In a development that could reshape how scientists study the mechanics of human movement, a team of biomedical engineers and computer scientists has unveiled an open-source software tool capable of tracking the tiny, intricate bones of the wrist as they glide and rotate during motion. The new tool, called Hierarchical 3D Registration, or 3DH, is a module within the SlicerAutoscoperM extension of the widely used 3DSlicer computing environment. Described in a study published in BioMedical Engineering OnLine, the software was validated against previously established methods using both sequential three-dimensional computed tomography, or 3DCT, scans of living participants performing thumb tasks and dynamic four-dimensional CT, or 4DCT, images of cadaveric wrists moving through flexion and extension. The results demonstrate that the open-source approach can match the accuracy of independent, previously published methodologies while offering researchers everywhere a free and reproducible path to quantifying skeletal kinematics.

The wrist has long been one of the most challenging joints for biomechanists to study. It contains eight small carpal bones, including the scaphoid and lunate, along with the radius of the forearm and the metacarpal bones of the hand, all articulating in a tightly packed arrangement that allows the remarkable range of motion humans rely on every day. Understanding how these bones move relative to one another, a field of study known as arthrokinematics, is essential for unraveling the causes of wrist instability, osteoarthritis, ligament injuries and the long-term outcomes of surgical repairs. Yet measuring those movements in living, moving joints has historically required either invasive techniques, labor-intensive manual analysis or proprietary software that few laboratories can access or verify.

Quantitative tracking of skeletal structures in vivo, with both efficiency and high temporal resolution, matters to a wide range of questions in biomechanical research, and the authors of the new study argue that the field has been held back by a shortage of accessible, semi-automated registration tools. Image registration, the computational process of aligning one three-dimensional image volume to another, lies at the heart of motion tracking from CT data. When a bone moves between two scans, its position and orientation change, and a registration algorithm must find the transformation that best overlays the bone’s image from one time point onto the next. Done well, this yields precise measurements of rotation and translation. Done poorly or inconsistently, it introduces errors that can swamp the small movements of carpal bones, which may shift by only a few millimeters and a few degrees during normal function.

The new 3DH module addresses this problem with a hierarchical strategy for registering bones across sequences of image volumes. The researchers presented and evaluated the software using two distinct classes of data drawn from previous studies. The first consisted of sequential 3DCT datasets of human participants performing various thumb tasks, in which the radius, the first metacarpal and the trapezium were the target bones. The second consisted of dynamic 4DCT datasets capturing simulated flexion-extension motion in cadaveric wrists, providing time-resolved, three-dimensional image volumes of the scaphoid and lunate in motion. By testing the tool on both static sequential scans and true dynamic imaging, the team was able to demonstrate that 3DH is not limited to a single imaging paradigm but can serve as a unified tracking solution across modalities.

To determine whether the new open-source tool could be trusted, the investigators compared the arthrokinematics computed with 3DH against values that had been previously calculated using independent approaches for the same 3DCT and 4DCT datasets. The statistical framework for this comparison was the Bland-Altman analysis, a standard method for assessing agreement between two measurement techniques. Rather than simply correlating the two sets of results, Bland-Altman analysis quantifies the mean bias, the systematic difference between methods, and the 95 percent limits of agreement, the range within which most individual differences between the methods are expected to fall. This approach is particularly well suited to method-comparison studies because it reveals whether a new technique introduces systematic error and how widely individual measurements might deviate from those of the established gold standard.

The performance figures reported in the study are striking. For the 3DCT data, in which the radius, first metacarpal and trapezium were tracked, the mean bias for the helical angle, a single-axis description of three-dimensional rotation, was just minus 0.06 degrees, with 95 percent limits of agreement ranging between minus 1.87 degrees and 1.74 degrees. Translation measurements fared nearly as well: the mean bias was 0.12 millimeters, with limits of agreement spanning from minus 2.23 millimeters to 2.00 millimeters. In practical terms, the open-source tool reproduced the rotational and translational measurements of the earlier methods to within roughly two degrees and two millimeters at the extremes, differences small enough to preserve the biological signal in most wrist kinematics experiments.

The dynamic 4DCT results told a similar story for the carpal bones. Tracking the scaphoid and lunate through simulated flexion-extension, 3DH produced mean biases of the Euler angle rotations, which decompose three-dimensional rotation into three orthogonal components, of less than 0.77 degrees about any direction. The widest 95 percent limits of agreement ranged between minus 7.11 degrees and 8.65 degrees. For translations, mean biases were less than 0.5 millimeters in any direction, with the widest limits of agreement ranging between minus 2.36 millimeters and 1.43 millimeters. The wider limits of agreement for rotations in the dynamic data likely reflect the added challenges of time-resolved imaging, including motion blur and the smaller, more complexly shaped carpal bones, but the near-zero mean biases indicate that the tool does not systematically drift away from established values.

Crucially, all target bones were successfully tracked in both the 3DCT and 4DCT datasets, meaning the hierarchical registration algorithm converged on correct solutions in every case tested. This robustness matters because registration algorithms can occasionally fail, locking onto incorrect alignments or diverging entirely, and such failures typically demand time-consuming manual intervention. A semi-automated tool that reliably succeeds across diverse bones, motion tasks and imaging protocols removes a significant bottleneck. Manual bone tracking in CT volume sequences can take hours per bone per trial, and the labor multiplies rapidly in studies involving multiple participants, multiple bones and dozens of motion cycles. By automating this process within a validated framework, 3DH promises to accelerate the pace at which wrist biomechanics data can be collected and analyzed.

The choice to build the module inside 3DSlicer, and specifically within the SlicerAutoscoperM extension, is a deliberate strategic decision with implications well beyond the wrist. 3DSlicer is a free, open-source platform for medical image computing and visualization with a large international user base spanning research, clinical and educational communities. By embedding 3DH in this ecosystem, the developers have made the tool immediately available to any laboratory already working with 3DSlicer, complete with the platform’s established infrastructure for loading DICOM data, segmenting anatomical structures and visualizing results. The authors note that the ability to track bones from both 3DCT and 4DCT data sources, with agreement to prior methodologies, demonstrates that the 3DH approach is a capable tool that could promote collaboration and consistency of results through open-source software.

Consistency is a quietly urgent issue in biomechanics. When different laboratories use different proprietary tools with undisclosed algorithms, it becomes difficult to distinguish true biological differences between study populations from mere methodological discrepancies. Open-source registration changes that calculus: the code can be inspected, verified, extended and reused, and results from different institutions become directly comparable because they are produced by identical algorithms. This transparency also lowers the barrier to entry for research groups in settings that cannot afford commercial licenses, potentially broadening the global community capable of conducting rigorous skeletal kinematics research.

The study itself emerged from a collaboration spanning several institutions. The team included Joseph J. Crisco, Amy M. Morton and John D. Holtgrewe of the Department of Orthopedics at The Warren Alpert Medical School of Brown University and Rhode Island Hospital; Cesar Lopez, Andrew Thoreson and Kristin D. Zhao of the Assistive and Restorative Technology Laboratory and the Department of Physical Medicine and Rehabilitation at the Mayo Clinic in Rochester, Minnesota; and Jean-Christophe Fillion-Robin and Beatriz Paniagua of Kitware Inc. in Clifton Park, New York. The involvement of Kitware, a company known for its leadership in open-source scientific computing, underscores the blend of academic rigor and software engineering expertise required to produce a research tool robust enough for widespread distribution. Funding was provided by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under award numbers AR059185, AR078924 and AR071338.

The implications of this work extend naturally to the future of musculoskeletal imaging. As newer CT technologies, including photon-counting detector systems, deliver ever-finer spatial resolution and as dynamic 4DCT becomes more widely available, the volume of image data available for kinematic analysis will grow dramatically. Tools like 3DH, which are fast, semi-automated and validated across modalities, will be essential for converting that imaging bounty into quantitative biomechanical knowledge. For patients with wrist disorders, the downstream promise is equally tangible: better measurements of how carpal bones actually move in living joints can inform more accurate diagnoses, improved surgical planning, more realistic computational models of joint loading and ultimately better-designed treatments for conditions ranging from scapholunate instability to thumb-base arthritis. With the tool now openly available in the 3DSlicer environment, researchers around the world can begin putting it to work.

The Denisovans, the enigmatic archaic human relatives first identified from a finger bone and a handful of teeth in a Siberian cave, have long been known almost exclusively through their DNA. Now, a landmark study published in Nature has pulled back the curtain on how these ancient hominins actually lived — what they ate, how they hunted, what tools they made, and the environments they braved — through an extraordinary assemblage of fossils and artefacts from Bianfu Cave in Yunnan Province, southwestern China. The findings paint the most detailed portrait to date of Denisovan biology, behaviour and ecology in eastern Asia, and suggest that these vanished people left a deep genetic and cultural imprint on later populations across Southeast Asia and Oceania.

The research team, led by Qingfeng Ruan, Hua Li and Song Xing and their colleagues, conducted a comprehensive analysis of hominin fossils and the associated archaeological remains recovered from Bianfu Cave. The hominin fossils themselves date to roughly 167,000 to 134,000 years ago, placing them firmly in the Middle Pleistocene. But the cultural sequence preserved in the cave’s sediments is far longer, stretching from approximately 190,000 years ago to about 70,000 years ago. That span — some 120,000 years of continuous or near-continuous occupation — makes Bianfu Cave one of the longest cultural records that can most parsimoniously be attributed to Denisovans anywhere in the world.

The Denisovan identity of the fossils was established through two independent and technically demanding lines of evidence. The first involved analysis of the enamel–dentine junction, the internal boundary between the enamel cap and the underlying dentine of the teeth. This three-dimensional morphological interface acts something like a fingerprint of evolutionary relationships, revealing diagnostic traits of the tooth’s internal architecture that are invisible from the outside but characteristic of particular hominin lineages. The second line of evidence came from palaeoproteomics, the recovery and identification of ancient proteins preserved within fossilized tissues. Because proteins are far more abundant and durable than DNA, they can survive in warm, humid environments like those of southern China, where genetic material degrades rapidly. Sequencing these ancient proteins allowed the researchers to place the Bianfu hominins on the Denisovan branch of the human family tree with confidence.

The fossil collection itself is remarkable. It includes four teeth that substantially expand the known range of dental variation within the Denisovan lineage — a critical contribution, since Denisovans were previously represented by a very small number of dental specimens, making it difficult to gauge how variable their teeth could be. The new fossils also include two cranial fragments and a radius, a forearm bone, providing rare glimpses of Denisovan skull and postcranial anatomy that were previously almost entirely unknown. Together, these remains transform Denisovans from a genome and a scatter of isolated teeth into a population with recognizable skeletal biology.

Beyond the fossils, the archaeological record from Bianfu Cave offers something equally precious: a window into Denisovan daily life. The researchers reconstructed the ancient environment using faunal remains and pollen preserved in the cave deposits. These records indicate that across Marine Isotope Stages 6 through 4 — a period encompassing major glacial–interglacial climatic swings between roughly 190,000 and 57,000 years ago — the landscape around the cave was dominated by coniferous forest or a mosaic of forest and steppe. This was a cool, variable environment that Denisovans exploited successfully across tens of thousands of years, demonstrating a resilience to climatic change that adds to growing evidence of their adaptability, from the high altitudes of the Tibetan Plateau to the subtropical latitudes of mainland Southeast Asia.

The subsistence evidence is particularly striking. Analysis of the animal bones from the cave shows that its inhabitants practised specialized hunting focused on medium- to large-bodied prey. This was not opportunistic scavenging but a sustained, targeted hunting strategy — one that would have required knowledge of animal behaviour, coordinated effort and effective weaponry. Specialized big-game hunting of this kind has often been associated with modern humans and late Neanderthals in western Eurasia, so finding it in a Denisovan context more than 150,000 years ago dramatically narrows the perceived behavioural gap between these archaic lineages and our own species.

Yet the technological strategy documented at Bianfu Cave is, at first glance, surprisingly simple. The stone tool assemblage is characterized by expedient core reduction and tool production — that is, the inhabitants struck flakes from cores as needed, used them, and discarded them, rather than investing in carefully shaped, curated tools that were carried, maintained and resharpened over time. Alongside this, the Denisovans of Bianfu Cave made pervasive use of unmodified bones, employing skeletal elements of their prey directly as tools without altering them beforehand.

Taken together, the researchers interpret this pattern as an adaptive system that prioritized the exploitation of what they call object affordance — the inherent functional possibilities already present in a raw material or object — over intensive tool manufacture. A suitably shaped bone fragment could serve as a hammer, a digging implement or a scraping tool with little or no modification; a sharp flake struck from a convenient core could cut and butcher without further refinement. In the rich, resource-diverse environments of subtropical southwestern China, where raw materials and animal resources were plentiful, such a flexible, low-investment strategy may have been not a sign of simplicity but a highly efficient way of life. It suggests that Denisovans measured success not by the sophistication of their toolkit but by how skilfully they read and exploited the opportunities their landscape provided.

This behavioural portrait gains added significance when set against the wider geography of Denisovan presence. Over the past decade and a half, Denisovans have been documented across a staggering range: from the original genomic discoveries in the Altai Mountains of Siberia, first reported in 2010, to the mandible from the Tibetan Plateau described in 2019, Denisovan DNA recovered from sediments at Baishiya Karst Cave in 2020, a molar from the Annamite Chain of northern Laos announced in 2022, a male Denisovan mandible from Pleistocene Taiwan reported in 2025, and proteomic identification of the 146,000-year-old Harbin cranium, also in 2025. Bianfu Cave now anchors this story firmly in southwestern China, a region long suspected to have been a Denisovan heartland and a critical corridor for hominin movements between mainland Asia and Southeast Asia.

The study’s authors argue that the region’s Denisovans point to a substantial genetic and cultural legacy in later human populations. Modern populations in Southeast Asia and Oceania carry the highest known proportions of Denisovan ancestry anywhere in the world — a legacy of interbreeding that occurred as modern humans moved through and beyond island Southeast Asia. The new evidence from Bianfu Cave suggests that the ways of life practised by Denisovans in this region — their hunting strategies, their flexible technological approaches, their hard-won knowledge of tropical and subtropical environments — may have been part of the inheritance, alongside their genes, that helped later human populations thrive as they spread across the diverse and challenging landscapes of Asia and the Pacific.

For palaeoanthropologists, the significance of the Bianfu Cave discoveries is difficult to overstate. Denisovans have been the least anatomically documented of all known hominin groups — a lineage defined first by DNA and only slowly fleshed out with fragmentary fossils. The new teeth, cranial fragments and radius from Yunnan, combined with an unmatched 120,000-year archaeological sequence, mean that researchers can finally study Denisovans not just as a genome or a set of molecular signatures but as a living, adapting population: hunters of large game, makers and users of expedient tools, and inhabitants of conifer forests that swung between glacial cold and interglacial warmth. As analysis of the Bianfu material continues, it promises to reshape our understanding of how archaic humans lived in eastern Asia and how deeply their story is woven into our own.

A major trial found that implantable cardioverter-defibrillators may benefit some patients younger than 70 with mild-to-moderate heart dysfunction and cardiac scarring. A more than decade-long randomized trial suggests that implantable cardioverter-defibrillators may benefit some patients who have heart scarring and only mild-to-moderate loss of heart function, a group that currently falls outside standard treatment guidelines. [...]

Those who remember PC gaming in the days before Steam may have fond memories of the "big box" packaging that made those games stand out as shelf eye candy and sought-after in-home collectibles. Today, PC gaming platform GOG is hoping to reignite some of the nostalgia for those days by adding downloadable 3D models and printable, foldable high-res packaging scans for a handful of PC games on the platform.

Thus far, the "Printable Big Box" collection on GOG is limited to just five games: Armikrog, Descent, Hitman: Codename 47, Myst, and Star Trek: 25th Anniversary. Players browsing those games on the GOG store will now see "3D Boxes" listed among the included "Goodies" before purchase. Owners of those games on GOG can download a ZIP file containing a version of the box art as an animated 3D object file (in an easy-to-open GLB format) and a "flat printable template" in PDF, SVG, and Adobe Illustrator formats. Those downloads allow for what GOG calls "a proper papercraft project for anyone who misses shelves full of PC game boxes."

The 3D model animates to show the front flap opening. Credit: GOG

"Physical media doesn’t survive the way software can—boxes get lost, crushed, thrown out," GOG writes in today's announcement. "Scanning them, and letting you print your own, is preservation in a different shape: the same mission that keeps these games running on modern PCs, applied to the objects they used to come in."

Read full article

Comments

© GOG

Depression in later life may be an early warning from the brain, appearing years before Alzheimer’s begins to affect memory and thinking A new study in JNeurosci found that faster accumulation of tau, a protein closely tied to Alzheimer’s, was associated with worsening depressive symptoms in older adults who remained cognitively healthy. Teodora Markova of [...]

Pat Fauquet is the next stop onThe Road to Macstock is a conversation about her session, “Pills, Processes, and Procedures,” explaining how Apple devices can help manage medications, appointments, research, and communication with doctors. Drawing from years of health challenges and tech experience, she highlights the Apple Watch, iPhone, iPad, and Mac as practical tools for personal advocacy, caregiving, and preparedness. This might be the most important session this year. 

💾

The Road to Macstock takes a serious turn as Mike T. Rose previews his workshop on becoming a digital caretaker for family members and loved ones. Mike discusses scams targeting older adults, password sharing, remote support, Apple Passwords, financial preparedness, and the need to balance protection, respect, independence, and trust when helping others manage their digital lives. 

💾

Expansion microscopy has spent the past decade changing the rules of super-resolution imaging. Instead of relying solely on increasingly sophisticated optics, the technique physically enlarges biological specimens so that molecules separated by nanometers become easier to distinguish with ordinary fluorescence microscopes. Now, researchers have introduced a method designed to push that concept toward one of the most difficult goals in microscopy: seeing the shapes of individual proteins directly.

The approach, described by Ahmed H. Shaib, Mahmoud M. Alawieh and Stefan O. Rizzoli in Nature Protocols, combines one-step nanoscale expansion microscopy with fluctuation-based super-resolution analysis. The resulting workflow, called ONE microscopy, is intended to make molecular-scale imaging more accessible to laboratories that do not have access to cryo-electron microscopes or specialized high-end optical systems.

Traditional light microscopy is limited by diffraction, a physical effect that causes light from two nearby objects to blur together when they are too close. Even modern super-resolution methods can require complex instruments, intense labeling strategies or extensive computational processing. Expansion microscopy takes a different route. Researchers anchor biological molecules within a swellable polymer network and then expand the material, increasing the physical distance between fluorescent labels.

The expansion step does not automatically reveal every molecular detail. A protein’s structure can still be represented by only a small number of fluorescent signals, and conventional images may contain noise, background fluorescence and motion-related fluctuations. ONE microscopy addresses these limitations by analyzing changes in fluorescence intensity over time. These fluctuations contain information about the presence, position and behavior of labeled molecules that may not be obvious in a single frame.

In practical terms, the method links chemical preparation, physical enlargement and computational analysis into a single workflow. Samples are embedded in a polymer gel, labeled to identify the proteins of interest and then expanded. After expansion, researchers acquire image sequences using conventional fluorescence equipment. Specialized software analyzes the temporal variation in the recorded signals, extracting spatial information beyond what a standard diffraction-limited image would provide.

This combination is significant because it shifts the focus from simply locating a protein to examining its overall shape. For many biological questions, knowing that a protein is present is not enough. Its size, orientation and structural organization can determine how it interacts with membranes, vesicles, organelles or neighboring proteins. Directly observing those features could help researchers investigate molecular machines in their native cellular environments rather than relying exclusively on purified samples or averaged structural models.

The protocol is designed to work across a broad range of biological materials, including purified proteins, cultured cells and tissues. That flexibility could make the technique useful for researchers studying protein organization at multiple scales. A purified protein might provide a controlled test of shape reconstruction, while cells and tissues could reveal how the same protein is arranged amid the crowded and complex environment of living biology.

A major part of the reported advance is the accompanying software package. Fluctuation-based imaging can be powerful, but its usefulness depends on reliable data processing, and computational analysis has often been a barrier for non-specialist users. The authors present the software as stable and user-friendly, with the goal of making the analysis more efficient, reproducible and practical for laboratories using standard fluorescence microscopes.

ONE microscopy does not replace cryo-electron microscopy, which remains capable of resolving structures at atomic or near-atomic scales under appropriate conditions. Nor does it eliminate the challenges associated with labeling, gel chemistry, image quality and sample preparation. Expansion can introduce distortions, and successful imaging depends on preserving the relationship between the fluorescent labels and the underlying structures. Nevertheless, the method offers a complementary strategy: rather than averaging thousands or millions of molecules, it aims to examine individual protein shapes through expanded, fluorescence-labeled specimens.

The researchers describe the workflow as a practical framework for protein imaging on conventional equipment. By combining physical separation of fluorophores with information extracted from fluorescence fluctuations, the method brings nanoscale structural analysis closer to routine biological imaging. Its broader impact may come not from replacing existing forms of super-resolution, but from making a previously specialized capability more reproducible and attainable for laboratories investigating how individual proteins operate inside cells and tissues.

Subject of Research: One-step nanoscale expansion microscopy for visualizing individual protein shapes using conventional fluorescence microscopes.

Article Title: ONE microscopy.

Article References: Shaib, A.H., Alawieh, M.M. & Rizzoli, S.O. ONE microscopy. Nature Protocols (2026). https://doi.org/10.1038/s41596-026-01399-x

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41596-026-01399-x

Keywords: expansion microscopy, ExM, nanoscale imaging, super-resolution microscopy, fluctuation-based analysis, protein structure, fluorescence microscopy, single-protein imaging, cryo-electron microscopy, biological imaging

Foundation models are promising to become the “universal translators” of biomedical data—but a new perspective argues that medicine should resist treating them as all-knowing digital doctors. In a review published in Nature Biomedical Engineering, researchers describe how large, adaptable artificial intelligence systems are reshaping biomedical imaging while warning that impressive benchmark scores can conceal serious weaknesses in clinical environments. Their central message is direct: foundation models are most likely to improve healthcare by augmenting specialists, not replacing them.

Foundation models are designed to learn broad representations from enormous and diverse datasets before being adapted to specific tasks. In biomedical imaging, those data may include magnetic resonance imaging, computed tomography, X-rays, ultrasound, digital pathology slides and ophthalmic images. The same underlying model could theoretically identify tumors, segment organs, estimate disease risk, retrieve similar cases and generate clinical reports. The vision becomes even broader when imaging is combined with pathology, electronic health records and genomic information, creating a composite system intended to analyze a patient across multiple biological scales.

That ambition reflects a major change from traditional medical AI. Earlier systems were generally built for one narrowly defined task, such as detecting pneumonia on chest radiographs or outlining a brain lesion on MRI scans. Foundation models instead attempt to create a reusable backbone that can be transferred across diseases, hospitals and imaging technologies. Technically, they learn statistical patterns in high-dimensional data, often through self-supervised training in which the system predicts missing, transformed or associated information. Afterward, developers fine-tune or prompt the model for particular clinical applications.

But biomedical imaging is not a single, uniform world. Images differ according to scanner manufacturer, acquisition protocol, patient population, disease prevalence and local clinical practice. A model trained primarily on data from large academic hospitals may perform very differently in community clinics, rural settings or countries with limited equipment. This problem, known as domain shift, can occur even when images appear visually similar. Small changes in image quality, contrast, hardware or patient demographics may alter the statistical distribution learned by the model and undermine its predictions.

The authors introduce a framework called real-world evaluation and assessment of foundation models, or REAL-FM, to examine whether these systems are ready for practical use. Rather than focusing only on accuracy scores, REAL-FM considers several dimensions at once, including the quality and representativeness of training data, technical readiness, clinical value, integration into existing workflows and responsible artificial intelligence. The framework is intended to help clinicians interpret claims about new models and to push developers toward evaluations that resemble actual medical practice.

One of the sharpest distinctions in the perspective is between pattern recognition and causal reasoning. Foundation models can become extraordinarily skilled at recognizing visual associations—for example, linking a particular texture or anatomical feature with a diagnosis present in their training data. Yet association is not the same as understanding why a disease occurs, how it will progress or whether a treatment will benefit an individual patient. A model may identify a correlation that is valid in one hospital but reflects a hidden confounder, such as a scanner type, reporting convention or patient-selection pattern, rather than a biological signal.

This limitation becomes especially dangerous when models are moved beyond simplified benchmarks. Many benchmark datasets provide carefully curated images, clear labels and narrowly defined tasks. Clinical care is messier: scans can be incomplete, diagnoses can be uncertain, records may contain contradictory information and several conditions may coexist. The perspective highlights a lack of verified generalization across such conditions, along with a shortage of prospective, outcome-based validation. In a prospective study, an AI system would be evaluated while care is actually being delivered, with researchers measuring whether it improves diagnostic accuracy, treatment decisions, patient outcomes or workflow efficiency.

Data scarcity is another obstacle. The largest foundation models require vast quantities of data, but medical information is difficult to collect, standardize and share. Patient records are fragmented across institutions, imaging data may lack reliable annotations and rare diseases are inherently underrepresented. Privacy regulations and governance requirements further complicate the creation of centralized datasets. As a result, a model may appear broadly capable while remaining poorly tested in the populations and clinical circumstances where errors could have the greatest consequences.

The researchers argue that human oversight therefore remains indispensable. In practical terms, foundation models may be most useful as clinical assistants that prioritize images for review, highlight suspicious regions, summarize longitudinal records or offer a second opinion that specialists can interrogate. Safe deployment will require transparent reporting of uncertainty, monitoring for performance drift, mechanisms for correcting errors and clear accountability when recommendations influence care. The future envisioned by the authors is not a single monolithic medical oracle, but a coordinated ecosystem of specialized AI systems, each evaluated for a defined clinical role and connected to expert-led workflows.

The perspective arrives as the biomedical AI field races toward increasingly general-purpose systems. Its warning is not that foundation models lack value, but that technical scale alone cannot establish clinical reliability. Before these models can be trusted with consequential decisions, they must demonstrate robustness across domains, usefulness in real workflows, safety under unexpected conditions and measurable benefits for patients. The proposed REAL-FM framework offers a way to separate viral demonstrations from durable medical progress—by asking not only what a model can recognize, but where it works, why it works and whether it makes care better.

Subject of Research: Foundation models and their real-world evaluation in biomedical imaging and clinical medicine.

Article Title: Foundation models in biomedical imaging: turning hype into reality

Article References: Muneer, A., Zhang, K., Hamdi, I. et al. Foundation models in biomedical imaging: turning hype into reality. Nature Biomedical Engineering 10, 1557–1575 (2026). https://doi.org/10.1038/s41551-026-01762-z

Image Credits: AI Generated

DOI: 10.1038/s41551-026-01762-z

Keywords: foundation models, biomedical imaging, medical artificial intelligence, clinical AI, multimodal AI, domain shift, causal reasoning, clinical validation, responsible AI, REAL-FM

Schizophrenia may leave a distinctive biological footprint across the brain’s communication network. Using specialized PET imaging, researchers have mapped where synaptic connections are most reduced in people with schizophrenia and identified a possible region from which the damage begins to spread. The study, which involved a Rutgers professor, was published in Molecular Psychiatry. Synapses are [...]

How we sleep may have lasting impacts for our brain health as we age. A new University of

Astronomers have found a third giant planet around Beta Pictoris, one of the best-known young planetary systems near Earth. The discovery is striking because the planet was not first picked out as a bright dot in a telescope image. NASA’s James Webb Space Telescope found it in the spectrum of light coming from the system: […]

The post Webb Found a Hidden Planet in Beta Pictoris By Reading Its Atmosphere appeared first on Curiosmos.

Reaching 90 without dementia may feel like clearing the final hurdle. But the risk does not necessarily end there. Dementia risk does not disappear after age 90. Even among people who reach their 10th decade without cognitive symptoms, the likelihood of developing the condition can vary substantially by sex, race, ethnicity, and genetics, according to [...]

The protein boom may be pushing many people to consume far more than their bodies need, and new research suggests that excess could come at a hidden cost. Protein has become one of the food industry’s favorite selling points. It is added to cereal, coffee, snack bars, desserts, and even water, often with the suggestion [...]

Changing what you eat may begin reshaping how your genes behave in as little as a few weeks, even though your DNA itself remains unchanged. In a randomized clinical trial, researchers found that healthy adults who followed a vegan diet for one month developed changes in chemical markers attached to their DNA. Compared with participants [...]

A study of more than 5,500 people suggests that monitoring hormone levels and managing weight may help reduce the risk of premature death. For adults over 50, excess abdominal fat and low vitamin D levels may form a particularly dangerous combination. A study of more than 5,500 people found that participants with both conditions had [...]

The wildfires afflicting communities across the world can now be tracked more easily by anyone using Europe’s free and open satellite imagery service—all thanks to a wildfire visualization script originating from nearly a decade ago.

The Copernicus Browser provides free access to the latest satellite imagery taken by the European Space Agency’s Sentinel satellites for the European Union’s Earth observation program. The three satellites of the Sentinel-2 mission have proven especially useful for wildfire monitoring by providing multi-spectral imaging at resolutions as granular as 10 meters in visible light and near-infrared bands—and the Copernicus Browser previously offered 10 default visualization layers using different combinations of spectral bands.

Now Copernicus Browser has also added a straightforward “wildfires” visualization layer for Sentinel-2 imagery that went live on August 4. Anyone clicking on this layer while looking at a wildfire location can see active fires in white or yellow, burning vegetation in red, and burned landscapes in dark brown or black.

Read full article

Comments

© European Union | Copernicus Sentinel-2

The wildfires afflicting communities across the world can now be tracked more easily by anyone using Europe’s free and open satellite imagery service—all thanks to a wildfire visualization script originating from nearly a decade ago.

The Copernicus Browser provides free access to the latest satellite imagery taken by the European Space Agency’s Sentinel satellites for the European Union’s Earth observation program. The three satellites of the Sentinel-2 mission have proven especially useful for wildfire monitoring by providing multi-spectral imaging at resolutions as granular as 10 meters in visible light and near-infrared bands—and the Copernicus Browser previously offered 10 default visualization layers using different combinations of spectral bands.

Now Copernicus Browser has also added a straightforward “wildfires” visualization layer for Sentinel-2 imagery that went live on August 4. Anyone clicking on this layer while looking at a wildfire location can see active fires in white or yellow, burning vegetation in red, and burned landscapes in dark brown or black.

Read full article

Comments

© European Union | Copernicus Sentinel-2