The device belongs to a family of structures known as photonic crystals, periodic arrangements of dielectric materials that control the flow of light in ways natural materials cannot. The concept, famously described by Joannopoulos, Villeneuve, and Fan in Nature in 1997, relies on a photonic bandgap, a range of wavelengths that the periodic structure simply refuses to transmit. When a defect is introduced into the otherwise perfect lattice, light of a specific wavelength becomes trapped at that defect, forming a nanocavity that acts like an extremely selective optical resonator.

The proposed biosensor is based on a two-dimensional rod-in-air photonic crystal arranged in a triangular lattice configuration. Silicon rods stand in air, and the periodic arrangement creates a wide photonic bandgap with strong optical confinement within the sensing region. When a biological sample with a particular refractive index is placed in the cavity, the trapped resonance shifts to a new wavelength. By measuring that shift, the sensor can determine the refractive index of the sample with extraordinary precision, and because different cell types have different refractive indices, the same principle can be used to discriminate cancer cells from healthy ones.

The design process was systematic and rigorous. The photonic band structure was evaluated using the plane-wave expansion method, a computational technique that determines which wavelengths the periodic structure can and cannot transmit. Sensor performance was then analyzed through finite-difference time-domain simulations, which model how light actually behaves in the device over time. The researchers varied the rod radius within the sensing region at a constant lattice parameter, followed by an analysis of lattice constant variations while maintaining the optimized rod radius. Both two-dimensional and three-dimensional configurations were explored to validate the robustness of the design.

The performance figures are remarkable. For the fixed lattice constant configuration, the sensor achieves a sensitivity of 478.57 nanometers per refractive index unit, a quality factor of 1.6230 × 10^7, a detection limit of 1.91922 × 10^-8 refractive index units, and a figure of merit of 5.21043 × 10^6 per refractive index unit. In comparison, the variable lattice constant design provides a higher sensitivity of 965 nanometers per refractive index unit, along with a quality factor of 9.5621 × 10^5, a detection limit of 2.13245 × 10^-7 refractive index units, and a figure of merit of 4.68944 × 10^5 per refractive index unit.

These numbers deserve unpacking. Sensitivity describes how much the resonance wavelength shifts per unit change in refractive index, so a higher sensitivity means the device responds more dramatically to the presence of a biological sample. The quality factor, often abbreviated as Q, describes how sharply the resonance is defined, and a higher Q means the device can distinguish between very similar refractive indices. The detection limit describes the smallest refractive index change the device can reliably detect, and the figure of merit combines sensitivity and quality factor into a single performance score. The fixed lattice constant design achieves an extraordinarily high Q, while the variable lattice constant design achieves nearly double the sensitivity.

The practical challenges associated with photonic crystal fabrication were comprehensively analyzed. Photonic crystals require extremely precise periodic structures, and even small deviations from the ideal geometry can degrade performance. The researchers systematically investigated the impact of structural imperfections on sensor performance, providing a realistic assessment of how the device would behave in a laboratory setting. This kind of analysis is essential for translating computational designs into practical devices, and it distinguishes this work from purely theoretical studies that ignore fabrication realities.

The compact footprint of 82.84 square micrometers is a key advantage. On-chip photonic systems require devices that can be integrated densely, and the small size of this biosensor makes it a promising candidate for integration into advanced on-chip photonic systems. The device could potentially be combined with microfluidic channels that deliver biological samples directly to the sensing region, creating a complete diagnostic platform on a single chip.

The principle of refractive-index-based cancer cell discrimination is grounded in established biology. Cell nuclei have lower refractive index and mass density than cytoplasm, as demonstrated by Schürmann and colleagues in 2016, and cancer cells differ from healthy cells in their internal composition and organization. By measuring the refractive index of a cell with sufficient precision, the device can potentially identify whether that cell is cancerous. The approach is label-free, meaning no chemical tags or fluorescent markers are required, which simplifies the diagnostic process and avoids potential interference from labeling chemicals.

This work builds on a growing body of research into photonic crystal biosensors for biomedical applications. Previous studies have explored photonic crystal platforms for glucose detection, blood component analysis, urine biomarker detection, and cancer cell identification. The current design distinguishes itself through its combination of high quality factor, high sensitivity, low detection limit, and compact footprint, along with its realistic assessment of fabrication imperfections. The researchers, based at the Indian Institute of Technology (Indian School of Mines) Dhanbad and Dibrugarh University, acknowledge the Department of Electronics Engineering at IIT (ISM) Dhanbad for providing the research facilities and computational resources necessary to carry out this work.

Subject of Research: Computation & theory

Article Title: Design of a high-Q photonic crystal nanocavity biosensor for refractive-index-based cancer cell discrimination

Article References: Kumar, S., Mondal, H., & Sen, M. (2026). Design of a high-Q photonic crystal nanocavity biosensor for refractive-index-based cancer cell discrimination. Journal of Materials Science, 61(42), 32962-32984. https://doi.org/10.1007/s10853-026-13745-1

Image Credits: AI Generated

DOI: 10.1007/s10853-026-13745-1

Keywords: photonics, biosensors, cancer detection, nanocavities, silicon photonics, refractive index sensing, Design, high-Q, photonic, crystal, nanocavity, biosensor

The study, published in Scientific Reports, tackles a deceptively simple question: can a computer reliably tell an acute lymphoblastic leukemia blast apart from a healthy hematogone in a peripheral blood smear image? The answer matters because microscopic screening remains one of the most accessible diagnostic tools in oncology, yet it depends entirely on human expertise and human attention. A pathologist scanning hundreds of cells can miss subtle morphological differences, and fatigue, workload, and experience all shape what gets noticed. An automated assistant that flags suspicious cells could support that screening process, catching patterns that even trained eyes might overlook.

What makes MoViTHAFX-Net unusual is its dual architecture. On one side sits MobileNet, a lightweight convolutional neural network originally designed for mobile devices, which excels at extracting local morphological information. It examines the fine details of a cell: the texture of the nucleus, the shape of the membrane, the granularity of the cytoplasm. On the other side sits a Vision Transformer, or ViT, a model that treats an image as a sequence of patches and models long-range contextual relationships between them. Where MobileNet sees the tree, the ViT sees the forest. It understands how distant regions of a cell relate to each other, capturing the global structure that defines a leukemic blast.

But simply running two networks in parallel and gluing their outputs together isn’t enough. The researchers introduced a component called Hierarchical Attention Fusion, or HAF, which first refines the features from both branches and then adaptively combines three learnable morphology-guided feature projections before the final binary classification. In plain terms, HAF doesn’t just merge the two streams of information; it decides, cell by cell, how much weight to give each type of feature. A blast with an unusual nuclear shape might trigger stronger attention to the MobileNet branch, while a cell whose overall structure is telling might shift focus to the transformer. This adaptive weighting is what gives the hybrid model its edge over a direct concatenation of the two networks, as the ablation results in the study demonstrate.

The training data came from the C-NMC 2019 dataset, a benchmark collection of 10,661 labelled cell images from peripheral blood smears. The researchers used an image-level stratified partition, splitting the data into 7462 training images, 1600 validation images, and 1599 held-out test images. Data augmentation, which artificially expands the training set by applying transformations like rotations and flips, was restricted to the training partition to keep the validation and test sets clean. This careful separation ensures the model isn’t accidentally memorizing images it has already seen, a common pitfall in machine learning studies.

On the held-out test partition, the model correctly classified 1583 of 1599 images, corresponding to 99.00 percent accuracy. The precision, recall, and F1-score all came in at approximately 0.990, and the area under the curve, a measure of the model’s ability to distinguish between classes across all possible thresholds, reached 0.99. These numbers mean the model almost never mistakes a healthy cell for a leukemic one, and almost never lets a leukemic cell slip past undetected. In diagnostic terms, that balance between false alarms and missed detections is exactly what a screening tool needs.

Yet the researchers are careful about what these numbers actually prove. Because the reported custom split is image-level rather than subject-disjoint, and evaluation is limited to a single benchmark dataset, the results should be interpreted as internal benchmark performance rather than evidence of patient-independent or multicentre clinical generalization. In other words, the model excels at the specific dataset it was trained on, but whether it would perform equally well on blood smears from different hospitals, different staining protocols, or different patient populations remains an open question. This kind of transparent reporting is rare and valuable in a field where inflated claims often outpace clinical reality.

The team also added a qualitative layer to the system: Grad-CAM++ visualization, a post hoc technique that highlights the image regions contributing to each prediction. When the model flags a cell as a leukemic blast, Grad-CAM++ generates a heatmap showing which parts of the image drove that decision, typically the nucleus or other morphologically significant regions. This transparency is crucial for building trust with pathologists, who need to understand why an AI system reached a conclusion before they act on it. It transforms the model from a black box into something closer to a collaborative partner that shows its work.

The implications reach beyond leukemia. The hybrid CNN-transformer approach with adaptive attention fusion could apply to any classification problem where both local detail and global structure matter, from detecting tumors in histopathology slides to identifying defects in industrial imaging. The architecture demonstrates that combining complementary neural network designs, rather than betting on a single paradigm, can yield measurable gains. As ablation results within the same experimental setting indicate an incremental advantage of Hierarchical Attention Fusion over direct CNN-ViT concatenation, the study offers a concrete design lesson for future hybrid models.

For now, MoViTHAFX-Net remains a research tool, not a clinical one. But the trajectory is clear: as hybrid architectures mature and datasets grow more diverse, models like this could move from benchmark datasets to real diagnostic workflows, supporting microscopic screening in laboratories around the world. The study, published open access, received no external funding and was conducted by A. Elavarasi and N. Jaisankar of the School of Computer Science and Engineering at Vellore Institute of Technology. It represents a step toward a future where the first line of defense against leukemia isn’t a tired human eye, but a tireless hybrid intelligence that sees both the tree and the forest.

Subject of Research: Deep hybrid MobileNet and vision transformer model for acute lymphoblastic leukemia classification from blood smear images.

Article Title: A deep hybrid feature extraction model based on MobileNet and vision transformer for acute lymphoblastic leukemia prediction

Article References: Elavarasi, A., & Jaisankar, N. (2026). A deep hybrid feature extraction model based on MobileNet and vision transformer for acute lymphoblastic leukemia prediction. Scientific Reports. https://doi.org/10.1038/s41598-026-73064-y

Image Credits: AI Generated

DOI: 10.1038/s41598-026-73064-y

Keywords: acute lymphoblastic leukemia, MobileNet, vision transformer, deep learning, C-NMC 2019, Grad-CAM++, hierarchical attention fusion, medical imaging, blood smear, machine learning, cancer detection, deep

The research, led by Yunze Wang, Yuying Hou, Xinyu Liu, Wen Zhang, and Xionghui Zhou of Huazhong Agricultural University together with Qingjiao Li of Sun Yat-Sen University’s Eighth Affiliated Hospital, tackles a practical problem that has quietly limited the field. Whole-genome methylation sequencing of plasma cfDNA is attractive because a single assay can capture both DNA methylation and fragment-level features. But nobody had rigorously tested whether the chemical steps used to read methylation actually preserve the fragmentomic signals that other methods rely on.

The team systematically compared two methylation library preparation protocols, whole-genome bisulfite sequencing (WGBS) and enzymatic methyl sequencing (EM-seq), against conventional whole-genome sequencing (WGS). Their verdict was nuanced. Canonical cfDNA fragmentomic patterns, such as characteristic fragment size distributions and end motifs, were partially preserved across the methylation protocols, but the degree of preservation varied from feature to feature. In other words, methylation sequencing is not a perfect stand-in for WGS when it comes to fragmentation, and researchers combining both modalities from one assay need to know exactly which signals survive the chemistry.

Beyond the technical audit, the study probed the biology linking the two modalities. DNA methylation levels turned out to be associated with characteristic fragment length and end-motif patterns, confirming that how a DNA molecule is chemically decorated influences how it is cut, protected, and degraded in the blood. This coupling is not a nuisance; it is information. Tumor cells, with their altered chromatin and epigenetic landscapes, release DNA that carries coordinated epigenetic and physical fingerprints, and the study set out to organize those fingerprints into a framework a computer model can actually use.

The organizing principle the researchers chose is the methylation haplotype block, or MHB. Rather than treating each CpG site or each fragment independently, the team delineated MHBs directly from plasma cfDNA. These blocks, stretches of the genome where methylation states tend to be inherited together across reads, exhibited tissue-associated regulatory features and distinct patterns of chromatin accessibility, repeat enrichment, and methylation heterogeneity. That makes them biologically interpretable units, anchored in real regulatory architecture rather than arbitrary genomic windows.

Within these MHB regions, methylation and fragmentomic signals showed coordinated yet complementary patterns. Methylation tells the model which tissue the DNA came from; fragmentation tells it how that DNA was processed and cleared. Neither signal alone captures the full picture, but together they form a richer portrait of what is happening in a patient’s body. This insight guided the construction of FAME, the Fragmentomics And Methylation Ensemble, an interpretable multimodal diagnostic model that integrates MHB-centered methylation and fragmentomic features for cancer detection and tissue-of-origin prediction.

The researchers did not stop at a compact model. They extended FAME with complementary genome-wide methylation and fragmentomic features, producing FAME-GW, a genome-wide-augmented version. Across cross-validation, independent validation cohorts, and head-to-head comparisons with established methods including THEMIS (THorough Epigenetic Marker Integration Solution), SPOT-MAS (screening for the presence of tumor by methylation and size), and DELFI (DNA evaluation of fragments for early interception), the FAME-based models achieved robust diagnostic performance. FAME delivered a compact and interpretable representation, while FAME-GW delivered the strongest overall predictive performance.

What makes this work stand out in the crowded liquid biopsy field is its insistence on interpretability. Many high-performing cancer detection models are black boxes, blending thousands of features in ways clinicians cannot inspect. By centering the analysis on MHBs, units with known regulatory and chromatin properties, the FAME framework lets researchers trace a prediction back to biologically meaningful regions. A clinician or scientist can ask which blocks drove a cancer call and what their methylation and fragmentation patterns look like, rather than trusting an opaque score.

The study also carries a practical message for anyone designing cfDNA assays. Because methylation library preparation partially preserves fragmentomic signals, a single whole-genome methylation sequencing run can serve double duty, capturing both epigenetic and fragment-level information from one precious plasma sample. That is a meaningful efficiency gain for multi-cancer early detection programs, where sample volume is limited and every read counts. But the feature-by-feature variation in preservation means assay developers must validate each fragmentomic signal they intend to use under their specific protocol.

The work, published on 15 October 2026 as an open-access article in Genome Medicine, was supported by the Fundamental Research Funds for the Central Universities and the Huazhong Agricultural University Scientific and Technological Self-innovation Foundation. The in-house cohort drew healthy controls from The Eighth Affiliated Hospital of Sun Yat-sen University and gastric cancer samples from Fujian Medical University Union Hospital, with ethics approval and written informed consent obtained for all participants. As methylation haplotype blocks move from a research framework toward clinical assay design, the study’s central conclusion is likely to shape the next generation of cfDNA diagnostics: the epigenetic and fragmentomic stories written in circulating DNA are not separate chapters but a single, coupled narrative, and models that read them together, in interpretable units, are the ones most likely to reach the clinic.

Subject of Research: Integrating coupled methylation and fragmentomic signals in plasma cell-free DNA for interpretable cancer diagnosis

Article Title: Decoding coupled epigenetic-fragmentomic signals in circulating cell-free DNA for interpretable cancer diagnosis

Article References: Wang, Y., Hou, Y., Liu, X., Li, Q., Zhang, W., & Zhou, X. (2026). Decoding coupled epigenetic-fragmentomic signals in circulating cell-free DNA for interpretable cancer diagnosis. Genome Medicine. https://doi.org/10.1186/s13073-026-01793-w

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01793-w

Keywords: cell-free DNA, DNA methylation, fragmentomics, methylation haplotype blocks, liquid biopsy, cancer detection, whole-genome bisulfite sequencing, EM-seq, FAME model, tissue of origin, Genome Medicine, interpretable machine learning

The compound at the heart of the work is azulene, a naturally occurring molecule responsible for the vivid blue colour of certain fungi and corals. Among the species that carry it is Lactarius indigo, the indigo milk cap mushroom, whose unusual pigmentation has long fascinated chemists. Azulene also appears in some skincare products, where it is used to calm irritation. But it is the molecule’s optical behaviour, rather than its soothing reputation, that has made it attractive to the Bath team: azulene absorbs light at one wavelength and emits light at a different, typically longer wavelength, which is precisely the property needed for a molecule to serve as a fluorophore.

Fluorescent probes are the workhorses of modern biological imaging. Each probe is built from two functional parts: a receptor, which is engineered to detect or bind a particular substance of interest inside or around a cell, and a fluorophore, the light-handling component that turns that binding event into a visible signal. When illuminated under a microscope, the fluorophore absorbs light of a specific wavelength and re-emits light at a longer wavelength, appearing as a characteristic colour. It is this coloured glow that allows researchers to track molecules, map cellular structures and monitor disease-related processes that would otherwise remain invisible.

Building on azulene’s unusual optical properties, the Bath group, led by Professor Simon Lewis, Professor Tony James and Dr Dan Pantos in the Department of Chemistry, had previously developed the AzuFluor platform, a family of azulene-based probes capable of visualising important biological processes. The new study, whose first author is postdoctoral researcher Dr Chiara Faverio, takes that platform a significant step further. Dr Faverio synthesised and analysed a wide range of new azulene derivatives, allowing the team to systematically explore how subtle structural changes influence the fluorescence colour and brightness of the resulting molecules.

That systematic exploration produced a headline result: the team created a molecule five times brighter than the original azulene molecule. Brightness matters enormously in fluorescence imaging, because a brighter probe can be used at lower concentrations, produces cleaner images with less background noise, and makes faint biological signals easier to detect. The five-fold improvement gives the researchers what they describe as a promising candidate for future sensing and imaging applications, and demonstrates that the structural rules they uncovered can be used to tune probe performance in predictable ways.

Crucially, the study did more than produce one brighter molecule. By comparing the behaviour of many derivatives, the researchers identified the molecular features that control both brightness and emission colour in azulene fluorophores. That structure-property relationship is the real prize of the work, because it converts what had been a largely empirical exercise into a design discipline. Chemists wanting a probe that glows in a particular colour, or at a particular intensity, can now consult a set of established principles rather than relying on trial and error, dramatically shortening the path from concept to working imaging tool.

The implications reach into some of the most challenging areas of medicine. Neurodegenerative diseases such as Alzheimer’s and Parkinson’s involve molecular changes inside cells that unfold over long periods and are difficult to observe directly, while cancer research depends on tracking signalling events and molecular interactions that drive uncontrolled growth. Probes bright enough and sensitive enough to light up these processes in real time, inside living cells, give researchers a window on disease mechanisms that static measurements cannot provide. The Bath team’s blueprint is intended to make such probes easier to design on demand.

Dr Faverio framed the significance of the advance in terms of what brighter probes make possible in the laboratory. Fluorescent probes, she noted, act like molecular beacons, lighting up in response to specific biological targets and allowing researchers to observe processes that would otherwise remain invisible. Being able to design brighter, more sensitive probes, she said, opens up exciting new opportunities for studying the cellular changes associated with neurodegenerative diseases, cancer and other serious health conditions. Her synthesis and analysis of the new azulene derivatives formed the experimental backbone of the study, providing the systematic dataset on which the design principles rest.

Professor Lewis placed the work in the context of the group’s longer engagement with the molecule. When the team began exploring azulene fluorophores in 2019, he said, they recognised the compound’s unusual and exciting potential. The new publication, he added, reveals the molecular features that control brightness and emission, providing a clear roadmap for designing increasingly powerful probes for sensing and bioimaging applications. In other words, seven years of accumulated chemistry on a single blue scaffold has now been distilled into a set of rules that other laboratories can pick up and apply.

The publication also marks a personal milestone. It is the fiftieth paper of 2026 for Professor Tony D. James, capping an exceptionally productive year spanning supramolecular chemistry, sensing technologies, fluorescence imaging and molecular diagnostics. James is widely known for his work on fluorescent probes for biologically important molecules, and the azulene programme exemplifies the blend of synthetic chemistry and molecular recognition on which that reputation rests. The authors note that the University of Bath owns the trademark on the AzuFluor name, underlining how the research has moved from curiosity-driven chemistry towards a recognised platform technology. With the structure-property relationships now mapped, the team’s blueprint points towards a next generation of fluorescent materials in which colour and brightness can be engineered to order, giving disease researchers brighter beacons with which to illuminate the molecular events that matter.

Subject of Research: Structure-property relationships in azulene-based fluorescent probes for biological imaging and disease detection

Article Title: Chemists develop blueprint for designing powerful fluorescent materials for better disease detection

Article References: Chemists develop blueprint for designing powerful fluorescent materials for better disease detection. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: azulene, fluorescent probes, fluorophores, bioimaging, chemical sensing, University of Bath, Alzheimer's disease, Parkinson's disease, cancer detection, structure-property relationships, Organic Chemistry Frontiers, EPSRC

The review, led by Emma U. Hammarlund of Lund University together with Per Malmberg, Julhash U. Kazi, Nicole R. Posth and Kenneth J. Pienta of Johns Hopkins, traces the intellectual journey of stable isotope science from its origins in early twentieth-century physics to its potential future at the bedside. The central claim is bold but carefully argued: the same fractionation principles that allow geologists to reconstruct ancient climates from seashells could allow clinicians to read altered physiology directly from blood, urine, and tumor tissue, without administering any labelled substance at all.

The story begins with puzzling discrepancies in atomic weights noticed by chemists in the late nineteenth century. Ernest Rutherford and Frederick Soddy showed in a burst of papers between 1902 and 1903 that radioactive atoms transform into new forms of matter and eventually into stable atoms. Soddy went further, concluding that chemical identity depends on nuclear charge rather than atomic weight, and in a short 1913 letter to Nature he introduced the term isotope, meaning same place, a word suggested to him by Dr. Margaret Todd. That same year, J. J. Thomson recorded faint traces of a mass-22 neon isotope on a photographic plate, and in 1914 Stefanie Horovitz, working at the Radium Institute in Vienna, used painstaking gravimetric analyses to confirm Soddy’s prediction that lead derived from uranium was slightly lighter than ordinary lead. Francis William Aston’s mass spectrograph, which earned him the 1922 Nobel Prize, then revealed that many elements are mixtures of isotopes, laying the groundwork for isotope analysis with steadily increasing precision.

Human physiology was the first field to put stable isotopes to work. In the 1930s, Rudolf Schoenheimer and David Rittenberg used newly described heavy isotopes of hydrogen and water to trace metabolism in rats and mice, detecting the labels through simple density measurements before mass spectrometers were even widely available. Their work revealed that the body’s fats and proteins are in a continuous dynamic state of synthesis and breakdown, and it established isotope tracer biochemistry in medicine, a paradigm that still dominates clinical applications today in fields from gastroenterology to metabolic research.

The deeper power of isotopes emerged once scientists understood that physical and biological processes fractionate them in predictable ways. In the late 1940s, Harold Urey showed that small mass differences alter bond vibrational energies, causing heavier isotopes to concentrate in the phases where they are most strongly bound. Because this equilibrium fractionation is temperature dependent, Urey proposed that oxygen isotopes in calcium carbonate shells could reconstruct past temperatures, the first paleothermometer. In parallel, Jacob Bigeleisen formulated the quantitative theory of kinetic isotope effects, showing that lighter isotopes react faster, diffuse more easily, and can be preferentially removed in one-directional processes. Together, these two frameworks turned isotope ratios into narratives of source, transformation, and constraint.

What made the method truly robust, the authors emphasize, was not just instrumentation but a globally shared reference system. Absolute isotope ratios are notoriously hard to measure accurately, so the community built common scales anchored in standards such as PDB for carbon, Standard Mean Ocean Water for water isotopes, and Canyon Diablo Troilite for sulfur. From the 1960s onward, the International Atomic Energy Agency, NIST, and the US Geological Survey established replacement standards and distributed them worldwide, and laboratories adopted two-point or multipoint normalization to correct for instrument bias. Few analytical methods achieve this degree of comparability across laboratories and decades, with pH and temperature among the closest parallels. It is this ecosystem of shared references that allowed isotope data to travel, and it is precisely what medicine has only partly adopted.

The applications across the natural sciences illustrate the principle. Rare sulfur isotopes in sedimentary rock record the rise of Earth’s ozone layer, because ultraviolet-driven photolysis produced a mass-independent signature that vanished once the ozone shield formed. Carbon isotopes connect biological carbon fixation to the sedimentary record and distinguish C3 from C4 plant consumption in diets. In biology, archaeology, and forensics, isotope ratios in feathers, bones, and hair reveal provenance and movement: hydrogen isotopes in feathers link migratory birds to their breeding regions, strontium and oxygen isotopes trace prehistoric human migration, and hydrogen and oxygen in scalp hair reflect local drinking water averaged over weeks to months, mapped through geographic isoscapes. In epidemiology, carbon, nitrogen, and sulfur isotopes can distinguish food sources and dietary patterns more objectively than self-report, and lead isotopes can identify whether exposure came from paint or pipes.

Medicine, the authors argue, is not an outlier to this story but its next logical step. The clinical use of stable isotopes has largely remained within the tracer paradigm, from the 13C-urea breath test, in which urease-positive Helicobacter pylori hydrolyses labelled urea to release detectable 13CO2 in exhaled breath, to 13C-labelled substrates tracing metabolic flux through glycolysis and the TCA cycle, and hyperpolarized 13C-MRI and deuterium metabolic imaging mapping tumor metabolism without radiation burden. But fractionation related to disease offers something tracers cannot: a read-out of altered physiology without any administered label. Although isotope effects are small at the level of individual reactions, they accumulate across metabolic networks and become detectable in metabolites, tissues, and excreted products.

The clinical examples already span several isotope systems and tumor types. Serum calcium isotope fractionation tracks bone mineral balance: bone mineralization preferentially incorporates the lighter calcium isotope into the crystal lattice, leaving blood enriched in the heavier isotope, so when osteolytic resorption exceeds formation in active multiple myeloma, isotopically light calcium is released back into circulation. Serum values are accordingly lower in patients with active myeloma, around minus 0.75 per mille versus minus 0.66 per mille in remission, tracking disease activity earlier than bone-density imaging. In hepatocellular carcinoma, blood copper and sulfur isotope compositions differ from controls, and laser-ablation multicollector ICP-MS can now map copper isotope variation across tumor regions. Breast tumors are enriched in light zinc isotopes, with metallothionein-linked zinc sequestration proposed as a mechanism, and bladder cancer studies combining plasma copper isotope composition with machine learning have achieved high classification performance. These signals are modest, typically a few tenths of a per mille, but they exceed the roughly 0.05 to 0.1 per mille precision of some mass spectrometry instruments, and their magnitude is comparable to fractionation within the human body itself.

The challenges are real, and the review does not shy away from them. Natural-abundance isotope ratios integrate diet, geography, age, sex, renal and hepatic function, inflammation, medications, and microbiome composition, so clinical claims will require matched disease controls, longitudinal sampling, paired concentration measurements, and explicit metadata capture. Operational barriers include instrument cost, acid digestion, chromatographic isolation of analytes, and limited throughput. Nor are isotope ratios likely to replace circulating tumor DNA, methylation assays, or conventional metabolomics as stand-alone screens; their more realistic niche is mechanism-linked adjunct information integrated with multiomics and machine learning. Yet the authors see the balance shifting. High-throughput workflows, from LC/GC-IRMS to ICP-MS and Orbitrap-based isotopologue analysis, may not need the maximal precision of geochemistry if predictive performance comes from multidimensional isotope features fused with clinical and genomic data. Combined with exposomics and whole genome sequencing, they suggest, isotope signals could even seed a new field of carcinogenomics linking genetic susceptibility to environmental exposure. If medicine learns to read isotopes the way Earth scientists do, a blood sample may one day report not just what is present, but which processes are shifting in disease.

Subject of Research: Stable isotope fractionation as mechanistic biomarkers in medical oncology

Article Title: Stable isotopes from Earth science to bedside medicine: beyond tracers and toward mechanistic biomarkers in medical oncology

Article References: Hammarlund, E. U., Malmberg, P., Kazi, J. U., Posth, N. R., & Pienta, K. J. (2026). Stable isotopes from Earth science to bedside medicine: beyond tracers and toward mechanistic biomarkers in medical oncology. Medical Oncology, 43(11), Article 332. https://doi.org/10.1007/s12032-026-03436-x

Image Credits: AI Generated

DOI: 10.1007/s12032-026-03436-x

Keywords: stable isotopes, isotope fractionation, mechanistic biomarkers, medical oncology, mass spectrometry, cancer detection, multiple myeloma, calcium isotopes, machine learning, liquid biopsy, reference materials, exposomics

The ASPIRE Prize, whose full name is the APEC Science Prize for Innovation, Research and Education, is awarded each year to an outstanding early-career scientist who has demonstrated both scientific excellence and a strong commitment to cross-border collaboration within the APEC region. Each member economy may nominate one scientist under the age of 40, and a single winner is chosen through a regional selection process. The structure of the prize is deliberately designed to spotlight researchers whose work transcends national boundaries, and Luo’s selection at this early stage of his academic career underscores the global reach of the computational tools his laboratory has built. According to APEC, the prize represents not only excellence in science and technology but also the organization’s collective aspiration to address global challenges through collaboration and forward-looking leadership.

This year’s award specifically recognizes outstanding contributions that harness artificial intelligence and data science to strengthen industries and build resilient economies for the future, a theme that sits squarely at the center of Luo’s research program. His work focuses on developing AI-powered computational tools that have transformed the analysis of complex and highly diverse genomic data. Genomic datasets are among the largest and most computationally demanding in modern biology: a single human genome consists of roughly three billion base pairs, and sequencing instruments produce fragments that must be assembled, aligned, and interpreted against reference sequences before any biological meaning can be extracted. The algorithms Luo and his collaborators have developed address precisely this bottleneck, and their breakthroughs have made genomic sequencing significantly faster, more accurate, and highly cost-effective than previous approaches allowed.

The practical consequences of that speed and accuracy extend far beyond the laboratory. Faster and cheaper sequencing lowers the cost per genome to the point where large-scale clinical deployment becomes feasible, which in turn opens the door to applications in healthcare, food security, and biotechnology. Luo’s innovations are described as instrumental in translating complex genomic research into practical, clinical solutions for disease prevention, precision medicine, and clinical decision support. In precision medicine, the ability to rapidly characterize a patient’s genome can inform drug selection and dosing; in disease prevention, population-scale sequencing can reveal inherited risk factors before symptoms appear; and in agriculture and food security, genomic tools underpin the breeding of resilient crops and livestock and the monitoring of pathogens that threaten food supplies.

The adoption figures attached to Luo’s work illustrate how a research algorithm can evolve into global infrastructure. The open-source AI tools developed by his laboratory have been downloaded more than 10 million times worldwide, with approximately 60 percent of those downloads originating from other APEC economies. Open-source distribution of this kind is significant for two reasons. First, it means the tools are not locked behind commercial licensing, so research groups, hospitals, and public health agencies across economies with very different resource levels can integrate them into their own pipelines. Second, the concentration of downloads within the APEC region demonstrates the kind of cross-border scientific diffusion that the ASPIRE Prize is explicitly designed to reward, with the laureate’s software functioning as shared technical infrastructure across the very economies the prize seeks to connect.

Those tools are now widely integrated into global healthcare, agricultural, and medical biotechnology industries, and Luo has described how they are moving from research settings into clinical environments. In a statement accompanying the announcement, he said that the algorithms developed by his lab are now being integrated into clinical decision-support systems that can help detect early-stage cancer and rare diseases in newborns. For him, he added, that is what innovation is about: turning research into tools that can help save lives. The two clinical scenarios he identified are among the most demanding in genomic medicine. Early-stage cancer detection requires algorithms sensitive enough to distinguish faint tumor-derived signals from abundant background genetic material, while newborn screening for rare diseases depends on rapid turnaround, because many rare genetic conditions are most treatable within days or weeks of birth.

The technical challenge that Luo’s career has been built around is a familiar one to anyone working in bioinformatics: raw sequencing data is voluminous, noisy, and fragmented, and the quality of every downstream conclusion depends on how well it is processed. Sequence assembly and alignment, the core problems his algorithms target, are computationally hard problems that scale with both genome size and dataset volume. As sequencing technology has shifted toward producing ever larger volumes of data at lower cost, the computational step has increasingly become the limiting factor in genomic workflows. AI-driven approaches, including machine learning methods that can correct errors, resolve repetitive regions, and classify variants, have emerged as a way to push past those limits, and Luo’s laboratory has positioned itself at the leading edge of that shift, producing tools that are simultaneously faster, more accurate, and cheaper to run than the alternatives.

Luo’s path to this recognition reflects the international character of contemporary computational biology. He completed his PhD training in bioinformatics at the University of Hong Kong between 2010 and 2015, working with Professor Tak-Wah Lam, and then undertook postdoctoral training with Professors Steven Salzberg and Michael Schatz at the Center of Computational Biology at Johns Hopkins University from 2016 to 2017. Salzberg and Schatz are among the most prominent figures in genome informatics, and that training environment placed Luo at the intersection of algorithm design and large-scale genome assembly at a moment when the field was being transformed by high-throughput sequencing. He is now an Associate Professor in HKU’s School of Computing and Data Science, in addition to his leadership roles in the AI and Data Science Division, where his work spans bioinformatics algorithms and clinical informatics.

His publication record and the recognition it has attracted trace the trajectory of a researcher whose influence has grown rapidly. Luo has published more than 100 papers, ten of which have each achieved over a thousand citations, a benchmark that places those papers among the most influential in their fields. He has been identified as a Top 1 Percent Scholar Worldwide by Clarivate Analytics every year since 2019, was selected by Baidu Research as one of the Worldwide Top 150 Chinese Young Scholars in AI, was named one of the Top 10 Innovators Under 35 in Asia Pacific by MIT Technology Review in 2019, and was recognized in Forbes’ 30 Under 30 Asia list in the healthcare and science category in 2017. Taken together, these honors chart a career in which algorithmic innovation, clinical application, and international collaboration have advanced in parallel rather than in sequence.

For the Asia-Pacific research community, the 2026 ASPIRE Prize carries a message about where the region’s scientific priorities are heading. APEC’s framing of this year’s award, emphasizing artificial intelligence and data science as tools for strengthening industries and building resilient economies, signals that the organization sees computational biology not as a niche discipline but as strategic infrastructure for public health, agriculture, and biotechnology across its member economies. Luo’s own account of his motivation, that innovation matters when research becomes a tool that can help save lives, captures the ethos the prize is intended to celebrate. With his open-source tools already embedded in clinical decision-support systems and downloaded millions of times across the region, the 2026 laureate offers a concrete example of how early-career research, when built openly and aimed at real clinical problems, can ripple outward into healthcare systems and industries far beyond the laboratory where it began.

Subject of Research: Artificial intelligence technologies for genomic analysis and clinical decision support

Article Title: HKU scholar named sole recipient of 2026 APEC ASPIRE Prize

Article References: HKU scholar named sole recipient of 2026 APEC ASPIRE Prize. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: APEC ASPIRE Prize, Ruibang Luo, genomics, artificial intelligence, bioinformatics, precision medicine, clinical decision support, open-source software, The University of Hong Kong, newborn screening, cancer detection, data science

The clinical problem the team set out to solve is both common and consequential. Population-based screening programs have dramatically increased the detection of large advanced adenomas, and in the Netherlands alone roughly 4,000 patients each year are diagnosed with polyps larger than two centimeters or early-stage rectal cancer. Treatment for these lesions depends entirely on their biological nature. Polyps containing only low-grade dysplasia, a mild precancerous change, can be removed piecemeal with standard endoscopic techniques available in virtually every hospital. But polyps harboring high-grade dysplasia or early invasive carcinoma demand en-bloc resection, in which the entire lesion is removed in one piece, using sophisticated methods such as endoscopic submucosal dissection or endoscopic intermuscular dissection. These procedures are more complex, more expensive, and restricted to a small cadre of specially trained endoscopists at tertiary centers.

When the distinction is missed, the consequences cascade. If a malignant polyp is understaged and removed piecemeal, pathologists may be unable to assess whether the margins are clear, whether the cancer was completely excised, or whether features indicating lymph node metastasis risk are present. Patients may then require additional major abdominal surgery with substantial complication risks. Even among experts, between 13 and 22 percent of polyps containing high-grade dysplasia or invasive carcinoma go unrecognized, and in community settings that miss rate has been reported to reach a staggering 81 percent. The reverse error is equally costly: overstaging benign polyps triggers unnecessary referrals and operations. Current Dutch guidelines therefore recommend considering en-bloc resection for all rectal polyps larger than two centimeters, yet only about 10 percent of these lesions ultimately prove to contain a malignant component. A large American study found that a quarter of patients undergoing major abdominal surgery for a suspected malignant polyp had benign disease.

The Leiden team’s strategy replaces guesswork with molecular targeting. Their agent, SGM-101, is a chimeric anti-CEA antibody conjugated to a fluorophore called BM-105, which emits light in the near-infrared spectrum. CEA was chosen because earlier work showed it is overexpressed in approximately 75 percent of high-grade dysplasia and early invasive carcinoma cases, while being nearly absent in low-grade dysplasia and normal mucosa in 66 and 98 percent of cases, respectively. When infused intravenously several days before the procedure, the antibody circulates through the body and binds to CEA on malignant cells. A camera system tuned to the fluorophore’s emission wavelength then excites the dye and renders the labeled tissue visible in real time, effectively painting cancer cells in light that the human eye would otherwise never perceive.

In the trial, 22 patients scheduled for en-bloc endoscopic resection of advanced rectal polyps received either 5 or 10 milligrams of SGM-101 via a 30-minute intravenous infusion, three to five days before their procedure. Thirteen of the polyps ultimately proved to contain high-grade dysplasia or invasive carcinoma, while nine contained only low-grade dysplasia, providing both positive and negative test cases. Because standard flexible endoscopes are not yet equipped with fluorescence detection, the researchers used a rigid Quest-Olympus laparoscope introduced transanally through a balloon trocar for in vivo imaging. After resection, specimens were imaged ex vivo on the back table, then sliced into 3-to-5-millimeter bread-loaf sections that were imaged with a Li-Cor PEARL system, allowing direct correlation of fluorescence signals with stained histological slides.

The quantitative results were compelling. On bread-loaf analysis, the median fluorescence intensity rose in step with histological grade, from 0.05 in normal mucosa to 0.14 in low-grade dysplasia and 0.31 in high-grade dysplasia or invasive carcinoma, with all three values significantly different from one another. The primary endpoint, a tumour-to-low-grade-dysplasia ratio, reached a median of 1.7 across the cohort, rising to 2.5 in patients who received the 10-milligram dose compared with 1.3 in the 5-milligram group, a statistically significant difference. The tumour-to-background ratio of 3.7 indicated strong contrast between malignant tissue and normal mucosa. Overall, fluorescence imaging detected high-grade dysplasia and invasive carcinoma with 86 percent sensitivity and 88 percent specificity, remarkable figures for a first-in-context application.

Perhaps most striking was the technology’s ability to correct expert errors. Routine white-light endoscopy, performed by highly experienced gastroenterologists, misclassified the polyp stage in three of the 22 patients. Fluorescence imaging with the clinical Quest system correctly identified the stage in all three cases: two benign polyps wrongly suspected of malignancy showed no fluorescent signal, while one polyp judged benign on white light but harboring high-grade dysplasia lit up precisely at the malignant focus. The imaging did fail in three other cases where conventional assessment was correct, largely due to biological heterogeneity in CEA expression and technical limitations of the rigid scope, which could not visualize three polyps at all because of its restricted angulation.

Those limitations underscore that this remains an early chapter rather than a finished story. The sample was small, with only nine polyps eligible for the primary ratio analysis, and the findings are explicitly hypothesis-generating. The rigid laparoscope restricted inclusion to distal rectal lesions and made positioning awkward. The intravenous agent must be given days before the procedure, whereas an endoscopist would ideally decide during the examination whether fluorescence is needed. CEA expression itself is not uniform across colorectal neoplasia, which likely explains the false negatives. Encouragingly, in the final two patients the team tested a prototype flexible fiber fluorescence endoscope from Olympus ex vivo, and its signals were broadly comparable to the rigid system, supporting a path toward true in vivo fluorescence colonoscopy. The researchers also note that fluorescence ratios showed only a non-significant trend toward predicting invasion depth, a question that will require larger, adequately powered studies.

The broader context makes the advance especially timely. Fluorescence-guided surgery has matured rapidly, with multiple approved agents and commercial camera systems now in operating rooms, but fluorescence-guided endoscopy lags behind, in part because no flexible fluorescence-capable endoscope has yet been commercialized. Other groups have charted parallel routes: a c-Met-targeted tracer once revealed dysplastic colon lesions invisible under white light, a VEGF-A-targeted antibody has assessed treatment response in rectal cancer, and the international CLASSICA trial is exploring the perfusion dye indocyanine green for tissue classification, a tracer spectrally compatible with SGM-101. The Leiden team envisions two realistic implementation pathways: selective use by screening endoscopists facing difficult-to-classify polyps, helping decide between piecemeal removal and referral for en-bloc resection, and use by experts during resection itself to guide the choice between dissection techniques. If larger prospective trials confirm these results, molecular light could transform one of endoscopy’s most persistent uncertainties into a visible answer, sparing countless patients unnecessary surgery while catching dangerous lesions before they escape.

Subject of Research: CEA-targeted near-infrared fluorescence endoscopy for detecting high-grade dysplasia and early invasive rectal cancer

Article Title: CEA-targeted fluorescence endoscopy for the detection of high-grade dysplasia and early invasive rectal cancer: A proof-of-concept trial

Article References: Warmerdam, M. I., Dijkhuis, T. H., Linders, D. J. G., Pesch, C. M. W., Crobach, A. S. L. P., Matoshi, A., Basiliya, K., van den Berg, S. H., Cailler, F., Peeters, K. C. M. J., Holman, F. A., Mieog, J. S. D., Vahrmeijer, A. L., Hilling, D. E., & Boonstra, J. J. (2026). CEA-targeted fluorescence endoscopy for the detection of high-grade dysplasia and early invasive rectal cancer: A proof-of-concept trial. European Journal of Nuclear Medicine and Molecular Imaging. https://doi.org/10.1007/s00259-026-08155-z

Image Credits: AI Generated

DOI: 10.1007/s00259-026-08155-z

Keywords: fluorescence-guided endoscopy, CEA-targeted imaging, SGM-101, rectal cancer, colorectal polyps, high-grade dysplasia, near-infrared fluorescence, endoscopic submucosal dissection, molecular imaging, clinical trial, cancer detection, Leiden University Medical Center

The review, led by Muhammad Danyal Ahsan and colleagues at Weill Cornell Medicine, was preregistered with PROSPERO and conducted according to the PRISMA 2020 reporting standard. The team searched MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials from their inception through August 2025, screening 18,231 abstracts to identify 17 original research studies published between 2005 and 2024. The studies came from eight countries, with Germany contributing nine of the seventeen, and sample sizes ranged from 22 to 7,287 participants. Critically, the researchers only included studies of people who had no established indication for the scan, excluding patients with known hereditary cancer syndromes, people undergoing cancer staging or recurrence surveillance, and studies that selectively reported results from a single organ system.

The headline numbers are striking. Across the pooled data, only 16.9 percent of scanned individuals had no findings reported, while 78.0 percent had at least one finding. That sounds alarming until the nature of those findings is examined. The most common abnormalities were degenerative spinal disease in 22.0 percent of scans, subcortical cysts in 15.4 percent, and bone marrow edema in 6.2 percent, conditions that are ubiquitous in aging populations and rarely change clinical management when discovered incidentally. The authors estimate that roughly 62 to 67 percent of people undergoing whole-body MRI without a medical indication will receive findings that are not actionable at all.

A smaller but clinically meaningful fraction of scans did flag something requiring attention. Non-malignant actionable findings, defined as non-cancerous abnormalities that prompted laboratory testing, additional imaging, or procedural follow-up, were reported in 14.8 percent of screened individuals. These were most frequently located in the gastrointestinal system, the musculoskeletal system, and the lymphatic or vascular system. The most common specific actionable findings were hepatic steatosis, or fatty liver, in 8.8 percent, pulmonary nodules in 4.8 percent, and uterine fibroids in 4.3 percent. Among the people with such findings, 58.2 percent went on to pursue additional laboratory work or imaging, and 20.8 percent underwent a subsequent procedure such as endoscopy, biopsy, resection, or surgery.

Herein lies the central problem the review exposes: none of the included studies reported the results of that downstream evaluation. Because diagnostic resolution was never documented, the authors could not calculate false-positive rates, overdiagnosis rates, or any measure of whether the follow-up testing ultimately helped or harmed patients. This is a fundamental gap for a screening technology. Screening is only justified when it demonstrably improves clinical outcomes, and the history of medicine is littered with tests that found plenty of abnormalities without extending a single life. Prostate-specific antigen screening and routine diabetes screening in asymptomatic adults, for example, have not proven effective at reducing all-cause mortality, despite detecting real biochemical and metabolic signals.

The cancer detection numbers offer a more nuanced picture. Among the 4,463 participants in the eleven studies that reported pathologically confirmed oncologic outcomes, malignancy was detected in 1.4 percent, with a 95 percent confidence interval of 0.9 to 2.2 percent. The most common cancers found were genitourinary tumors, accounting for 21.8 percent of detections, followed by cancers of the female reproductive system at 19.5 percent and gastrointestinal cancers at 16.5 percent. Notably, the authors point out that roughly 60 percent of the malignancies discovered were in organ systems for which the US Preventive Services Task Force offers no Grade A or B screening recommendation, since current guidelines cover only breast, cervical, colon, and lung cancers. Fewer than 20 percent of US cancers are diagnosed through existing screening programs, which is precisely the gap that multi-organ scanning hopes to fill.

For context, the pooled malignancy yield of whole-body MRI sits within or above the per-examination detection rates of some recommended screening tests. Mammography detects roughly 0.2 to 0.8 percent of cancers per screening round, and colonoscopy detects approximately 0.14 to 0.20 percent. But the authors are careful to emphasize that diagnostic yield alone establishes nothing about clinical benefit. A scan that finds more cancers is not automatically a scan that saves lives; the detected cancers must be ones that would have become lethal without early detection, and the harms of chasing false positives and indolent lesions must be weighed against any mortality reduction. Those calculations simply cannot be made with the current evidence.

The methodological quality of the underlying literature is another major caveat. None of the seventeen included studies was free from risk of bias across all domains when assessed with Joanna Briggs Institute tools, and limitations included concerns about sampling, sample size, and confounding. Statistical heterogeneity between studies was extreme, with I-squared values exceeding 90 percent across most analyses, reflecting differences in the use of contrast agents, diffusion-weighted imaging, vascular protocols, body coverage, thresholds for calling a finding actionable, and outcome definitions. The confidence intervals around the pooled estimates are correspondingly wide; the 95 percent interval for the proportion of scans with any finding spans from 44.9 to 93.9 percent. A funnel plot for the malignancy analyses also showed asymmetry consistent with publication bias or selective reporting, meaning studies that found more cancers may have been more likely to be published.

Generalizability is a further concern. All of the included studies came from non-US healthcare systems in East Asia, Western Europe, and the Middle East, and several involved employer-sponsored, private, or otherwise selected cohorts rather than representative community samples. The pooled cohort was 64.9 percent male, no studies reported race and ethnicity data, and only seven of the seventeen studies explicitly described their populations as asymptomatic. Meanwhile, the commercial landscape that the review is meant to inform is dominated by direct-to-consumer services such as Prenuvo and Function Health, which charge approximately 500 to 3,000 dollars per scan with variable protocols and are often bundled with additional clinical assessments. Marketing campaigns and celebrity endorsements have amplified public interest, yet no professional society, including the American College of Radiology, the USPSTF, or the National Comprehensive Cancer Network, endorses whole-body MRI for the general population.

The authors argue that the technology itself is not the problem; the evidence base is. Whole-body MRI is radiation-free, images multiple organ systems in a single session, and could theoretically address real barriers to preventive care, since organ-specific screening requires multiple visits, is sometimes invasive, and suffers from suboptimal adherence. Advances in artificial intelligence may further reduce acquisition time and standardize interpretation, and the authors suggest that a consensus reporting taxonomy extending the oncologic ONCO-RADS system to cardiovascular, musculoskeletal, and metabolic abnormalities could reduce variability and unnecessary follow-up. But until high-quality observational studies or randomized controlled trials track patients from initial finding through downstream workup to hard outcomes such as morbidity, mortality, overdiagnosis, and cost, the review concludes that most whole-body MRI examinations in people without a medical indication reveal either nothing or nothing that matters, and the clinical value of the scan remains an open question rather than a settled fact.

Subject of Research: Diagnostic yield of whole-body MRI as a multi-disease screening tool in people without an evidence-based indication

Article Title: The Diagnostic Yield of Whole-Body MRI as a Multi-disease Screening Modality: A Systematic Review and Meta-analysis

Article References: Ahsan, M. D., Frontera, N., Gurram, A., Sharaf, S., Qazi, M., Patel, D., Yang, K., Gioia, S., Murray, I. R., McGonigle, R., Guniganti, P., Dodelzon, K., Ma, X., Kierans, A. S., Diamond, L. C., Godwin, K., Shih, G., Sabuncu, M., Lantos, J., … Sharaf, R. N. (2026). The Diagnostic Yield of Whole-Body MRI as a Multi-disease Screening Modality: A Systematic Review and Meta-analysis. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10802-y

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10802-y

Keywords: whole-body MRI, screening, diagnostic yield, meta-analysis, overdiagnosis, cancer detection, incidental findings, preventive medicine, medical imaging, false positives, systematic review, direct-to-consumer health

The device, developed by Isidora Panez-Toro and colleagues, targets circulating tumor cells, or CTCs, the wandering seeds of metastasis that shed from primary tumors into the bloodstream. These cells are extraordinarily scarce, often fewer than ten per ten milliliters of blood, and they survive in circulation for less than an hour before immune attack and shear stress destroy them. Catching them has long been compared to finding a handful of specific needles in a haystack the size of a stadium. Yet their capture matters enormously, because CTCs carry information about tumor biology that can guide therapy selection and track how a cancer evolves in real time.

The team’s approach rests on a physical phenomenon first described in 1961, when Segré and Silberberg observed that particles flowing through a straight channel migrate, seemingly spontaneously, to orderly lateral positions. Scientists later worked out the forces behind this inertial focusing: a shear-induced lift that pushes particles away from the channel’s centerline and a wall-induced lift that repels them from the walls. When the channel is coiled into a spiral, a secondary rotating flow known as Dean flow adds a drag force that nudges particles sideways. The delicate balance between these lift forces and Dean drag depends on particle size, which means that cells of different dimensions settle into different equilibrium positions and can be steered into separate outlets, purely by fluid dynamics, with no labels, magnets, or lasers.

What distinguishes the new work is its geometry. The researchers built two devices sharing identical inlets, channel cross-sections, and outlet architecture, but with different spiral shapes: a classic Archimedean, or circular, spiral and an oval one. The oval design introduces an oscillation in the radius of curvature with each loop, which modulates the secondary Dean forces and, the authors hypothesized, could sharpen the redistribution of cells and particles. Both chips feature six counterclockwise loops, a rectangular channel 530 micrometers wide and 140 micrometers tall, and three outlets at the periphery. The round spiral’s channel runs 21 centimeters; the oval’s stretches to 25 centimeters, with a short radius of 6.5 millimeters and a long radius of 11 millimeters.

To model blood, the team used fluorescent polystyrene beads of three sizes: 10 micrometers to mimic platelets and red blood cells, 13 micrometers to stand in for white blood cells, and 26 micrometers to represent tumor cells. The beads were suspended in bovine serum albumin solutions chosen to reproduce the density and viscosity of whole blood, which is a non-Newtonian fluid, and the devices were tested under both symmetric flow, with two syringes each delivering 800 microliters per minute, and asymmetric flow, with a fast buffer stream of 2800 microliters per minute meeting a slower particle stream of 600 microliters per minute. Under asymmetric conditions, the round spiral concentrated 13-micrometer beads at the middle outlet with about 90 percent relative isolation, while 26-micrometer beads went to the inner outlet, and the oval chip achieved inner-outlet isolation above 97 percent for the large beads.

The researchers then diluted the carrier fluid, dropping the albumin concentration from 30 percent to 16, 8, and 4 percent to simulate the blood dilution common in other protocols. The results were telling: lower density degraded the focusing of mid-sized particles and, in the oval device, split their distribution between outlets, while large particles remained robustly focused at the inner outlet. In other words, concentrating the sample rather than diluting it preserves the hydrodynamic precision, a finding that runs counter to the widespread practice of diluting blood tenfold or more before microfluidic processing, a step that inflates processing times to hours and risks losing precious cells.

Real cells, however, are not rigid beads. When the team flowed GFP-expressing MNNG-HOS osteosarcoma cells, which average about 17 micrometers in diameter, through the round spiral, the cells behaved like the 13-micrometer particles rather than the 26-micrometer ones, likely because cellular deformability lets them squeeze under hydraulic forces that would deflect a solid bead. The picture grew more complex in whole blood spiked with tumor cells: under asymmetric flow, cells scattered between the middle and outer outlets, an inversion the authors attribute to the sheer number of blood cells overwhelming the equilibrium forces. Switching to symmetric flow restored order, and the oval spiral proved superior, directing 38,026 of 40,000 spiked cells to the inner outlet, compared with 31,657 for the round design.

The decisive breakthrough came from narrowing the channel. A redesigned oval spiral with the width cut from 530 to 320 micrometers, raising the channel’s aspect ratio from 0.26 to 0.43, sharpened the focusing of both particle classes dramatically. In mixed solutions, 13-micrometer beads isolated at the middle outlet with 97.48 percent efficiency and 26-micrometer beads at the inner outlet with 82.73 percent when the albumin concentration was 16 percent. Most importantly, when blood samples were spiked with as few as 500 osteosarcoma cells, the narrower oval chip recovered them at the inner outlet with a relative isolation index of 84.83 percent, and the isolated cells remained viable and kept proliferating in culture afterward, a critical requirement for downstream molecular analysis.

The study is candid about its limits. The team did not directly measure internal pressure or channel deformation in the soft PDMS polymer, and rigid beads cannot fully reproduce the stiffness, buoyancy, and wall interactions of living cells. Residual red blood cells still contaminate the tumor-cell fraction, a problem other groups have tackled by cascading multiple spirals or adding downstream filters. Tumor-cell heterogeneity, with CTC sizes varying by tumor type and within a single sample, means no universal size-based sorter will ever be perfectly pure. The authors suggest that stiffer materials such as polymethyl methacrylate and more physiological cell models will be needed to validate the design’s robustness.

Even so, the implications are considerable. Commercial CTC platforms such as CellSearch and Parsortix achieve good enrichment but involve numerous processing steps and high per-test costs, and most microfluidic alternatives demand red blood cell lysis, centrifugation, or massive dilution before a sample ever reaches the chip. The oval spiral device accepts minimally manipulated blood and delivers enriched tumor cells in under 30 minutes, at the low cost that soft-lithography fabrication allows. If the performance holds in patient samples across tumor types, the technology could become a standard benchtop tool for real-time monitoring of tumor evolution and treatment response, turning a routine blood draw into a window on a cancer’s changing biology.

Subject of Research: Inertial spiral microfluidic separation of circulating tumor cells from liquid biopsies

Article Title: An inertial‐based spiral device for separating rare tumor cells from liquid biopsies

Article References: Panez‐Toro, I., Griscom, L., Heymann, D., & Muñoz‐García, J. (2026). An inertial‐based spiral device for separating rare tumor cells from liquid biopsies. Bioengineering & Translational Medicine, Article e70175. https://doi.org/10.1002/btm2.70175

Image Credits: AI Generated

DOI: 10.1002/btm2.70175

Keywords: liquid biopsy, circulating tumor cells, microfluidics, inertial focusing, spiral microchannel, Dean flow, cancer detection, cell sorting, osteosarcoma, biomedical engineering, early diagnosis, PDMS

The core difficulty stems from how these images are labeled. A whole-slide image can be a gigapixel mosaic of tissue, far too large to feed directly into a neural network. The standard workaround is to slice the slide into thousands of small tiles and then classify the entire slide using only a single label, such as tumor or normal. This setup is known as multiple instance learning, or MIL. The slide is treated as a bag of instances, and the algorithm must figure out which tiles inside the bag reveal the disease. It is a weakly supervised problem: nobody tells the model exactly where the tumor is, only that somewhere in this vast tissue landscape, evidence exists.

Attention-based MIL methods, popularized by landmark architectures such as ABMIL and TransMIL, let the network assign importance weights to individual tiles, effectively letting the model highlight the most suspicious regions. But this approach carries a hidden flaw that the Nanchang team calls the spatial context dilemma. Because attention is spatially agnostic, the model can latch onto isolated artifacts, such as a folding crease, an air bubble, or a stray ink mark, and treat them as diagnostic gold. Conversely, when researchers try to inject broad global context so the model understands the whole slide, the signal can oversmooth small focal lesions or blend together biologically unrelated regions of tissue. The model ends up either too narrow or too blurry, and both failure modes degrade diagnostic accuracy.

The solution proposed by Li and Li is called Neighbor-Constrained MIL, or NCMIL, a dual-path framework that marries two complementary views of the same slide. The first path uses a Nyströmformer, an efficient transformer variant that approximates self-attention at a fraction of the computational cost, to capture global slide-level context across the thousands of tiles. The second path applies Neighbor-Constrained Attention, a mechanism that forces each tile to attend only to its physically adjacent neighbors on the slide. The key insight is that diagnostically relevant evidence in histopathology is not scattered randomly; it is spatially organized, forming coherent structures such as tumor nests, invasive fronts, and immune cell infiltrates that pathologists read as patterns, not as isolated pixels.

Three specific design choices give NCMIL its power. First, the model defines a fixed physical neighborhood around each tile and enforces it with a hard attention mask, meaning tiles outside the neighborhood are mathematically forbidden from contributing to a tile’s local representation. This preserves the tissue’s true topology rather than letting attention wander across the slide. Second, the framework performs similarity-weighted neighbor aggregation using frozen tile embeddings, which are precomputed feature vectors that do not change during training. If a neighboring tile is spatially adjacent but morphologically inconsistent, for example a fat cell next to a tumor cell, its influence is automatically suppressed. This prevents the model from averaging away the very signals that matter. Third, an adaptive local-global fusion module dynamically balances the fidelity of the local tumor microenvironment against the broader slide-level context, learning how much of each signal to trust for a given image.

What makes this approach elegant from an engineering standpoint is that it achieves spatial coherence without an explicit graph-construction stage. Earlier spatially aware methods built explicit graphs connecting tiles, which adds computational overhead and hyperparameter complexity. NCMIL instead bakes the spatial prior directly into the attention mechanism as an inductive bias, a built-in assumption about how the world is structured. The neighborhood constraint acts like a soft version of the way a pathologist’s eye moves across tissue, sampling locally while maintaining an awareness of the whole field. The frozen embeddings used for similarity weighting also mean the model can leverage powerful self-supervised or foundation-model features without retraining them, keeping the pipeline efficient.

To validate the design, the authors ran extensive five-fold cross-validation on four public histopathology benchmarks, a rigorous protocol in which the data is repeatedly split into training and testing partitions to ensure results are not a fluke of a single split. NCMIL achieved the best overall performance among all evaluated baselines, with absolute improvements of up to 1.3 AUC points, 2.2 F1 points, and 2.2 accuracy points over the strongest competitor. In a field where methods are often separated by fractions of a point, gains of this magnitude across three complementary metrics, which measure ranking quality, the balance of precision and recall, and raw classification correctness respectively, represent a meaningful advance rather than statistical noise.

The clinical significance of spatially coherent attention extends far beyond leaderboard numbers. Decades of cancer research, from the hallmarks of cancer framework to studies of the tumor immune microenvironment, have established that the arrangement of cells within tissue carries prognostic weight. The type, density, and location of tumor-infiltrating lymphocytes, for instance, can predict patient outcomes in colorectal and other cancers. An AI system that over-attends to artifacts or blurs together unrelated regions cannot capture these biologically meaningful spatial patterns. By constraining attention to respect tissue topology, NCMIL moves machine reading of slides closer to how human experts actually reason, potentially improving the reliability of automated screening, subtyping, and biomarker discovery workflows.

The work also arrives at a moment when computational pathology is scaling rapidly. Weakly supervised deep learning has already demonstrated clinical-grade performance in detecting lymph node metastases in breast cancer and predicting microsatellite instability directly from histology in gastrointestinal tumors. More recently, foundation models trained on millions of slides have pushed the field toward general-purpose pathology AI. Yet these large models still depend on downstream architectures that aggregate tile features sensibly, and the aggregation step is precisely where the spatial context dilemma bites. A better aggregator, in other words, amplifies the value of every upstream foundation model, which is why architectural contributions like NCMIL may have outsized influence as the ecosystem matures.

Limitations and open questions remain, as with any methodological advance. The fixed physical neighborhood is a strong prior, and while it protects against artifacts and oversmoothing, one might ask whether it could hinder the detection of genuinely long-range interactions, such as correlations between distant tumor regions, though the parallel global path via the Nyströmformer is designed to cover exactly that gap. The authors report no conflicts of interest, and the work was supported by the National Natural Science Foundation of China and the Natural Science Foundation of Jiangxi Province. Whether neighbor-constrained attention becomes a standard component of future pathology pipelines will depend on replication across larger and more diverse cohorts, but the study offers a compelling answer to a question every computational pathologist has faced: sometimes, to understand the big picture, a machine must first learn to look closely at its neighbors.

Subject of Research: Neighbor-constrained attention for multiple instance learning in whole-slide histopathology image classification

Article Title: Neighbor-constrained attention for multiple instance learning in whole-slide histopathology

Article References: Li, M., & Li, W. (2026). Neighbor-constrained attention for multiple instance learning in whole-slide histopathology. Medical & Biological Engineering & Computing. https://doi.org/10.1007/s11517-026-03627-8

Image Credits: AI Generated

DOI: 10.1007/s11517-026-03627-8

Keywords: computational pathology, whole-slide imaging, multiple instance learning, attention mechanism, deep learning, tumor microenvironment, cancer detection, histopathology, machine learning, weak supervision, transformers, medical imaging