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. [...]

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

Medical imaging plays a pivotal role in today’s healthcare system. CT scans, ultrasounds, X-rays, MRIs, and other diagnostic

Thousands of years ago, a striking blue pigment began appearing in the art that adorned the walls of ancient Egyptian tombs, and eventually on the crown of the famous bust of Nefertiti, wife of the 18th dynasty pharaoh Akhenaten and half-sister to one of Ancient Egypt’s most famous kings: Tutankhamun.

However, the ancient artists who first used this remarkable pigment, known today as “Egyptian blue,” had likely never guessed that thousands of years into the future, this unique substance would have applications in modern biomedical research.

That is precisely what one team of researchers discovered several years ago, in a study that revealed how this striking pigment is more than just a modern medical marvel with origins in the ancient world—it also provides evidence of ancient nanotechnology in early Egypt.

The Enigma of Egyptian Blue

“Egyptian blue” is the common term for a pigment first developed in ancient Egypt around 2500 B.C., made from calcium copper silicate. Originally, it was created to provide an artificial alternative to a more costly form of blue pigment from the era, which was produced using lapis lazuli.

However, ancient Egyptian scientists discovered that by heating limestone, sand, and copper, they could produce a far less expensive form of blue pigment, and soon its use spread throughout the ancient Mediterranean world, making its way into the craftsmanship in parts of the ancient world that ranged from Mesopotamia to Greece and Rome.

An Ancient Technology Finds Modern Biomedical Applications

Although Egyptian blue’s obvious practical uses in art were derived from its unique and stunning appearance, time would eventually reveal that it had other uses. Specifically, in a 2020 study published in Nature Communications, a team of researchers reported on their discovery that Egyptian blue in its powdered form can be separated into nanoscale sheets roughly 100,000 times thinner than a human hair.

This is significant because these tiny particles of the substance exhibit significant fluorescence under near-infrared light, which can be used by researchers to track them within biological samples.

Chemist Sebastian Kruss of Georg-August University Göttingen, the lead author of the 2020 study, remarked at the time that he was certain such capabilities “will be of growing interest for material scientists as well as biomedical research” in the years ahead.

Ancient Nanotech’s Survival Through the Centuries

One reason the pigment is so useful is also the likely reason behind its discovery, and that is Egyptian blue’s durability. For many millennia, samples of this unique material have managed to survive, and the material’s nanoscale form remains stable even over such long periods.

In the past, earlier studies had demonstrated that the material fluoresces in near-infrared wavelengths. Additionally, it was known to scientists that the pigment could be broken down into flakes after being stirred in hot water for periods of several days.

Building on this knowledge, the research detailed in the 2020 paper sought to go even further by examining whether those particles could function as fluorescent markers in living systems.

Near-Infrared Applications

One thing that scientists find useful about near-infrared imaging is that it provides an effective way to examine biological material by leveraging wavelengths that fall just beyond the ranges of the visible spectrum.

Although some molecules can already be used as fluorescent markers, Egyptian blue nanoparticles generate significantly brighter signals than several common alternatives. Additionally, unlike other fluorescent compounds, the pigment can resist photobleaching, which is important because this allows its glow to persist even during prolonged observation.

A pair of experiments was used by the team to demonstrate the material’s potential. The first involved using Egyptian blue particles injected within a fruit fly embryo, which were tracked as it developed. In another, the particles were added to plant leaves, allowing comparisons with the performance of a conventional near-infrared imaging molecule.

In this experiment, the team found that the Egyptian blue produced a bright signal that remained visible even against the leaves’ natural fluorescence—the comparison molecule, by contrast, was heavily obscured.

According to Kruss, the tiny Egyptian Blue particles prove to be “very useful glowing labels in biomedical research,” calling the pigment a “very powerful fluorophore.”

An Ancient Discovery Breaking New Ground in the 21st Century

According to Kruss and his colleagues, such particles could eventually help scientists observe processes including embryonic development and cell division, while potentially supporting applications such as image-guided surgery and the labeling of cancerous tissue.

Altogether, this enigmatic form of ancient nanotechnology is still playing a role in helping develop the cutting-edge biomedical technologies of tomorrow, even more than 5,000 years after its initial discovery.

Kruss and his team’s original paper from 2020, “Exfoliated near infrared fluorescent silicate nanosheets for (bio)photonics,” was published in Nature.

Micah Hanks is the Editor-in-Chief and Co-Founder of The Debrief. A longtime reporter on science, defense, and technology with a focus on space and astronomy, he can be reached at micah@thedebrief.org. Follow him on X @MicahHanks, and at micahhanks.com.

A negative scan does not always mean recurrent prostate cancer is nowhere to be found. Sometimes, the disease may simply be too small or too faint to appear yet. New research suggests that repeating a PSMA PET scan later can reveal cancer that was missed the first time. Among patients with rising prostate-specific antigen (PSA) [...]

Often hailed as “the window to the soul,” the human eye has long fascinated scientists and philosophers alike. However, beyond its poetic allure, emerging research suggests that the eyes provide a vital window into the health of the brain. A groundbreaking new study spearheaded by the University of Florida’s biomedical engineering department has unveiled a transformative approach to predicting Alzheimer’s disease risk factors by analyzing retinal photographs with cutting-edge deep learning algorithms. This development promises to revolutionize early detection and intervention for one of the most devastating neurodegenerative diseases.

Traditionally, diagnosing Alzheimer’s disease has been a challenge due to its insidious onset and prolonged development over decades. Most diagnostic protocols focus on detecting the disease in its later stages, by which time significant, irreversible brain damage has already occurred. Recognizing this limitation, Dr. Ruogu Fang and her interdisciplinary team leveraged advances in artificial intelligence and biomedicine to explore the retina—a part of the central nervous system accessible via non-invasive imaging—as a potential biosensor for Alzheimer’s risk.

The study capitalized on a rich database of over 40,000 retinal images collected from the UK Biobank, a comprehensive health resource with longitudinal patient data. Retinal photography is commonly conducted during routine eye exams, diabetes management, glaucoma monitoring, and cataract assessments. By harnessing these widely available, inexpensive images, the research team circumvented the need for costly, invasive, or logistically complex diagnostic tools such as MRI or PET scans, thus holding promise for scalable, population-wide screening.

Employing deep learning, a subset of machine learning characterized by neural networks modeled after the human brain, the researchers developed an AI framework that scrutinized retinal morphology. This model was able to discern subtle variations in retinal arteries, veins, and the optic nerve head, features imperceptible to the human eye but correlated strongly with known risk factors for Alzheimer’s disease. Such morphological markers serve as proxies for neurovascular health, which is crucial given Alzheimer’s recognized association with vascular impairment.

One of the most striking revelations of this research was the model’s capacity not only to predict intrinsic biological traits such as sex and blood pressure but also modifiable lifestyle factors that influence Alzheimer’s risk. The AI effectively inferred behaviors like smoking, alcohol consumption, and even insomnia from retinal images alone. This objective measurement circumvents the notorious unreliability of self-reported lifestyle data, often plagued by bias and inaccuracies, thereby enriching risk assessment precision.

As Dr. Fang points out, retinal morphology functions less like a mere clinical questionnaire and more like an integrative biological sensor reflecting a cumulative lifetime burden of neurovascular and lifestyle insults. This conceptual shift underscores the retina’s role in encapsulating decades of physiological data, transforming retinal imaging into a powerful tool for capturing both present health and future vulnerability.

Furthermore, the implications extend beyond risk prediction. Identifying individuals with early signs of retinal changes could trigger timely clinical interventions long before cognitive symptoms manifest. Such preclinical detection enables the implementation of protective lifestyle adjustments, pharmacological therapies, or cognitive training programs designed to delay or mitigate Alzheimer’s progression.

The pioneering work built on previous findings from Fang’s group that demonstrated retinal images’ utility in detecting active Alzheimer’s disease cases. However, this new model pioneers the field by focusing on prodromal markers and risk factors instead of established dementia, bridging a critical gap in neurodegenerative diagnostics.

Importantly, the use of retinal photographs democratizes neurovascular health monitoring. Since retinal imaging is non-invasive, accessible, and affordable, it offers the prospect of integrating brain health assessment into routine eye care, thus expanding preventive neurology beyond specialized clinical settings.

This research also exemplifies the synergy between computational science and biomedical engineering. By combining sophisticated image analysis algorithms with large-scale patient data, it transcends traditional diagnostic constraints, marrying technology with medicine in a profoundly impactful manner.

While challenges remain, including the need for further validation across diverse populations and integration with other biomarkers, this study sets a new paradigm for early Alzheimer’s disease risk stratification. The possibility of preemptive interventions guided by retinal AI analysis heralds a future where the devastating trajectory of Alzheimer’s may be altered through timely and personalized healthcare.

In summary, the novel application of deep learning to retinal photographs opens an unprecedented window into brain health, heralding a new era of precision neurodiagnostics. It moves us closer to the long-sought goal of identifying Alzheimer’s risk decades in advance, fostering opportunities for early intervention and ultimately, better patient outcomes.

---

Subject of Research: People

Article Title: Prediction of Alzheimer’s disease risk factors from retinal images via deep learning: Development and validation of biologically relevant morphological associations in the UK Biobank

News Publication Date: 16-Jun-2026

Web References:
DOI: 10.1177/13872877261457650

---

Keywords

Alzheimer’s disease, Neurodegenerative diseases, Neurological disorders, Eye, Retina, Artificial intelligence, Image analysis, Deep learning, Biomedical engineering, Neurovascular integrity, Retinal biomarkers, Predictive modeling