The new study tackles a problem that has quietly undermined many apparently successful AI systems for lung sound classification: diagnostic imbalance. On the widely used ICBHI 2017 benchmark, a collection of 920 annotated recordings from 128 patients totaling about 5.5 hours of audio, the classes are severely skewed, recording conditions vary widely, and sampling rates are inconsistent. Most prior models have chased high overall accuracy or average scores while quietly sacrificing one side of the diagnostic equation. When a model prioritizes sensitivity, it catches nearly every pathological sound but floods clinics with false alarms; when it prioritizes specificity, it avoids false positives but risks missing the very crackles and wheezes that signal disease. Either failure mode introduces systematic bias, over-detecting or under-detecting illness even when headline metrics look impressive.

To correct this, the researchers built a hybrid architecture that fuses two complementary ways of reading a lung sound. The raw audio is first resampled to a uniform 10 kilohertz, filtered with a zero-phase tenth-order Butterworth band-pass filter between 50 and 2000 hertz, segmented into fixed five-second windows, and converted into log-Mel spectrograms of 128 frequency bands, resized to 224 by 224 pixels so the audio becomes something an image model can digest. A convolutional neural network branch, with five-by-five and three-by-three convolutional kernels followed by max-pooling, then extracts fine-grained local spectro-temporal textures, precisely the transient signatures of crackles and the musical sustained patterns of wheezes. A Swin Transformer branch processes the same spectrogram through hierarchical, non-overlapping windows using shifted-window multi-head self-attention, capturing long-range contextual dependencies across the entire respiratory cycle that convolutions alone tend to miss.

The choice of the Swin Transformer, rather than a conventional Vision Transformer, is deliberate. Standard self-attention carries a quadratic computational cost and demands large datasets, which the ICBHI benchmark cannot provide; earlier attempts with Vision Transformers saw performance collapse to an average score of just 43.13 percent under official evaluation. The Swin variant computes attention locally within windows and exchanges information across them through its shifted-window mechanism, offering an efficient middle ground. In the fusion model, the 512-dimensional feature vectors from the CNN branch and the Swin branch are concatenated and passed through a multilayer perceptron and a SoftMax layer to produce the final classification. A pretrained variant initializes the CNN branch with EfficientNet-B0 weights and the Swin branch with Swin Transformer V2 weights, both trained on ImageNet-1k, giving the model a strong starting representation despite the modest size of the medical dataset.

What distinguishes the evaluation is its rigor. The team strictly followed the official ICBHI protocol, splitting data 60 percent for training and 40 percent for testing at the patient level, so that no respiratory cycle from a given patient could leak into the test set. They ran four tasks: four-class classification of respiratory cycles as normal, wheeze, crackle, or both; a binary anomaly-versus-normal version; three-class disease classification separating healthy, chronic, and infectious conditions; and a binary healthy-versus-unhealthy task. Beyond the standard sensitivity, specificity, average score, and harmonic score, they added AUC, macro F1, and a novel metric of their own devising, the Diagnostic Balance Index, which approaches one when sensitivity and specificity converge and falls toward zero as the model’s bias grows. Every experiment was repeated across five random seeds to confirm the results were stable rather than lucky.

The findings are striking. In the hardest four-class task, where the best prior protocol-compliant models reported sensitivities as low as 25 or 26 percent against specificities above 68 percent, the pretrained fusion model achieved the highest sensitivity in the field, outperforming existing methods by 5.8 percentage points, with true-positive detection of crackles improving by 28.55 percent and wheezes by 14.2 percent relative to the previous state of the art. In disease-level classification, the gains were even more dramatic: the pretrained fusion model reached an average score of 87.5 plus or minus 2.0 percent and a harmonic score of 87.2 plus or minus 2.1 percent on the three-class task, and 91.0 plus or minus 1.7 percent and 90.6 plus or minus 1.9 percent on the binary task, improving on prior models by up to 3.5 percentage points in harmonic score while keeping sensitivity and specificity in close alignment.

The baseline comparisons illuminate why the fusion works. The 1D-CNN, which reads the spectrogram as a temporal sequence, and the standalone Swin Transformer, which is data-hungry, both performed poorly. The 2D-CNN was competitive on average score, reaching 45.4 plus or minus 0.75 percent on the four-class task, but its sensitivity of 42.9 plus or minus 6.3 percent revealed high uncertainty and a clear bias toward specificity. Only the fusion models, pretrained and non-pretrained alike, delivered balanced detection across true positives and true negatives in every task, and the pretrained variant consistently outperformed the non-pretrained one, confirming that transfer learning from natural images meaningfully compensates for scarce medical audio data.

These numbers matter because of what they mean at the bedside. A stethoscope augmented by such a model could flag abnormal sounds without systematically missing disease in some patients or crying wolf on others, the twin failure modes that erode clinical trust. The authors are candid about the remaining hurdles: the model still confuses wheezes and crackles with one another, and the late-stage concatenation fusion limits deeper interaction between the two branches. They propose that future work explore early or intermediate fusion, gated fusion, cross-attention, and feature alignment, alongside validation on larger, more heterogeneous datasets and under real-world acoustic conditions, before deployment in clinical decision-support systems.

The broader significance extends beyond pulmonology. The study demonstrates that in medical AI, the metric you optimize quietly determines the medicine you practice, and that a model reporting a high average score can still be clinically dangerous if its sensitivity and specificity are wildly asymmetric. By introducing the Diagnostic Balance Index and showing that a carefully engineered CNN-Transformer hybrid can hold that balance steady on one of the field’s most stubborn benchmarks, the POSTECH team has set a new reference point for what robust automated auscultation should look like. As respiratory diseases continue to strain health systems worldwide, algorithms that listen to the lungs with both local precision and global context, and that err evenly rather than in one dangerous direction, may soon become the second opinion every stethoscope deserves.

Subject of Research: Fusion of CNN and swin transformer for robust lung sound classification

Article Title: Fusion of CNN and swin transformer for robust lung sound classification

Article References: Sitotaw, A. N., Hong, M., Im, S., & Kang, D. (2026). Fusion of CNN and swin transformer for robust lung sound classification. Biomedical Engineering Letters. https://doi.org/10.1007/s13534-026-00616-6

Image Credits: AI Generated

DOI: 10.1007/s13534-026-00616-6

Keywords: Fusion, swin, transformer, robust, lung, sound, classification, scientific research

The new research, conducted at Rabin Medical Center, the largest solid organ transplant center in Israel, and published in Archives of Osteoporosis, followed 39 adults who received teriparatide, a recombinant analogue of human parathyroid hormone, after kidney, lung, or liver transplantation. The investigators report that patients treated with the drug for a median of more than 22 months gained significant bone mineral density at the lumbar spine, femoral neck, and total hip, while clinicians observed acceptable metabolic safety and no episodes of graft rejection during therapy. The findings offer some of the most detailed evidence to date on anabolic osteoporosis treatment in a population that has long been underrepresented in bone research.

The scale of the skeletal problem after transplantation is difficult to overstate. In this cohort, 90 percent of patients had already suffered at least one fragility fracture before starting teriparatide, and 79 percent had multiple broken bones. The immune-suppressing glucocorticoids that nearly all transplant recipients take, at an average dose of around 7 milligrams of prednisone-equivalent per day in this study, directly impair the bone-forming cells called osteoblasts while accelerating bone resorption. Combined with pre-existing organ dysfunction, vitamin D deficiency, and the metabolic upheaval of end-stage organ disease, the result is a skeleton under relentless attack in patients who often cannot tolerate standard therapies.

Teriparatide works differently from the antiresorptive drugs, such as bisphosphonates, that are usually the first line of defense against post-transplant bone loss. Rather than slowing bone breakdown, it stimulates new bone formation by mimicking intermittent pulses of parathyroid hormone, a signal that activates osteoblasts when delivered in daily injections. This anabolic mechanism has made it a first-line option for severe glucocorticoid-induced osteoporosis in the general population. But transplant medicine has harbored a persistent worry: laboratory and clinical studies have suggested that bones in chronic kidney disease and after transplantation may become resistant to parathyroid hormone signaling, a condition linked to adynamic bone disease in which the cellular machinery for building bone falls silent. Skeptics questioned whether an anabolic drug could overcome that resistance.

The new data argue that, in many patients, it can. Among the subset of participants with paired bone density scans taken at the start and end of therapy, lumbar spine bone mineral density rose by an average of 11 percent, femoral neck density by 8 percent, and total hip density by 11 percent, gains that the authors note are comparable to responses reported in non-transplant populations and exceed what antiresorptive drugs typically achieve in transplant recipients. Statistically, the increases reached significance at all three skeletal sites. Treatment was started a median of 36 months after transplantation, once graft function and immunosuppressive regimens had often stabilized, and continued for a median of 22.3 months, a longer exposure than in most previous reports.

Fracture outcomes, though limited by small numbers, added a reassuring note. Six patients broke bones after starting teriparatide, but only two of those fractures occurred while the drug was actively being taken; the other four happened months to years after treatment had stopped. Given that nearly every participant entered the study with an already shattered fracture history, the apparent stabilization of fracture incidence during therapy is clinically meaningful, even though the retrospective design and small sample size prevented the study from formally demonstrating fracture reduction.

Safety, particularly for the transplanted organ itself, was a central question. Kidney function, measured by estimated glomerular filtration rate, declined modestly during the first six months of therapy, from an average of 73 to 64 milliliters per minute per 1.73 square meters, and then stabilized through the remainder of treatment and follow-up. Crucially, a contemporaneous matched group of transplant recipients who never took teriparatide showed a broadly similar decline over the same period, suggesting the change reflected the natural course of post-transplant kidney function rather than a drug effect. No patient progressed to end-stage kidney disease or required dialysis, and metabolic complications were rare: a single case of hypercalcemia, which resolved when the drug was stopped, and no cases of significant hypercalciuria despite most patients taking calcium and vitamin D supplements.

The study’s authors are careful to frame these results within the limits of a single-center, retrospective design. Patients were not randomly assigned; teriparatide was reserved for those with the most severe disease, often after failure of or intolerance to bisphosphonates, a selection reinforced by national reimbursement criteria in Israel. The matched control group was used only to contextualize kidney function trends, not to compare treatment effects, and causality cannot be inferred. Yet the very severity of the treated cohort strengthens the clinical signal: these were patients whose skeletons had already failed, and they still built substantial new bone. Only one randomized trial has previously tested teriparatide after transplantation, a short six-month study in kidney recipients begun during the intense immunosuppression of early recovery, which found no density gains but did prevent bone loss compared with placebo.

By including lung and liver recipients alongside kidney patients, the new study extends the evidence base to populations for which no dedicated osteoporosis treatment guidelines exist, in contrast to the organ-specific guidelines published for heart, kidney, and liver transplant care. The results align with recent observational work showing that teriparatide outperformed alendronate in renal transplant recipients and produced meaningful density gains in patients with low bone turnover, and they add histological and clinical weight to case reports of improved bone formation in heart transplant patients. Taken together, the picture emerging is that skeletal resistance to parathyroid hormone after transplantation, while a real phenomenon, is not an insurmountable barrier when therapy is timed after graft stabilization and sustained for nearly two years.

For clinicians, the message is one of selective optimism rather than blanket prescription. Teriparatide is expensive, requires daily injections, and carries a limited treatment duration, so it is unlikely to become a universal post-transplant remedy. But for the growing population of transplant survivors with severe osteoporosis, prior fractures, and few remaining options, the study suggests a viable path: an anabolic window that can rebuild what immunosuppression has torn down, administered with routine monitoring of calcium and kidney function. The authors call for prospective trials to define the optimal timing, duration, and sequencing of anabolic therapy after transplantation, questions that will only grow more urgent as transplant medicine continues to extend survival and shifts its focus toward the long-term quality of the lives it saves.

Subject of Research: Teriparatide treatment of severe osteoporosis in solid organ transplant recipients

Article Title: Teriparatide treatment of osteoporosis in solid organ transplant recipients—a single-center experience

Article References: Diker Cohen, T., Shraga-Slutzky, I., Kaminer, K., Gorshtein, A., Dotan, I., & Tsvetov, G. (2026). Teriparatide treatment of osteoporosis in solid organ transplant recipients—a single-center experience. Archives of Osteoporosis, 21(1), Article 136. https://doi.org/10.1007/s11657-026-01751-4

Image Credits: AI Generated

DOI: 10.1007/s11657-026-01751-4

Keywords: teriparatide, osteoporosis, solid organ transplantation, bone mineral density, fragility fractures, glucocorticoid-induced osteoporosis, anabolic therapy, kidney transplant, lung transplant, liver transplant, kidney function, transplant immunosuppression

The study, conducted by a team of radiologists at the University of Kansas Medical Center and published in CVIR Oncology, followed five patients with primary lung cancer who underwent pulsed electric field ablation. All procedures were performed by two experienced attending physicians, each with more than five years of experience using the Aliya System developed by Galvanize Therapeutics, and a device representative was present at every case to ensure that settings and operation matched the manufacturer’s specifications. Every ablation was carried out under computed tomography guidance, in line with the standard of care for image-guided tumor ablation. The cases represent the institution’s early clinical experience with the modality in patients selected jointly by the interventional radiology and oncology services. The results, measured primarily by change in tumor size on follow-up imaging using RECIST 1.1 criteria, where an increase of 20 percent or more in the sum of longest diameters counts as growth, were uniformly discouraging.

The first patient was an 82-year-old woman with a history of metastatic colon cancer who was found to have a primary pulmonary adenocarcinoma measuring 1.1 by 0.8 centimeters before treatment. Forty-three days after the ablation, follow-up imaging showed the tumor essentially stable at 1.2 by 0.9 centimeters. Stability might sound like partial success, but the authors note that effective ablation is expected to produce a decrease in tumor size over time, not merely a plateau. The second patient, a 72-year-old man with Erdheim-Chester interstitial lung disease and a bronchogenic carcinoma measuring 1.2 by 1.0 centimeters, initially appeared stable on a CT scan 55 days after treatment. Yet a subsequent PET/CT scan at 188 days revealed persistent metabolic activity and clear enlargement to 1.9 by 1.8 centimeters, along with a new lesion suggestive of metastasis. A later scan showed the tumor back to 1.2 by 1.1 centimeters, but the researchers classified the case as a failure, and the patient died 244 days after the procedure.

The remaining three cases were similarly bleak. Patient 3, a 47-year-old woman with metastatic non-small cell lung cancer, underwent ablation of a large tumor measuring 5.7 by 5.1 centimeters. She elected to pursue palliative care shortly afterward and died before any follow-up imaging could be obtained, leaving her outcome unmeasurable but unpromising. Patient 4, a 73-year-old woman with a history of breast cancer who developed squamous cell carcinoma with nodal metastasis, began with a tumor of 2.0 by 1.9 centimeters. Thirty-five days after ablation, the lesion had grown to 3.7 by 2.9 centimeters, and by 88 days it had reached 4.1 by 3.3 centimeters. She died 91 days after treatment from complications of cancer progression. Patient 5, a 74-year-old man with recurrent adenocarcinoma and nodal metastasis, saw his 6.3 by 3.1 centimeter tumor expand to 8.6 by 2.8 centimeters within 97 days. In total, three patients showed tumor growth, one showed stability without regression, and one died before follow-up imaging.

To understand why these findings matter, it helps to look at the physics of irreversible electroporation. The technique delivers trains of short, high-intensity electric pulses through needle electrodes placed in or around the tumor. When the induced transmembrane potential exceeds a critical threshold, the lipid bilayer of the cell membrane destabilizes and forms irreversible nanoscale defects. Cells lose their ability to maintain homeostasis and die through apoptotic pathways rather than through the coagulative necrosis produced by heat-based methods. That distinction underpins the promised advantages: extracellular matrix proteins such as collagen and elastin survive the treatment, major blood vessels and airways are relatively spared, and the reduced inflammatory cascade theoretically lowers the risk of complications such as bronchopleural fistula or damage to mediastinal structures. In organs like the liver, pancreas and kidney, where tumors often sit perilously close to major vessels, these properties have generated genuine enthusiasm and a growing clinical literature.

Lung tissue, however, presents a different environment. The air-filled parenchyma has very different electrical conductivity than solid organs, which can alter the distribution of electric fields in ways that are difficult to model and verify. Tumor size also matters: several of the patients in this series had lesions considerably larger than the small nodules typically targeted for ablation, and field homogeneity degrades rapidly with lesion diameter. The authors acknowledge that the variability in tumor sizes and histologic types across their five cases, combined with advanced disease stage and comorbidities at the time of treatment, introduces selection bias that limits how confidently the results can be generalized. Still, the direction of the findings is hard to ignore, particularly given that they echo a larger and more rigorous study.

That study, the ALICE trial, was a prospective multicenter phase II investigation conducted at two European institutions and initially designed to enroll 36 patients with lung malignancies. At an interim analysis of the first 23 enrolled patients, the results were poor enough that the trial was terminated early. Fourteen of the 23 patients went on to develop progressive disease. The Kansas case series, despite its small size, is consistent with that signal and strengthens the suspicion that irreversible electroporation, at least as currently delivered, does not achieve reliable tumor control in the lung. The authors are careful with their language, stating that the findings may indicate pulsed electric field ablation may not be effective for primary lung cancer and may not reduce tumor size, but the pattern across both studies suggests the problem is not merely a matter of operator technique.

The limitations of the new report are real and worth weighing. Five patients cannot establish efficacy or the lack of it with statistical confidence. Tumor sizes ranged from about one centimeter to more than six centimeters, and histologies included adenocarcinoma, squamous cell carcinoma and bronchogenic carcinoma, each of which may respond differently to electric field therapy. Long-term follow-up was incomplete, with one patient dying before imaging and another dying within three months of treatment. The presence of a device representative during all cases, while intended to ensure proper operation, reflects the reality of early-adopter clinical programs and does not substitute for the independent oversight of a controlled trial. What the series does provide is an honest, unvarnished look at real-world outcomes from an experienced team using a commercial system under manufacturer-approved conditions.

Where does the field go from here? The authors suggest that future research could investigate differing techniques that might increase effectiveness, potentially including refined electrode configurations, optimized pulse parameters, or combination approaches pairing electroporation with chemotherapy, immunotherapy or thermal methods. Preclinical work continues to explore how tissue conductivity, pulse frequency and electrode spacing shape the ablation zone, and the broader literature on liver, renal and pancreatic applications remains active. For now, though, the message for patients and clinicians is cautious: a technology that works beautifully on the whiteboard, sparing vessels and avoiding heat sinks, still has to prove itself against the unforgiving biology of lung cancer. On current evidence, pulsed electric field ablation of primary lung tumors has not done so, and the search for non-thermal ablation options for lung cancer continues.

One biological nuance worth emphasizing is that apoptosis, the cell death pathway triggered by irreversible electroporation, unfolds over hours to days rather than instantly. This means that immediate post-procedural imaging can underestimate the ablation zone, and conversely, the absence of a shrinking mass on early follow-up may not fully capture cellular-level damage. The Kansas team relied on tumor size as the primary metric, which is a practical but relatively blunt instrument, since volumetric assessment and metabolic imaging can sometimes reveal responses that diameter measurements miss.

The question of why electroporation behaves differently in lung is also drawing scientific attention. Aerated alveoli create a heterogeneous impedance landscape, and atelectatic or post-ablative regions conduct current differently from normal parenchyma, potentially leaving viable tumor cells in under-treated margins. Electrical field modeling software exists to plan electrode placement, but its accuracy in lung is less validated than in solid abdominal organs. Airway proximity adds further complexity, as bronchi may act as conduits that distort field distribution.

Despite the disappointing results, pulsed electric field technology continues to advance in cardiology, where similar pulse delivery is used for cardiac ablation, and lessons from that rapidly maturing field may eventually inform oncologic applications. Until properly powered prospective studies demonstrate benefit, however, thermal ablation and other established therapies remain the standard for lung tumors deemed amenable to percutaneous treatment.

Subject of Research: Pulsed Electric Field (PEF) ablation of primary lung cancer: a case series

Article Title: Pulsed Electric Field (PEF) ablation of primary lung cancer: a case series

Article References: Pulsed Electric Field (PEF) ablation of primary lung cancer: a case series. (n.d.). https://doi.org/10.1007/s44343-026-00057-z

Image Credits: AI Generated

DOI: 10.1007/s44343-026-00057-z

Keywords: Pulsed, Electric, Field, ablation, primary, lung, cancer, case, series, scientific research

The study, led by M.A.J. Morsink, S. Fleischer, T.R. Nash and colleagues, presents what it describes as a practical toolbox for modelling fibrosis outside the body. Rather than treating fibrosis as a single disease, the authors examine the biological features shared across organs while recognizing that each tissue has its own cellular composition, mechanical environment and response to injury. This distinction is critical. A model that reproduces collagen accumulation in a dish may still fail to capture the electrical properties of the heart, the air–liquid interface of the lung, the filtration architecture of the kidney or the specialized immune environment of the liver. Faithful modelling therefore requires more than adding a profibrotic chemical to cultured cells.

At the centre of fibrosis is a complex interaction between injured parenchymal cells, immune cells, vascular cells and fibroblasts. Fibroblasts are connective-tissue cells that normally help maintain the extracellular matrix and support wound healing. During chronic injury, however, they can become activated into myofibroblasts, contractile cells that produce large quantities of collagen and other matrix proteins. Signalling molecules such as transforming growth factor beta are widely associated with this transition, but fibrosis is not controlled by one pathway alone. Cytokines, growth factors, oxidative stress, altered metabolism and signals from damaged cells can all contribute. The review emphasizes that in vitro systems should reproduce this multicellular communication rather than isolate a single cell type from its biological context.

The extracellular matrix itself is not simply a passive scar. Its composition, organization and physical properties can actively direct cell behaviour. As collagen and related proteins accumulate, the tissue becomes stiffer. That mechanical change can activate mechanosensitive pathways, including signalling through integrins and transcriptional regulators such as YAP and TAZ. In turn, cells may produce even more matrix, creating a self-reinforcing cycle in which biochemical and mechanical signals amplify one another. A useful fibrosis model must therefore measure not only the quantity of deposited matrix but also its organization, crosslinking, stiffness and ability to alter cellular function. Techniques such as three-dimensional biomaterials, tunable hydrogels, traction-force measurements and advanced microscopy can help investigators observe these changes in real time.

The authors identify three-dimensional human tissue models as an important bridge between conventional cell culture and animal studies. Two-dimensional cultures are relatively inexpensive and experimentally convenient, but they flatten cells onto artificial surfaces and often expose them to conditions unlike those found in living tissue. Three-dimensional systems can allow cells to adopt more realistic shapes, form tissue-like interfaces and experience gradients of oxygen, nutrients and signalling molecules. Organoids, engineered tissue constructs and organ-on-chip platforms can further introduce fluid flow, vascular-like channels or mechanical stimulation. These features may be especially valuable for modelling organs in which physical forces are central to disease, such as the breathing lung, the beating heart and the continuously perfused kidney.

Human cells are another major advantage of these systems. Primary cells obtained from patients can retain disease-associated characteristics, while induced pluripotent stem cells can be differentiated into several human cell types and combined into more complex models. Patient-derived material may reveal why some individuals develop aggressive fibrosis after injury while others recover with limited scarring. It may also expose differences in drug response that are invisible in genetically uniform laboratory animals or immortalized cell lines. Yet the review cautions that human relevance does not automatically guarantee biological accuracy. Cells can lose their identity during expansion, stem-cell-derived populations may remain immature and donor-to-donor variability can make results difficult to compare. Rigorous characterization is therefore essential.

A major design challenge is deciding which hallmarks of fibrosis a model must reproduce. The review proposes that researchers begin with the biology of the target organ and define measurable criteria before selecting a platform. These criteria may include persistent fibroblast activation, pathological extracellular matrix deposition, tissue stiffening, inflammatory signalling, altered vascular behaviour, epithelial or endothelial dysfunction and loss of organ-specific performance. In a lung model, for example, meaningful fibrosis should involve more than collagen production; it should also reflect changes in epithelial integrity, gas-exchange-related functions and the interaction between airway or alveolar cells and immune populations. In a cardiac model, contractility and electrical activity may be as important as matrix accumulation.

The practical value of a model depends equally on how its results are measured. Traditional endpoint assays can quantify collagen or profibrotic gene expression, but they may miss changes that emerge earlier or disappear by the time a sample is collected. The authors highlight the growing importance of live imaging, single-cell and spatial molecular profiling, secreted-protein analysis, mechanical testing and functional readouts. Combining these methods can reveal which cells initiate fibrosis, how signals move through a tissue and whether a treatment reverses disease-associated function rather than merely suppressing one marker. Standardized reporting of cell sources, matrix composition, culture conditions, disease induction methods and analysis pipelines will also be necessary if results from different laboratories are to be compared.

These models could reshape the search for antifibrotic therapies, a field in which promising laboratory findings have often failed to translate into durable clinical benefits. A well-designed human tissue system could be used to test candidate drugs across several stages of disease, from early inflammatory activation to established matrix remodelling. It might also help distinguish compounds that genuinely restore tissue function from those that simply reduce a biomarker. Because fibrosis frequently involves multiple cell types and feedback loops, future platforms may be particularly useful for evaluating drug combinations and identifying patients most likely to respond. The authors ultimately present in vitro fibrosis modelling not as a replacement for clinical or animal research, but as a complementary technology capable of narrowing the gap between molecular discovery and human disease.

The review’s broader message is that the next generation of fibrosis models should be designed as integrated biological systems rather than as collections of isolated measurements. Researchers will need to combine human cells, organ-specific architecture, immune interactions, extracellular-matrix biology, mechanical forces and functional outputs in a manner that remains experimentally manageable. Advances in microfluidics, biomaterials, imaging and computational analysis could make such systems increasingly precise, while shared standards may improve reproducibility and accelerate adoption beyond specialist laboratories. By defining the hallmarks that matter most and translating them into practical design choices, the authors offer a framework for building models that are not only technically sophisticated but clinically informative. For a disease responsible for progressive organ failure across the world, that shift could turn the laboratory dish into a more powerful window on how scarring begins, persists and might finally be stopped.

Subject of Research: Human in vitro models of fibrosis across multiple organs, including the heart, lung, kidney, liver and synovium.

Article Title: A practical toolbox for modelling fibrosis in vitro

Article References: Morsink, M.A.J., Fleischer, S., Nash, T.R. et al. A practical toolbox for modelling fibrosis in vitro. Nature Biomedical Engineering (2026). https://doi.org/10.1038/s41551-026-01749-w

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41551-026-01749-w

Keywords: Fibrosis, in vitro models, tissue engineering, extracellular matrix, organoids, organ-on-chip, mechanobiology, human disease modelling, drug discovery, regenerative medicine.