching, an energy-draining conversation between neighboring dopant ions that snuffs out the very glow the dopant was added to provide. The researchers therefore kept europium levels low, testing 2, 3, and 4 molar percent relative to calcium, and relied on the nanofiber geometry to place the emitting ions near the surface, where their light is easier to harvest.

The synthesis itself is a study in efficiency. Using a microwave-assisted hydrothermal method, the team combined calcium nitrate and glutamic acid, an amino acid that steers crystal growth, with potassium phosphate and potassium hydroxide solutions. Microwave irradiation heats the reaction mixture rapidly and uniformly, and sealed vials were reacted at 170 degrees Celsius for just 45 minutes at pressures of 10 to 11 bar. For the doped samples, a third solution of europium nitrate was added at carefully calculated stoichiometries. Each synthesis yielded roughly 0.11 to 0.17 grams of product, and the researchers ran three distinct methodologies, varying precursor temperatures, pH, and the way europium entered the recipe, to map how each parameter shaped the final material.

X-ray diffraction revealed that every sample crystallized predominantly in the hexagonal hydroxyapatite phase, with sharp Bragg reflections indicating high crystallinity. More intriguing was a systematic anomaly: the standard reference pattern predicts that the (300) reflection should reach only about 62 percent of the intensity of the dominant (211) peak, yet in these samples the (300) reflection frequently matched or exceeded it. This inversion is the fingerprint of preferential crystalline orientation, a consequence of the glutamic acid and microwave method working together. Because the (100) and (300) planes that terminate the nanofibers are rich in calcium ions, they are prime real estate for europium substitution, positioning the dopant where its emission can escape the crystal most easily.

Crystallite size and lattice strain told a subtler story. Using both the Scherrer equation and the Williamson-Hall method, which accounts for strain broadening that Scherrer ignores, the team found that Methodology 1 produced the largest, most pristine crystallites, around 280 angstroms, with microstrain below 0.02 percent in doped samples. Methodology 2, with its pH manipulation, yielded smaller crystallites as low as 152 angstroms and the highest microstrain measured, 0.186 percent, signaling dense defects and lattice distortion. Methodology 3 struck a balance, and its samples grew into strikingly well-formed hexagonal nanofibers hundreds of nanometers long, with sides of about 205 nanometers and long diagonals of 390 nanometers. Electron microscopy confirmed the trend: uniform smooth fibers in Methodology 1, fragmented disordered particles in Methodology 2, and large, aligned, well-faceted fibers in Methodology 3.

Dopant concentration sculpted morphology in unexpected ways. At 2 percent europium, the fibrous habit survived intact, suggesting the ions slipped neatly into calcium sites. At 3 percent, fibers thickened and agglomerated, with irregular particles and partially fused regions appearing as the mismatch between the larger Eu3+ radius and Ca2+ disrupted the balance between nucleation and growth. At 4 percent, however, the material rebounded, producing longer, thicker, more aligned fibers, apparently because excess europium promoted directional growth along the crystallographic c-axis, though localized agglomerates hinted at europium-rich amorphous pockets. Photoluminescence spectroscopy added another layer: the splitting pattern of the 5D0 to 7F1 transition showed more lines than theory predicts for a single site, indicating that europium occupies both calcium site types, and possibly interstitial positions, simultaneously.

With 3 percent europium selected as the optimal compromise, preserving the apatite phase, fibrous morphology, and detectable red emission, the team turned to quercetin. Three formulations were prepared by stirring 0.100 grams of the doped nanofibers with 0.005, 0.010, or 0.020 grams of quercetin in ethanol for 24 hours at 30 degrees Celsius, corresponding to weight ratios of 10:0.5, 10:1, and 10:2. After centrifugation, washing, and drying, electron microscopy showed the fibrous network largely preserved but decorated with small granular deposits absent from unloaded samples, evidence that quercetin adsorbs onto the fiber surfaces, likely through hydrogen bonding and electrostatic interactions with the calcium phosphate, rather than infiltrating pores. Energy-dispersive spectroscopy detected carbon consistent with the organic molecule, though its uneven distribution across analyzed regions suggested the adsorption is patchy rather than uniform.

Infrared spectroscopy was less conclusive. Because hydroxyapatite phosphate bands overlap with quercetin’s organic signals and the loaded amounts were small, the spectra showed only subtle changes: a phenolic OH signal near 1200 wavenumbers, aromatic CH stretches around 2959 to 2855, and weak carbon-carbon double-bond bands. The absence of new bands suggests the interaction is physical, mediated by hydrogen bonds, with no chemical reaction. Crucially, loading quercetin did not shift the positions of the characteristic europium emission transitions, meaning the optical tag survives the drug-loading process. That combination, a bone-like scaffold that glows red, carries a bioactive molecule on its surface, and keeps its luminescence intact, is precisely what a multifunctional theragnostic platform would need.

The authors are careful about limits. The physicochemical data represent single measurements without statistical replicates, the EDS technique is localized and semiquantitative, and nothing in the study establishes biological safety; cytocompatibility, ion-release, and cellular uptake experiments remain to be done before any clinical ambition can be entertained. Quantitative adsorption capacity, release kinetics, and reproducibility also await measurement. But as a proof of concept, the work demonstrates that a microwave-assisted, amino-acid-guided synthesis can deliver a luminescent calcium phosphate nanocarrier in under an hour, with morphology and dopant behavior controlled by recipe rather than chance. If the follow-up biology holds, the same mineral that builds skeletons may one day light the way for medicines traveling through the body.

Subject of Research: Europium-doped hydroxyapatite nanostructures for luminescent imaging and quercetin drug delivery

Article Title: Physicochemical and photoluminescent properties of Eu 3+ -doped hydroxyapatite nanostructures and preliminary quercetin adsorption study

Article References: Lazaro-Vidal, C., Méndez-Lozano, N., Alanis-Gómez, J. R., Hernández-Rosas, F., Rivera-Muñoz, E., & Velazquez-Castillo, R. (2026). Physicochemical and photoluminescent properties of Eu3+-doped hydroxyapatite nanostructures and preliminary quercetin adsorption study. Results in Chemistry, 31, Article 103956. https://doi.org/10.1016/j.rechem.2026.103956

Image Credits: AI Generated

DOI: 10.1016/j.rechem.2026.103956

Keywords: hydroxyapatite, europium doping, nanofibers, photoluminescence, quercetin, drug delivery, microwave-assisted hydrothermal synthesis, X-ray diffraction, biomedical imaging, calcium phosphate, nanomaterials, theragnostics

The scaffolds at the heart of the study are made of hydroxyapatite, a mineral chemically similar to natural bone, arranged as an orthogonal lattice of rods with pores between 100 and 400 micrometers wide. Implanted into mandibular bone defects in Yucatan minipigs, these structures invite new bone to grow into their macropores, and quantifying how much bone has infiltrated each scaffold is the key to comparing different lattice designs and biological additives. The problem is that bone and hydroxyapatite share nearly identical mineral composition, so their X-ray attenuation values overlap almost completely in micro-CT scans. Ingrown bone also overlaps in brightness with the surrounding soft tissue and noise. Manual labeling of the roughly 1,000 images per scaffold is prohibitively time consuming across studies involving tens of scaffolds, which is exactly where the U-Net convolutional neural network enters the picture.

The U-Net architecture, first introduced in 2015, remains a favorite in biomedical imaging because it delivers strong segmentation from small training sets. Its U-shaped design works like an autoencoder with a memory: an encoding path progressively shrinks the image while doubling its feature channels to detect patterns, a bottleneck connects the two halves, and a decoding path restores the original dimensions, with skip connections threading fine spatial detail from encoder to decoder at every level. The Illinois team used a four-level network starting with eight feature channels and reaching 128 at the bottleneck, ultimately classifying every pixel in a 512-by-512 patch as scaffold, bone, or background. Rather than inventing a new architecture, the researchers held the network constant and varied only the loss function, the mathematical rule that tells the model how wrong its predictions are during training.

Three loss functions were pitted against each other under identical conditions. Cross-entropy loss, the workhorse of classification, evaluates every pixel individually regardless of class balance. Soft Dice loss instead optimizes the overlap between prediction and ground truth, a strategy favored when one class, like bone, occupies only a small fraction of the image. A third, less common option called Accuracy loss directly rewards correct classification of the entire patch. The team trained each model for 1,000 epochs on 225 patches derived from 31 manually labeled images, using four NVIDIA V100 GPUs and confirming that all three losses converged without overfitting. A second expert annotator independently relabeled a subset to gauge human variability, finding a bone Dice score of 0.78 between annotators and an average difference in bone growth fraction of just 1.2 percentage points.

Then came the surprise. Standard evaluation metrics, the numbers researchers typically report to demonstrate that a segmentation model works, produced results that seemed catastrophic but were actually meaningless. Some patches scored a Dice coefficient of zero and a Precision of zero, values usually interpreted as total failure, yet visual inspection showed those patches were segmented perfectly well. The catch: those patches contained no bone in the ground truth at all. When the model predicted a handful of spurious bone pixels in a bone-free image, the Dice and Precision formulas collapsed to zero even though the biological error was negligible. Similar distortions plagued Recall, which tanked on patches where a few missed pixels represented a large fraction of a tiny bone area. Under extreme class imbalance, the familiar metrics were measuring the imbalance, not the model.

To escape this trap, the researchers invented an application-specific endpoint they call delta BGF: simply the difference between the bone growth fraction predicted by the model and the bone growth fraction in the ground truth. This is the number that actually matters for the science, because the entire purpose of segmenting the scans is to quantify bone regeneration and compare treatment groups. Bland-Altman analysis showed the systematic bias between predicted and true bone growth fraction was less than 1.5 percentage points for all three models, comparable to the disagreement between two human experts. When the team correlated the standard metrics against the absolute value of delta BGF, Accuracy emerged as the clear winner, with correlation coefficients reaching negative 0.8 for the Accuracy-loss model and negative 0.78 for the Dice model. A Steiger test confirmed Accuracy’s correlation was significantly stronger than those of Dice, Precision, and Recall in eight of nine comparisons, and bootstrap resampling across 10,000 samples showed the relationship was stable, with confidence intervals lying entirely below zero.

The loss function also shaped where the errors landed, which matters because different regions of the image carry different scientific weight. The Accuracy-loss model tended to overestimate bone, scattering false positives inside the scaffold boundary, precisely where the ingrown bone quantification happens. The cross-entropy and Dice models instead missed bone in the dense ring surrounding the scaffold, errors that are conveniently located outside the region of interest and can be corrected automatically, since scaffold pixels cannot legitimately exist beyond the scaffold boundary. Given this geography of mistakes, the cross-entropy model emerged as the recommended choice for future in vivo studies quantifying spatial variation in bone growth, combining clean class edges within the scaffold with correctable peripheral errors.

Robustness testing added another layer of confidence. The team created augmented test sets by rotating patches 15, 30, and 45 degrees, shifting contrast in both directions, and degrading resolution by factors of 0.66 and 0.5, simulating the natural variation in micro-CT scanning parameters and scaffold orientation. None of these perturbations produced a statistically significant change in Accuracy, Dice, Precision, Recall, or delta BGF. The systematic bias remained below 2.6 percentage points across all augmentations, with contrast changes producing the widest limits of agreement and rotation the narrowest, suggesting that intensity shifts stress the model more than orientation changes. For a field where scanning protocols differ between labs and even between experiments, this resilience means the trained models can plausibly be applied to new datasets without retraining from scratch.

Perhaps the most consequential result is biological rather than computational. Comparing against the group’s earlier in vivo studies, the smallest treatment differences those studies could detect were on the order of 8 to 20 percentage points in bone fraction, vastly larger than the 1.5 percentage point error of the AI segmentation. In other words, the model’s mistakes are too small to obscure genuine biological differences between scaffold designs, meaning future comparisons of bone regeneration should remain statistically sound. The study’s limitations are acknowledged candidly: the 54 test patches derive from only six images, so patch-level statistics treat clustered data as independent, and the authors frame their correlation findings as descriptive rather than confirmatory. Still, the message radiates well beyond bone scaffolds. Whenever deep learning is deployed to extract a quantitative measurement rather than a pretty picture, the evaluation metric should be the measurement itself. A model that aces the standard report card can still flunk the science, and a model that appears to fail it may be quietly doing excellent work.

Subject of Research: Evaluating U-Net deep learning segmentation of bone ingrowth in micro-CT images of hydroxyapatite scaffolds using loss functions and application-specific metrics

Article Title: Segmenting scaffolds with ingrown bone in micro-CT images: considerations of evaluation metrics and loss functions

Article References: Aboutaleb, S., Haug, N., Keni-McCray, P., McCray, A. R. C., Phillips, H., Sharping, S., Cohen, D., Norato, J., & Wagoner Johnson, A. (2026). Segmenting scaffolds with ingrown bone in micro-CT images: considerations of evaluation metrics and loss functions. Medical & Biological Engineering & Computing. https://doi.org/10.1007/s11517-026-03686-x

Image Credits: AI Generated

DOI: 10.1007/s11517-026-03686-x

Keywords: U-Net, semantic segmentation, micro-CT, bone regeneration, hydroxyapatite scaffold, loss functions, evaluation metrics, Dice coefficient, cross-entropy loss, Bland-Altman analysis, deep learning, biomedical imaging

The review organizes the field into eight distinct material families, each with its own physical principles, synthetic chemistry, and diagnostic strengths. These are quantum dots, gold nanostructures, carbon-based nanomaterials, silica-based nanoparticles, superparamagnetic iron oxide nanoparticles, aggregation-induced emission nanoparticles, nanobubbles, and polymeric nanomaterials. The classification is more than a cataloging exercise. By deconstructing the structure-property paradigms of each family, the authors establish design rules that connect the atomic-scale architecture of a nanoparticle to its observable performance inside a living body, from the wavelength of light it emits to the way it navigates the turbulent microenvironment of a tumor or an inflamed tissue.

Quantum dots remain among the most celebrated of these platforms. These semiconductor nanocrystals, typically only a few nanometers across, absorb light over broad spectral ranges and re-emit it at sharply defined wavelengths determined by their physical size. That size-tunable fluorescence, combined with exceptional brightness and resistance to photobleaching, allows researchers to track biological processes over long observation periods with single-particle sensitivity. The review emphasizes that modern synthetic methodologies now permit precise control over quantum dot composition, shell architecture, and surface chemistry, which in turn governs their optical quantum yield, colloidal stability, and crucially their biological compatibility. Surface functionalization strategies, such as coating the inorganic core with biocompatible ligands or amphiphilic polymers, transform otherwise toxic crystalline materials into probes that can circulate through the bloodstream and accumulate selectively at disease sites.

Gold nanomaterials occupy a complementary niche in the imaging arsenal. Depending on their shape, whether spherical nanoparticles, rods, shells, or stars, gold nanostructures support localized surface plasmon resonances, collective oscillations of conduction electrons that produce intense optical scattering and absorption. This property makes them powerful contrast agents for optical coherence tomography, photoacoustic imaging, and computed tomography, while their chemical inertness and ease of surface modification with thiolated molecules give synthetic chemists an unusually versatile handle. The authors highlight how the geometry of a gold nanoparticle can be tuned during synthesis to shift its plasmon resonance deep into the near-infrared window, the spectral region where biological tissue is most transparent, enabling imaging at centimeter depths with millimeter-scale resolution.

Carbon-based nanomaterials, including carbon dots, graphene quantum dots, and carbon nanotubes, bring a different set of advantages to the table. Derived from abundant and often biologically benign precursors, carbon dots exhibit strong fluorescence, low toxicity, and rich surface chemistry that supports straightforward conjugation to targeting ligands, drugs, or responsive groups. Their small hydrodynamic diameters promote efficient renal clearance, a biosafety attribute of growing importance as regulators scrutinize the long-term fate of engineered nanomaterials in the body. The review also situates silica-based nanoparticles within this design landscape, noting that mesoporous silica architectures provide exceptionally high surface areas and controllable pore structures that can host imaging agents, protect them from premature degradation, and release them in response to specific pathophysiological cues such as acidic pH or elevated glutathione levels.

Superparamagnetic iron oxide nanoparticles represent the workhorses of magnetic resonance imaging contrast enhancement. Because each nanoparticle behaves as a single magnetic domain, these materials generate strong local field inhomogeneities that shorten the relaxation times of nearby water protons, darkening the corresponding regions in T2-weighted and, for smaller cores, brightening them in T1-weighted magnetic resonance images. The authors stress that synthetic control over core size, crystallinity, and coating thickness directly determines the relaxivity, the quantitative measure of contrast efficiency, and that recent breakthroughs have pushed these agents toward molecular-level sensitivity in preclinical studies of cardiovascular disease, neuroinflammation, and cancer metastasis. Iron oxide also carries the advantage of clinically established chemistry, with several formulations already approved for human use in related applications.

Perhaps the most conceptually intriguing family covered in the review is that of aggregation-induced emission nanoparticles. Conventional organic fluorophores suffer from aggregation-caused quenching, a phenomenon in which densely packed dye molecules lose their luminescence through nonradiative energy dissipation. Aggregation-induced emission materials invert this behavior: they are nearly dark when molecularly dissolved but grow brilliantly luminous when their molecular motions are restricted, as happens when they aggregate into solid nanoparticles or bind to biological structures. This counterintuitive property makes them exceptionally resistant to photobleaching, ideal for long-term cell tracking, and highly effective as bright solid-state probes for fluorescence imaging deep within tissues. The review traces how synthetic strategies, including the incorporation of tetraphenylethylene and related propeller-shaped building blocks, have produced AIE nanoparticles with emission spanning the visible and near-infrared spectra and with built-in responsiveness to reactive oxygen species, enzymes, and other biomarkers of disease.

The remaining two families extend the multiscale reach of the platform concept in different physical directions. Nanobubbles, gas-filled or gas-generating structures on the scale of hundreds of nanometers, serve as contrast agents for ultrasound imaging, the most widely deployed and least expensive modality in clinical medicine, and their acoustic responsiveness can be harnessed both for diagnosis and for triggered delivery of therapeutic payloads. Polymeric nanomaterials, assembled from biodegradable polymers such as polylactic-co-glycolic acid, polyethylene glycol derivatives, and polypeptides, act as modular carriers that can integrate multiple imaging functions, fluorescence, magnetic resonance, photoacoustic, or positron emission signals, within a single tunable particle whose size, charge, and degradation profile are dictated by polymer design. The authors describe how such theranostic platforms, combining imaging with therapy, embody the spatiotemporal control over nanomaterial performance within pathophysiological microenvironments that the review identifies as a central design goal.

What distinguishes this synthesis is its insistence on connecting chemistry to clinical reality. The authors devote sustained attention to biosafety, examining how particle size, surface charge, protein corona formation, and degradation pathways determine biodistribution, organ accumulation, and clearance. They also confront the challenge of clinical scalability, noting that many spectacular results reported in the literature rely on synthesis routes that are difficult to reproduce at pharmaceutical manufacturing standards or on materials whose long-term toxicology remains incompletely characterized. Standardization of synthesis, rigorous batch-to-batch quality control, and translational toxicology emerge as prerequisites for the field’s next phase, and the review argues that interdisciplinary collaboration among chemists, materials scientists, biologists, and clinicians is the only realistic path from laboratory innovation to clinical practice.

The implications reach well beyond the imaging laboratory. Precision medicine depends on the ability to see disease early, characterize it molecularly, and monitor the response to therapy in real time, and synthetic nanoplatforms are increasingly the enabling technology for all three tasks. By establishing a theoretical foundation for the rational design of next-generation nanomaterials, the review by Lin and colleagues offers researchers a coherent framework for engineering probes whose behavior can be predicted from structure rather than discovered by trial and error. As the authors note, their hope is that this integrated perspective will inspire researchers across related fields and accelerate the translation of nanomaterial innovations into personalized healthcare management, a goal that now appears less like a distant aspiration and more like an engineering problem with a defined roadmap.

Subject of Research: Synthetic nanomaterial platforms for multiscale in vivo biomedical imaging

Article Title: Synthetic nanoplatforms for multiscale imaging in living organisms

Article References: Lin, F., Zhang, H., Zhan, J., Maimaitikelimu, X., Zhang, H., & Wang, H. (2026). Synthetic nanoplatforms for multiscale imaging in living organisms. Advanced Composites and Hybrid Materials. https://doi.org/10.1007/s42114-026-02055-4

Image Credits: AI Generated

DOI: 10.1007/s42114-026-02055-4

Keywords: nanomaterials, biomedical imaging, quantum dots, gold nanoparticles, aggregation-induced emission, iron oxide nanoparticles, nanobubbles, silica nanoparticles, carbon dots, surface functionalization, biosafety, precision medicine

The concept behind the new work is known as infrared-to-visible upconversion. Instead of detecting infrared photons electronically, an upconversion device absorbs them and re-emits their energy at shorter, visible wavelengths, where ordinary silicon sensors and even the naked eye can see it. The trick has been demonstrated before in bulk crystals and optical fibers, but those systems typically demand intense laser pumping, operate only in narrow spectral bands, and offer little spatial control over the conversion process. The result is a technology that has remained largely confined to laboratory demonstrations rather than practical imaging systems. The new study tackles each of these limitations by rethinking the device at the level of nanostructure design.

At the heart of the approach are lanthanide-doped upconversion nanoparticles, most commonly built from a sodium yttrium fluoride host lattice doped with ions such as ytterbium and erbium or ytterbium and thulium. These ions form a cascade: the sensitizer ion, typically ytterbium, absorbs a near-infrared photon around 980 nanometers and transfers that energy stepwise to an activator ion, which accumulates the excitation and finally emits a visible photon. Because the energy levels of lanthanide ions are shielded by outer electron shells, the emission is sharp, stable, and remarkably resistant to photobleaching. The nanoparticles can be synthesized with controlled sizes and shell architectures, and core-shell designs that physically separate dopant ions suppress a major loss channel known as surface quenching, in which excitation energy leaks away at particle surfaces before it can produce light.

On their own, however, these nanoparticles are inefficient. The transitions that lanthanide ions undergo are formally forbidden by quantum-mechanical selection rules, which makes absorption weak, and the stepwise energy transfer process competes with numerous decay pathways. This is where the metasurface enters. A metasurface is a two-dimensional array of engineered nanostructures, often metallic or dielectric pillars and antennas, patterned at a scale smaller than the wavelength of light. By adjusting the geometry, spacing, and material composition of these building blocks, researchers can sculpt how light behaves at the surface: concentrating it into tiny volumes, redirecting it, filtering specific wavelengths, or imposing precise phase shifts across a wavefront. Metasurfaces have already revolutionized flat optics, enabling ultrathin lenses and holograms, and the new work harnesses that same design freedom to supercharge upconversion.

The hybrid screens described in the study integrate the two components so that the metasurface acts as an optical antenna system for the nanoparticles. Resonant modes supported by the metasurface trap incoming infrared light near the surface, dramatically increasing the local electromagnetic field intensity exactly where the nanoparticles sit. Because upconversion is a nonlinear process, in which the emission rate scales steeply with excitation intensity, even a modest field enhancement translates into a large boost in output. The metasurface can also be tuned to match the absorption bands of the sensitizer ions and to extract the emitted visible light efficiently, reducing the losses that would otherwise trap the upconverted photons inside the structure. The researchers report that this combined electromagnetic and photonic engineering yields imaging screens with substantially enhanced brightness and sensitivity compared with films of nanoparticles alone.

What elevates the work from a materials demonstration to an imaging technology is the spatial dimension. Because metasurfaces are patterned with lithographic precision, the hybrid screens can be designed to do more than simply brighten an image. The authors show that the screens can impose controlled phase and amplitude modifications on the upconverted visible light, effectively performing optical processing at the moment of conversion. In one configuration, the screen functions as a direct infrared imager: infrared light from a scene strikes the screen, is converted locally into visible emission, and the resulting visible image can be captured with an ordinary camera or viewed directly. In another configuration, the metasurface patterning enables edge enhancement, a computational imaging operation in which the outlines and boundaries of objects are emphasized, all performed passively in optics without any digital processing.

This ability to merge light conversion with analog optical computation in a single thin film points toward a compelling vision of the future of imaging. Conventional infrared cameras chain together optics, detectors, amplifiers, and processors, each stage adding cost, weight, latency, and power consumption. A hybrid upconversion screen collapses much of that chain into a passive optical element. The infrared image is converted, enhanced, and even pre-processed before a single electron is moved. Such screens could be produced as coatings on standard camera lenses, integrated into smartphone modules, or deployed as large-area viewing panels that make invisible laser beams, thermal signatures, or biomedical fluorescence directly visible to the eye.

The potential applications span an unusually wide range. In night vision, low-cost, uncooled upconversion screens could complement or replace bulky image intensifier tubes, offering a lighter and potentially cheaper alternative for both military and civilian use. In medicine, near-infrared light penetrates tissue more deeply than visible light and scatters less, and upconversion nanoparticles are already explored as imaging probes and as agents for light-triggered therapies; screens that convert scattered near-infrared light into visible images could improve surgical guidance and diagnostics. In telecommunications, silicon photonic circuits and optical fibers operate in the near-infrared, and efficient, fast upconversion could allow infrared signals to be inspected visually or routed with visible-light components. In industrial settings, the screens could reveal hot spots, gas absorption features, or defects that are invisible under ordinary illumination, while in fundamental research they could serve as diagnostic foils for characterizing infrared laser beams and photonic devices.

The authors are candid about the challenges that remain before such devices become routine. Upconversion efficiency, even with metasurface enhancement, still falls short of what high-speed, low-light imaging would demand, and the nonlinear nature of the process means performance degrades at low illumination levels, precisely where night-vision applications matter most. The spectral bandwidth of lanthanide-based conversion is inherently narrow, tied to the discrete energy levels of the ions, so covering the broader infrared spectrum, including the mid-infrared region where thermal imaging lives, will require different material combinations or multi-resonant metasurface designs. Response time is another consideration: the excited-state lifetimes that make lanthanides stable emitters also limit how quickly the screens can follow rapidly changing scenes. Scaling the nanofabrication from centimeter-scale laboratory samples to large, uniform, low-cost panels is an engineering task in its own right.

Nevertheless, the demonstration marks a meaningful step in the convergence of two of nanophotonics’ most productive threads: flat metasurface optics and lanthanide luminescence. By treating the upconversion screen not as a passive phosphor but as an actively engineered optical element, the researchers have shown that the conversion of invisible light into visible images can be made brighter, more controllable, and more functional than previously imagined. If the efficiency and bandwidth gaps can be closed through continued materials and design refinement, hybrid metasurface-nanoparticle screens could reshape how we see the invisible half of the electromagnetic spectrum, turning infrared imaging from a specialized electronic undertaking into something as simple and ubiquitous as a sheet of smart glass.

Subject of Research: Hybrid metasurface–lanthanide nanoparticle screens for enhanced infrared-to-visible upconversion imaging

Article Title: Enhanced infrared-to-visible upconversion imaging via metasurface–lanthanide nanoparticle hybrid screens

Article References: Sefidmooye Azar, N., Parry, M., Qi, X., Lee, C., Lee, W. S. L., Russell, B., Luo, W., de Gille, R. W., Nelson, D., Balendhran, S., Meng, J., Tan, H., Bonin, G. O., Choi, D.-Y., Schuck, P. J., Chan, E. M., Cohen, B. E., Neshev, D. N., & Crozier, K. B. (2026). Enhanced infrared-to-visible upconversion imaging via metasurface–lanthanide nanoparticle hybrid screens. Light: Science & Applications, 15(1), Article 377. https://doi.org/10.1038/s41377-026-02449-5

Image Credits: AI Generated

DOI: 10.1038/s41377-026-02449-5

Keywords: upconversion, metasurfaces, lanthanide nanoparticles, infrared imaging, nanophotonics, night vision, optical computing, core-shell nanoparticles, photoluminescence, flat optics, biomedical imaging, light conversion

The work, led by Svetlana Malinovskaya, professor at the Charles V. Schaefer, Jr. School of Engineering and Science at Stevens, addresses a problem that has long plagued experimentalists working at the frontier of quantum control. “A laser is a device that creates a very narrow, highly directional beam of light,” Malinovskaya explains. “Unlike sunlight or light from regular bulbs or flashlights that scatters in all directions, a laser produces light in which all waves move together in a highly synchronized way allowing the light to be very focused and controlled with remarkable precision.”

When such a light wave reaches a quantum system, its packets of energy, called photons, are absorbed by the atoms and molecules in that system, lifting them into higher-energy states. This is precisely the mechanism by which scientists manipulate quantum behavior. But there is a catch. If the laser field is too intense, a single atom or molecule may interact with several photons at once. These simultaneous interactions, known as multiphoton processes, open additional pathways between the system’s energy levels, scrambling the intended dynamics and rendering the quantum system difficult, sometimes impossible, to predict or control.

“By shining laser light on molecules, we can excite molecular vibrations in a controlled way and learn about molecular properties,” says Malinovskaya. “But when very strong laser fields are used for precise quantum control, they can also trigger unwanted multiphoton processes allowing the molecule to access many different states and pathways, making its behavior much more difficult to predict and control.” In other words, the instrument of control becomes an instrument of chaos, and the very act of measurement or manipulation contaminates the result.

That unpredictability is more than an inconvenience; it is a roadblock for the technologies hoping to capitalize on quantum mechanics. “That’s not what we need, particularly for the precision measurements required in quantum computing or quantum sensing,” Malinovskaya says. “In those systems, every photon counts.” Quantum computers rely on the faithful preparation and manipulation of quantum states to perform calculations, while quantum sensors extract exquisitely faint signals from their environment, and even a small number of stray photon interactions can corrupt a computation or drown a measurement in noise. The ideal solution would use just enough light to steer a quantum system where it needs to go, without any excess energy spilling over into unwanted channels.

In their new study, Malinovskaya and her collaborators propose to do exactly that with what they describe as a “digitized” laser pulse. Their calculations show that a train of twelve short, low-intensity laser pulses can produce the same net effect on a quantum system as one long, intense pulse, but without pushing the atoms or molecules into the undesirable states that plague high-intensity approaches. The idea is conceptually similar to replacing a single powerful hammer blow with a rapid series of precisely timed taps that, together, accomplish the same task with far less collateral disturbance.

“Instead of using one very strong laser pulse, we suggest mimicking its effects with a carefully programmed sequence—or train—of weak pulses,” Malinovskaya explains. “Each pulse carries much less energy, but its timing, intensity, frequency and phase are precisely calculated and controlled.” According to the team’s analysis, this sequence of gentle nudges can drive the same gradual transfer of a quantum system from one state to another that a much stronger pulse would produce. The desired outcome is preserved, while the laser intensity at each step remains low enough to keep multiphoton processes from ever gaining a foothold.

The method is an outgrowth of a well-established control strategy known as adiabatic passage, in which a quantum system is guided slowly and smoothly between energy states so that it remains stable against small imperfections. By digitizing that smooth evolution into discrete steps, the researchers retain the robustness of the adiabatic approach while shedding its traditional dependence on high peak laser intensities. The result, described in a paper titled “Digitizing ultrafast adiabatic passage with a pulse train” published in the Journal of the Optical Society of America B on September 10, 2026, offers a theoretical blueprint that, if borne out experimentally, could reshape how quantum control experiments are designed across a wide range of platforms.

The potential applications stretch across the quantum technology landscape. Quantum sensors, quantum computers and quantum simulators all depend on reliable preparation and manipulation of quantum states, and the pulse-train technique could make those operations more accurate and repeatable. In molecular physics and spectroscopy, where intense laser pulses often generate background effects that interfere with measurements, the lower-intensity approach could yield cleaner data and sharper insights into molecular structure and behavior. The implications extend beyond physics laboratories as well: in biology and medicine, laser-based technologies are widely used for imaging and disease diagnosis, and reducing pulse intensity offers a straightforward way to minimize damage to sensitive cells and tissues, opening a path to safer optical diagnostics and therapies.

For now, the technique remains theoretical, but the paper lays out all of the necessary calculations in full detail, providing experimenters with a concrete roadmap. “The next step will be to actually test it,” Malinovskaya says. “When demonstrated experimentally, this approach will open a new way to precisely control quantum systems with weaker laser fields, making it easier to use in practical applications.” If the laboratory results match the theory, the digitized pulse train could become a standard tool in the quantum engineer’s kit, helping to bring the promise of quantum technologies—from unimaginably powerful computers to sensors capable of detecting the faintest whispers of nature—closer to everyday reality, one carefully timed pulse at a time.

Subject of Research: Digitized ultrafast adiabatic passage using low-intensity laser pulse trains for precise quantum state control

Article Title: Stevens researchers take step toward more precise, practical quantum technologies

Article References: Stevens researchers take step toward more precise, practical quantum technologies. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: quantum control, laser pulses, multiphoton processes, adiabatic passage, quantum computing, quantum sensing, spectroscopy, molecular physics, quantum states, biomedical imaging, Stevens Institute of Technology, Svetlana Malinovskaya