The system, named COV-DSNet, was developed jointly by researchers from the Department of Critical Care Medicine at Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, and the National Key Laboratory for Novel Software Technology at Nanjing University. It belongs to a family of algorithms known as three-dimensional convolutional neural networks, which process volumetric medical images as complete spatial objects rather than as stacks of independent two-dimensional slices. This distinction matters for pneumonia imaging, because the lesions of COVID-19 — ground-glass opacities, consolidations and the characteristic peripheral distributions of inflamed lung tissue — extend across many slices, and their overall burden and evolution are inherently volumetric phenomena.

Architecturally, COV-DSNet combines two ideas that have become central to modern medical image analysis. The first is mixed convolution, in which the network applies three-dimensional convolutional kernels to some layers and two-dimensional kernels to others, balancing the need to capture spatial context across the lung volume against the need to extract fine-grained textural features within individual slices. The second is the incorporation of attention mechanisms, computational modules that allow the network to learn which regions of an image deserve greater weight when making a decision. In effect, attention steers the model toward the diseased portions of the lung and away from irrelevant structures such as the heart, mediastinum or imaging artifacts, mimicking in a limited way the focused reading strategy of an experienced radiologist.

To train and evaluate the system, the team drew on data from 578 patients with COVID-19, combining pulmonary CT scans with clinical data screening. After applying their inclusion criteria, 297 CT scans from 156 patients entered the COV-DSNet analysis. The task posed to the network was binary staging: given a scan, classify the patient as being in the progression stage or the remission stage of the disease. This framing reflects the clinical reality that the two stages call for different management priorities — progression demands anticipation of deterioration and closer monitoring, while remission supports de-escalation and planning for discharge.

The performance figures reported in the study are grounded in a standard evaluation metric known as the area under the receiver operating characteristic curve, or AUC-ROC, which summarizes a model’s ability to discriminate between the two classes across all possible decision thresholds. Across their analyses, COV-DSNet achieved a mean AUC-ROC of 0.820, with a standard deviation of 0.030 and a 95 percent confidence interval spanning 0.792 to 0.844. In a randomly assigned verification set, the system reached an AUC-ROC of 0.864, with a 95 percent confidence interval of 0.750 to 0.947, alongside a sensitivity of 0.708 and a specificity of 0.921. The authors note that this specificity — the proportion of patients correctly identified as not being in progression — was higher than that of manual reading under the pre-specified single-center scope and reader setting of the study.

That emphasis on specificity is clinically meaningful. A staging tool that frequently raises false alarms would trigger unnecessary interventions, prolonged isolation and wasted resources, eroding the trust of the clinicians who are meant to use it. A sensitivity of 0.708 means the system missed roughly three in ten progression cases in the verification set, a figure the authors present transparently, but the high specificity suggests that when COV-DSNet declares a patient to be in progression, that judgment is usually warranted. In a triage context, such a signal could help prioritize which patients receive the closest attention during a surge of admissions.

Perhaps the most consequential finding of the study concerns what the AI-derived stage adds to conventional prognostic modeling. The researchers built a baseline clinical model using two established measures: the APACHE II score, a widely used severity-of-illness metric computed from physiological measurements and age, and the PaO2/FiO2 ratio, which quantifies how efficiently oxygen passes from the lungs into the blood. On its own, this clinical model predicted severe progression within seven days with an AUC of 0.908. When the COV-DSNet-derived stage diagnosis was added, the AUC rose to 0.978 — a substantial improvement in a prediction task that directly informs decisions about intensive monitoring and early escalation of care. The combined model also showed good calibration, with a bootstrap-corrected mean absolute error of 0.047, meaning its predicted probabilities tracked closely with observed outcomes rather than being systematically overconfident or underconfident.

The study also explored a second, less obvious application: infectivity assessment. Reverse transcription-polymerase chain reaction tests, the standard for detecting SARS-CoV-2, report a cycle threshold, or Ct value, which reflects how much viral genetic material is present in a sample. Lower Ct values generally indicate higher viral loads, and Ct values have been used to guide judgments about how infectious a patient is and how long isolation should continue. The researchers found that in a subset of COVID-19 patients, the AI-derived staging helped avoid misjudgments about infectivity that would have arisen from relying on the RT-PCR Ct value alone. This suggests that the structural state of the lungs, as captured by CT, carries information about the disease process that viral load measurements do not fully encompass, and that integrating imaging-based staging could refine infection control decisions at the bedside.

The authors are careful to frame their conclusions within the limits of the data. The study was conducted at a single center, Nanjing Drum Tower Hospital, and the comparison with manual reading was made under a pre-specified reader setting, so the reported advantage over human interpretation may not generalize to every clinical environment or every level of radiological expertise. The verification set was modest in size, as reflected in the width of its confidence interval, and the work was carried out on patients from a single institution’s population. External validation on independent cohorts from other hospitals and regions remains a necessary step before any claim of broad clinical readiness, and the authors position the system as a rapid staging tool demonstrated within their single-center study rather than as a finished diagnostic product.

Even with those caveats, the work points toward a broader role for volumetric deep learning in the management of respiratory infections. The authors suggest that their approach provides new ideas and solutions for the early clinical management of other pneumonias beyond COVID-19, since the core problem — inferring disease phase from the evolving appearance of lung tissue on CT — is not unique to SARS-CoV-2. If staging algorithms of this kind can be validated across institutions, they could become a routine layer of decision support, converting the raw three-dimensional information in a chest scan into an actionable estimate of where a patient stands in the arc of their illness. For clinicians confronting the next wave of severe respiratory disease, that could mean the difference between reacting to deterioration after it happens and anticipating it days in advance.

Subject of Research: Deep learning-based staging of COVID-19 using chest CT scans

Article Title: Enhanced coronavirus disease staging with deep learning of chest computed tomography scan images

Article References: Chen, M., Cai, H., Lu, H., Ling, T., Gao, Y., You, Y., Wang, Y., Cao, K., Shi, Y., Zhang, J., & Yu, W. (2026). Enhanced coronavirus disease staging with deep learning of chest computed tomography scan images. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02825-y

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02825-y

Keywords: COVID-19, deep learning, chest CT, 3D convolutional neural network, attention mechanisms, disease staging, artificial intelligence, BMC Medical Imaging, AUC-ROC, APACHE II, RT-PCR Ct value, pneumonia

The clinical logic behind the study is straightforward but consequential. Small cell lung cancer is exquisitely radiosensitive, so outside true emergencies patients with this subtype may derive limited benefit from surgical resection of brain metastases. Large tumors from non-small cell lung cancer, by contrast, may warrant surgical removal before radiotherapy to relieve mass effect and improve local control. And a subset of non-small cell lung cancers harboring EGFR driver mutations may allow clinicians to defer upfront radiation entirely while trializing a tyrosine kinase inhibitor with intracranial activity. Three subtypes, three fundamentally different treatment pathways, and a diagnostic pipeline too slow to inform the moment when those pathways are chosen. The researchers therefore set out to build a tool that could estimate subtype using data available on day one.

The study team retrospectively identified 305 patients with metastatic lung cancer diagnosed between 2016 and 2025 at a single institution. In an unusual and methodologically deliberate design choice, the 182 patients who did not have brain metastases at diagnosis formed the training cohort, while the entire 123-patient group presenting with synchronous brain metastases was reserved as a held-out testing cohort. This partitioning preserved the clinically relevant target population for honest evaluation, though it introduced the possibility of distribution shift between the two groups, a concern the investigators addressed with dedicated sensitivity analyses. Patients were classified into three groups: small cell lung cancer, EGFR-mutated NSCLC, and EGFR-wild-type NSCLC. EGFR was singled out for independent categorization because it is among the most prevalent driver mutations and, unlike KRAS, has multiple central-nervous-system-penetrating targeted therapies already approved.

Clinical variables included sex, age, race, smoking history, pack-years, and prior lung disease such as emphysema or interstitial lung disease. The imaging features were extracted from narrative diagnostic chest CT reports using a firewalled institutional large language model, Anthropic Claude Sonnet 4.6, blinded to patient identifiers and tumor subtype. The model determined the presence or absence of fifteen predefined features, including tumor location, pleural attachment, cavitation, lymphangitic spread, fibrosis, miliary pattern, emphysema, mediastinal lymphadenopathy, vessel encasement, superior vena cava involvement, and osseous metastases. To validate this extraction, investigators manually reviewed fifty randomly selected reports as a reference standard, finding 96.8 percent overall agreement and a Cohen’s kappa of 0.918 across 750 individual feature comparisons. Notably, the authors emphasize that the modeling approach itself does not depend on the use of a large language model; the tool simply enabled consistent, high-throughput data extraction.

Two machine-learning approaches were tested: LASSO-regularized logistic regression, which performs feature selection while guarding against overfitting, and random forest, an ensemble method that aggregates many decision trees. Each was trained on clinical variables alone and on the combined clinical-plus-imaging feature set, and performance was evaluated with one-versus-rest AUC analysis in the held-out brain metastasis cohort. The combined LASSO logistic regression model achieved testing AUCs of 0.887 for EGFR-mutated disease, 0.811 for small cell lung cancer, and 0.801 for EGFR-wild-type disease. The random forest reached 0.907, 0.783, and 0.810 respectively. Adding imaging features significantly improved discrimination for EGFR-mutated cancer and small cell histology under logistic regression, and for small cell histology under random forest, while clinical variables alone were largely sufficient for the EGFR-wild-type group. Neither modeling approach demonstrated clear superiority over the other.

The three-way classification analysis offered perhaps the most clinically interpretable results. Both multiclass models showed their highest positive predictive value for EGFR-mutated cancer, exceeding 80 percent, with negative predictive values approaching 90 percent. For small cell lung cancer, the pattern inverted: positive predictive value was low, around 33 to 36 percent, but negative predictive value remained high, above 89 percent. In practical terms, the models are better at ruling out small cell histology than at confirming it, and patients predicted to have EGFR-mutated disease may represent a particularly informative group for early multidisciplinary discussion. An exploratory analysis of misclassified EGFR-wild-type patients revealed that several harbored other driver mutations, including KRAS G12A, ERBB2 mutations, and EML4-ALK fusions, suggesting the EGFR-wild-type category may partly capture a broader oncogenic-driver phenotype rather than a homogeneous entity.

Decision curve analysis added a layer of clinical realism by quantifying net benefit across a range of threshold probabilities, representing the minimum confidence at which a clinician would act on a predicted subtype. For EGFR-mutated cancer, the multiclass random forest outperformed both the flag-all and flag-none reference strategies across the entire examined threshold range from 0.05 to 0.80. At a representative threshold of 0.30, the models corresponded to roughly 12 to 14 additional correct EGFR-positive identifications per 100 patients evaluated without an increase in false positives. For small cell lung cancer, benefit was confined to low thresholds below approximately 0.30 and turned negative at higher thresholds, a limitation the authors attribute largely to the small number of small cell cases, just 19 of 123 testing patients, which constrains attainable positive predictive value and widens confidence intervals.

The feature-level analyses revealed biologically plausible signals. Miliary pulmonary metastases and pleural attachment were associated with EGFR-mutated disease, consistent with previously described imaging phenotypes of this molecular subtype, while smoking history, greater pack-years, superior vena cava involvement, and central thoracic disease distribution were associated with small cell histology. Cavitation and pleural attachment leaned toward EGFR-wild-type tumors. The investigators also confronted their design’s weaknesses directly: the single-institution cohort may overrepresent EGFR-mutated disease, imaging features were drawn from unstructured radiology reports rather than standardized interpretation, and the EGFR-wild-type group is biologically heterogeneous. Covariate reweighting and size-matched sensitivity analyses showed minimal change in model discrimination, but the authors stress that these checks cannot prove the training and testing populations were equivalent.

The bottom line is carefully bounded. The researchers explicitly state that these models should not substitute for histologic or molecular diagnosis, nor serve as a stand-alone basis for treatment decisions, and that generalizability requires validation in independent external cohorts. Yet as a proof of concept, the study demonstrates something genuinely useful: clinical and radiographic information already available at initial presentation can carry meaningful probabilistic information about lung cancer subtype during the diagnostic limbo that follows a devastating synchronous brain metastasis diagnosis. In that window, when neurosurgeons, radiation oncologists, and medical oncologists must weigh surgery, radiosurgery, and targeted therapy against one another, even an imperfect early estimate could sharpen multidisciplinary conversations and help prioritize which patients need expedited molecular confirmation. Prospective implementation studies will determine whether that promise translates into better outcomes, but the study makes a compelling case that the answers clinicians need may already be hiding in the charts and scans they see on day one.

Subject of Research: Machine-learning prediction of lung cancer subtype from clinical and chest CT features to guide early management of synchronous brain metastases

Article Title: Predictive modeling of lung cancer subtype with integration of clinical and thoracic imaging features to guide early brain metastasis management

Article References: Park, J., Buclez, P., Lubisich, J., Hanubal, K., Inda, A., Hui, C., Harris, J., & Simon, A. (2026). Predictive modeling of lung cancer subtype with integration of clinical and thoracic imaging features to guide early brain metastasis management. Journal of Neuro-Oncology, 180(1), Article 5. https://doi.org/10.1007/s11060-026-05812-z

Image Credits: AI Generated

DOI: 10.1007/s11060-026-05812-z

Keywords: lung cancer, brain metastases, EGFR mutation, small cell lung cancer, non-small cell lung cancer, machine learning, chest CT, predictive modeling, LASSO logistic regression, random forest, radiology reports, Journal of Neuro-Oncology

Pulmonary agenesis is among the rarest congenital malformations of the respiratory system. It arises when the lung bud, the embryonic outgrowth of the foregut that normally branches into the airways and lung parenchyma, fails to develop on one side. The result is a complete absence of the affected lung tissue, airway, and vascular supply, rather than an underdeveloped lung that is merely small. The condition can affect either lung, and its reported incidence varies widely across older case series, in part because many affected infants die before a diagnosis is ever made. When one lung is missing, the remaining lung must carry the entire burden of gas exchange, and the cardiovascular system must adapt to a shifted distribution of blood flow. Survival is possible, particularly with unilateral agenesis and no other major anomalies, but the condition is frequently complicated by recurrent infections and, as in this case, pulmonary vascular disease.

The clinical course of the infant described in the report illustrates why the diagnosis is so often delayed. Shortly after birth she developed severe respiratory distress, a presentation that in a neonate is far more commonly explained by pneumonia, sepsis, transient tachypnea, or aspiration than by a missing lung. At the first facility she encountered, clinicians treated her for presumed pneumonia. When she failed to improve, she was referred onward to Tikur Anbessa Specialized Hospital, the tertiary referral center affiliated with Addis Ababa University, where the team led by pediatrician Beniam Tadewos Lamboro took a closer look at what her body was actually doing.

On admission, the physical examination offered the first decisive clues. The baby showed severe intercostal and subcostal retractions, the visible sinking of the chest wall between and beneath the ribs as she struggled to move air. More tellingly, auscultation revealed completely absent air entry on the left side of her chest. A silent hemithorax in a newborn is an alarming finding that demands an immediate explanation, whether that explanation is fluid, collapse, a diaphragmatic hernia, or the absence of lung tissue itself. The examining team recognized that a simple infection could not account for everything they were hearing and seeing.

Chest radiography then sharpened the picture considerably. The X-ray showed opacities in the right middle and lower lobes with air bronchograms, findings consistent with pneumonia, confirming that the infection was real. But it also showed complete opacification of the left hemithorax, together with two features that pointed away from infection and toward a structural problem: crowding of the ribs on the left side and a shift of the mediastinum, the central compartment of the chest containing the heart and great vessels, toward the left. These are classic signs of volume loss. When lung tissue is absent or collapsed, the structures around it fill the void, and the chest wall and mediastinum migrate toward the affected side. In pneumonia, by contrast, the affected side tends to hold its volume or even expand. The combination of infection on one side and volume loss on the other raised the suspicion of a congenital lung anomaly coexisting with pneumonia.

To settle the question, the team performed contrast-enhanced computed tomography of the chest. The CT scan delivered an unambiguous answer: the left main bronchus, the left lung, and the left pulmonary artery were completely absent. There was no vestige of left-sided pulmonary architecture at all. The scan also measured the ratio of the main pulmonary artery diameter to the aortic diameter, which exceeded 1, a radiographic marker associated with elevated pulmonary vascular resistance and pulmonary hypertension. This finding mattered enormously for prognosis and management, because it signaled that the single right lung was not only doing all the oxygenation work but was also facing abnormally high pressures in its circulation.

Echocardiography completed the hemodynamic assessment and revealed the downstream consequences of the missing lung. The right side of the heart, which pumps blood into the pulmonary arteries, was enlarged, a predictable consequence of pumping against increased resistance. The baby had a patent ductus arteriosus, a persistent fetal blood vessel connecting the aorta and the pulmonary artery, measuring 2.5 millimeters with bidirectional shunting, meaning blood was flowing back and forth across it rather than in a single direction. She also had severe tricuspid regurgitation, a leak of the valve between the right atrium and right ventricle, with a peak pressure gradient of 62 millimeters of mercury, a value that indicated severe pulmonary hypertension. In essence, the vascular bed of the lone right lung could not accommodate the entire cardiac output at normal pressures, and the heart was straining against that limitation.

Management proceeded along two parallel tracks. The team supported the infant’s breathing with continuous positive airway pressure, or CPAP, which keeps the airways and air sacs propped open and reduces the work of breathing without requiring invasive ventilation. At the same time, she received medical therapy aimed at lowering pulmonary vascular resistance and supporting cardiac function. This combination reflects the modern understanding of pulmonary agenesis as a cardiopulmonary condition rather than a purely anatomical one: the missing lung is fixed, but the pressures in the pulmonary circulation and the strain on the right heart are potentially modifiable with drugs. Notably, the infant survived the neonatal period, a stage at which many cases of pulmonary agenesis historically proved fatal.

Follow-up over the first nine months of life told a nuanced story. At six months, the child remained dependent on supplemental oxygen but had achieved age-appropriate growth and developmental milestones, an outcome the authors emphasize as meaningful given the severity of her initial presentation. Repeat echocardiography, however, showed that the patent ductus arteriosus had enlarged to 6 to 7 millimeters with persistent bidirectional shunting, indicating that the pulmonary vascular disease had not resolved and may have progressed. By nine months of age she was still on home oxygen and continuing medical therapy, but she had required no further hospitalizations. The report thus documents both the resilience of a child living with a single lung and the chronic, evolving nature of the pulmonary hypertension that accompanies it.

The authors draw a clear practical conclusion from this single case. Congenital lung agenesis can coexist with ordinary respiratory infections, and those infections can mask the underlying anomaly, delaying diagnosis and appropriate counseling. They argue that clinicians in resource-limited settings, where advanced imaging is not always available at first contact, should suspect congenital lung anomalies whenever a neonate or infant presents with persistent respiratory distress accompanied by radiographic signs of volume loss, specifically mediastinal shift and rib crowding on chest radiography. In such circumstances, they recommend considering CT imaging for confirmation. The case is also a broader argument for the value of publishing detailed clinical observations from settings where rare diseases are underreported: every documented case sharpens the pattern recognition of clinicians who may otherwise encounter such a presentation only once in a career, and for the family in this report, that sharpened recognition is what transformed a treatment-resistant pneumonia into a defined, manageable diagnosis.

Subject of Research: A rare case of neonatal left lung agenesis initially misdiagnosed as pneumonia in a resource-limited setting

Article Title: Neonatal left lung agenesis presenting as pneumonia in a resource-limited setting: a case report

Article References: Lamboro, B. T., Girma, S. M., Temesgen, N. E., Mengistu, A. G., & Mamo, Z. B. (2026). Neonatal left lung agenesis presenting as pneumonia in a resource-limited setting: a case report. BMC Pediatrics. https://doi.org/10.1186/s12887-026-07723-4

Image Credits: AI Generated

DOI: 10.1186/s12887-026-07723-4

Keywords: lung agenesis, neonatal respiratory distress, congenital anomalies, pulmonary hypertension, pneumonia, patent ductus arteriosus, chest CT, echocardiography, resource-limited settings, neonatology, case report, BMC Pediatrics

Ocrelizumab is a humanized monoclonal antibody that binds to the CD20 protein found on the surface of B cells, marking them for destruction and dramatically reducing the immune attacks on the nervous system that drive multiple sclerosis. Approved after landmark trials showed clear benefits in both relapsing and primary progressive forms of the disease, it has become one of the most prescribed high-efficacy therapies in neurology. Like all immunotherapies, however, it reshapes the immune system in ways that are still being mapped. Infections, infusion reactions, and a theoretical concern about certain malignancies have long been part of its safety profile. The new study suggests that the lung, an organ densely populated by immune surveillance cells, may occasionally become collateral damage in the altered immunological landscape that follows B-cell depletion.

Organizing pneumonia is not an infection at all, despite its name. It is a form of interstitial lung disease in which the tiny air sacs of the lung and their surrounding tissue become inflamed and then filled with plugs of loose connective tissue, a process pathologists describe as granulation tissue extending into the alveolar spaces. It can arise without any identifiable cause, in which case it is called cryptogenic, or it can be secondary to drugs, infections, radiation, or autoimmune disease. Patients typically develop a persistent fever, a dry cough, and breathlessness that fail to respond to antibiotics, which is precisely the pattern that should alert clinicians to look for something other than bacterial pneumonia. On chest computed tomography, the signature findings are patchy ground-glass opacities, areas of consolidation, and sometimes the striking reversed halo sign, in which a ring of denser tissue surrounds a clearer center.

The Milanese team, led by Enrico Petza and senior author Massimo Filippi, retrospectively collected clinical, radiological, laboratory, treatment, and outcome data from multiple sclerosis patients at their center who developed organizing pneumonia while on ocrelizumab. The five identified patients, three of them women, had a median age of 54 years, with ages ranging from 27 to 64. The condition did not appear early in treatment. On average, patients had been receiving ocrelizumab for about 5.1 years before symptoms emerged, with a standard deviation of 2.9 years, and the lung disease declared itself a mean of 10.8 weeks, with a standard deviation of 10.4 weeks, after the most recent infusion. That long latency is clinically important, because it means neither patient nor physician is likely to connect a fever appearing months after a routine infusion with the drug that caused it.

The clinical picture in all five cases was strikingly uniform. Each patient presented with a persistent fever that refused to yield to antibiotics, and some also had a nonproductive cough. Chest CT scans consistently revealed patchy ground-glass opacities and consolidations, the radiological hallmarks of organizing pneumonia. Treatment followed the standard approach for the condition: all patients received prednisone at doses of 0.5 to 1 milligram per kilogram of body weight per day, continued for a median of 30 days with a range of 15 to 60 days, followed by a gradual tapering of the dose. Corticosteroids suppress the inflammatory process and usually produce a prompt clinical response, and indeed the patients initially improved. Ocrelizumab was discontinued in every case, a decision that reflects the seriousness with which the treating neurologists viewed the association.

The follow-up data, however, reveal a more troubling story than the initial recovery might suggest. Over a median follow-up of 29 months, ranging from 4 to 47 months, four of the five patients experienced respiratory relapses, and residual radiological abnormalities persisted in their lungs despite treatment. Only one patient eventually restarted a disease-modifying therapy, beginning ozanimod, a sphingosine-1-phosphate receptor modulator, two years after the onset of the lung disease. The reluctance to re-expose these patients to immunotherapy is understandable, but it leaves them vulnerable to renewed multiple sclerosis activity, creating a genuine therapeutic dilemma. The persistence of lung changes on imaging also raises questions about whether organizing pneumonia associated with B-cell depletion follows a more indolent course than the idiopathic form of the disease.

To place their five cases in context, the researchers conducted a literature review and identified 17 previously reported cases of organizing pneumonia in ocrelizumab-treated patients. Descriptive comparison showed substantial consistency between the Milanese cohort and the published cases, in terms of timing, presentation, imaging findings, and response to corticosteroids. Earlier reports included patients whose lung disease followed COVID-19 infection during anti-CD20 therapy, complicating the picture, since viral infection itself can trigger secondary organizing pneumonia. But the accumulation of cases without any infectious trigger, occurring years into treatment and weeks after infusions, strengthens the argument that ocrelizumab itself can be the inciting agent. The pattern also echoes what has been described with rituximab, an older anti-CD20 antibody used in lymphoma and rheumatoid arthritis, which has been linked to organizing pneumonia in multiple published reports.

Why would depleting B cells inflame the lung? The authors point out that the mechanism remains unclear and call for further studies to clarify the potential role of CD20-positive lymphocyte depletion in the pathogenesis of organizing pneumonia. Several hypotheses are plausible. B cells are not merely antibody producers; they regulate other immune cells, present antigens, and secrete cytokines that maintain immune homeostasis in tissues. Removing them may disturb the delicate balance of immune surveillance in the lung, allowing aberrant repair processes to take hold. Alternatively, the timing of the lung injury weeks after infusion could reflect the repopulation phase, when immature B cells return and the immune system recalibrates. It is also possible that a subclinical infection, undetected in an immunologically altered host, seeds the inflammatory cascade. Disentangling these possibilities will require immunological studies of affected lung tissue and longitudinal immune monitoring of treated patients.

For clinicians, the practical message is vigilance. A multiple sclerosis patient on ocrelizumab who develops weeks of unexplained fever with a dry cough, especially when antibiotics fail, should undergo chest imaging, and the radiologist should be told to consider organizing pneumonia among the differential diagnoses. Early recognition matters because corticosteroids are effective and because continuing the offending drug may prolong or worsen the lung injury. The authors note that the condition remained rare, affecting fewer than one in a hundred patients at their center, and nothing in the study undermines the substantial efficacy that ocrelizumab has demonstrated in randomized trials. But rare complications of very common drugs translate into real numbers of affected people, and the long latency between treatment and disease makes spontaneous reporting and pooled surveillance essential.

The study also highlights a broader theme in modern neuroimmunology: as high-efficacy immunotherapies transform the prognosis of multiple sclerosis, the frontier of care is shifting toward the fine print of long-term safety. Interstitial lung disease has now been reported as a late occurrence in ocrelizumab-treated patients in other journals as well, suggesting that the signal is no longer anecdotal. Larger registries, systematic pulmonary follow-up, and mechanistic research into how B-cell depletion reshapes lung immunity will be needed to identify who is at risk and whether monitoring or prophylactic strategies can prevent the disease. For now, the Milanese case series offers clinicians a clear template of what to watch for, and it offers patients the reassurance that when the condition is caught, effective treatment exists, even if the road to full recovery can be longer than anyone would like.

Subject of Research: Organizing pneumonia associated with ocrelizumab treatment in multiple sclerosis patients

Article Title: Ocrelizumab-related organizing pneumonia in multiple sclerosis: insights from a case series and literature review

Article References: Petza, E., Guerrieri, S., Marino, V., Gattuso, I., Rubin, M., Nozzolillo, A., Rocca, M. A., Moiola, L., & Filippi, M. (2026). Ocrelizumab-related organizing pneumonia in multiple sclerosis: insights from a case series and literature review. Journal of Neurology, 273(10), Article 600. https://doi.org/10.1007/s00415-026-14130-3

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14130-3

Keywords: multiple sclerosis, ocrelizumab, organizing pneumonia, interstitial lung disease, anti-CD20 therapy, B-cell depletion, corticosteroids, chest CT, drug safety, neuroimmunology, case series, Journal of Neurology

The study, published in the International Journal of Legal Medicine, was led by Wen-tao Xia of the Academy of Forensic Science in Shanghai, together with co-first authors Ya-ning Sun and Lei Wan and their colleagues. Rather than building a system that performs a single narrow task, the team designed a multi-task learning model based on a three-dimensional residual neural network, known as 3D-ResNet18. A single network simultaneously performs three distinct jobs: it predicts the numerical age of a fracture in days, it classifies the fracture into its healing stage, and it identifies the type of fracture present. This architecture matters because the three tasks are deeply intertwined. The radiological signs that indicate a fresh fracture are the same biological processes that define its healing stage, and forcing the network to learn all three objectives at once encourages it to extract richer, more generalizable features from the image data than a single-purpose model could.

To train and validate the model, the researchers assembled a multicenter dataset of 1,848 rib fractures identified on chest CT scans. Multicenter data is a crucial ingredient in medical artificial intelligence because images acquired on different scanners, with different protocols and from different patient populations, vary in subtle ways. Models trained on data from a single hospital frequently fail when deployed elsewhere, a phenomenon that has undermined many promising AI tools. By drawing training examples from multiple sources and then testing the model on an external dataset it had never seen, the team directly addressed the generalizability question that separates laboratory demonstrations from clinically useful systems. The external test confirmed that the model’s performance held up beyond the institutions where it was developed.

The headline numbers are striking. In the fracture age prediction task, the model achieved a mean absolute error of 7.94 days, meaning that on average its estimate of how long ago a fracture occurred was off by less than eight days. In head-to-head comparison, human expert evaluation of the same fractures produced a mean absolute error of 10.65 days. The gap of more than two and a half days may sound modest, but in forensic contexts, where the difference between a fracture sustained during a documented incident and one sustained weeks earlier can determine the outcome of a case, that improvement carries real weight. For the healing stage classification task, the model reached an accuracy of 71.18 percent, and in the fracture type classification task it achieved 94.71 percent, a level of performance that approaches the reliability of specialized detection systems.

Underneath these results lies a biological story that the algorithm has effectively learned to read. When a rib breaks, the body initiates a predictable cascade of repair. In the earliest phase, hemorrhage and inflammation dominate, and the fracture line itself may be barely visible on CT. Over the following days and weeks, soft callus forms around the fracture ends, followed by hard, mineralized callus that gradually bridges the gap. Eventually the callus remodels, and the bone slowly returns toward its original architecture. Each of these stages leaves a distinct radiological signature: changes in fracture line sharpness, callus density and volume, and eventual bony union. Radiologists have long used these signs to make rough estimates of fracture age, but the transitions are gradual and their appearance varies with patient age, bone density, and health status. A convolutional neural network processing volumetric CT data can, in principle, detect patterns in these transitions that are too subtle or too distributed for consistent human recognition.

The choice of a 3D architecture is technically significant. Rib fractures are inherently three-dimensional injuries, and the diagnostic cues they present may span multiple consecutive CT slices. A two-dimensional network analyzing each slice in isolation would discard that spatial context. By feeding volumetric patches through a 3D residual network, the model can integrate information along the length of the rib and across the fracture site, capturing the full morphology of callus formation and fracture line evolution. The residual connections within ResNet18, a now-classic design that allows gradients to flow through very deep networks during training, help the model learn fine-grained image features without succumbing to the training instabilities that plagued earlier deep architectures. The study was implemented using MONAI, an open-source framework for medical deep learning, reflecting the growing standardization of tools in this field.

The forensic implications extend well beyond rib fractures alone. The cited literature surrounding this work paints a sobering picture of why objective fracture dating is needed. Radiologists have documented imaging patterns of thoracic injuries in survivors of intimate partner violence, and separate studies have examined the radiological appearance of physical elder abuse and nonaccidental injury in the elderly. Missed rib fractures on initial trauma CT scans are a recognized problem with both clinical and forensic consequences. In cases of suspected abuse, the timing of injuries is often central to the investigation: a fracture consistent with the account given by a caregiver supports one narrative, while a fracture that predates or postdates the reported incident undermines it. Prior research has even documented self-inflicted long bone fractures in insurance fraud cases, where dating an injury could expose deception. A reproducible, quantitative tool for fracture dating gives investigators and courts evidence that does not depend on which expert happened to review the scan.

The new model also builds on a clear trajectory of prior work. Earlier studies by some of the same research community demonstrated that convolutional neural networks could detect and classify rib fractures on CT with high accuracy, and a separate multicenter study showed that a radiomics-based combined model could estimate rib fracture age from CT scans. Work from other groups has applied deep learning to estimate time-since-injury in pediatric fractures, one of the most sensitive applications in forensic radiology. The present study advances this line of research by combining age regression, healing stage classification, and fracture type identification in a single multi-task framework, and by demonstrating that the combined model outperforms manual expert evaluation on the age estimation task in external testing.

Limitations and caveats remain, as they do for any medical AI system. The raw data could not be publicly shared because of participant privacy, ethical restrictions, and legal constraints, though the authors state that de-identified data are available upon reasonable request subject to ethics approval and a data use agreement. Fracture healing is known to vary with factors such as age and bone mineral density, and the model’s performance across the full spectrum of patient populations and healing conditions will require continued scrutiny as the tool moves toward practical use. The study protocol was approved by the Ethics Committee of the Academy of Forensic Science with a waiver of informed consent, and the work was supported by China’s National Key R&D Program and several national and Shanghai-level research grants.

Still, the direction of travel is unmistakable. Artificial intelligence systems have already proven themselves at detecting rib fractures, localizing them anatomically, and flagging injuries that tired human readers miss. With this study, the same technology now demonstrates that it can read the clock embedded in healing bone more precisely than the experts who trained the field. For emergency physicians triaging trauma patients, for forensic pathologists reconstructing the sequence of events in a death investigation, and for child protection teams weighing whether an injury matches the story they have been told, a day-level estimate of fracture age delivered automatically from an existing chest CT could become a routine part of the diagnostic workflow. The rib has always kept a record of when it was broken. What has changed is that a neural network can now read that record with a precision the human eye could never consistently achieve.

Subject of Research: Deep learning estimation of rib fracture age from chest CT for forensic medicine

Article Title: A multi-task learning–based deep learning model for precise estimation of rib fracture age on chest CT

Article References: Sun, Y.-N., Wan, L., Yu, X.-Y., Wang, M.-W., Zhu, K., Li, Y.-Z., Liu, T.-A., Sheng, Y.-L., & Xia, W.-T. (2026). A multi-task learning–based deep learning model for precise estimation of rib fracture age on chest CT. International Journal of Legal Medicine. https://doi.org/10.1007/s00414-026-03962-3

Image Credits: AI Generated

DOI: 10.1007/s00414-026-03962-3

Keywords: deep learning, rib fracture, forensic medicine, chest CT, fracture age estimation, multi-task learning, 3D-ResNet18, fracture healing, artificial intelligence, legal medicine, computer vision, trauma imaging

The research team, led by Elahe Meftah and colleagues at Amir Oncology Teaching Hospital in Shiraz, combed through medical records of immunocompromised children under 18 who were screened for invasive pulmonary aspergillosis between April 2016 and December 2023. Their analysis focused on 141 children classified as having proven or probable disease according to established diagnostic criteria. The median age was just eight years, and nearly two thirds of the patients were male. Only a small fraction, 14.2 percent, had definitively proven infection confirmed by direct demonstration of the fungus, underscoring how difficult this diagnosis remains even in specialized centers. Cases labeled as possible aspergillosis were analyzed separately and descriptively, keeping the main cohort diagnostically rigorous.

One of the study’s central questions was what chest computed tomography actually reveals in children with this infection, and how those images relate to survival. Radiologists reviewed the available scans using predefined criteria, and a clear hierarchy of findings emerged. Pulmonary nodules, rounded shadows that mark where the fungus has invaded lung tissue, were by far the most common abnormality, present in 75 of the 141 patients. Consolidation, the dense opacification of lung tissue typically seen in pneumonia, came second at 44 patients. The famous halo sign, a nodule surrounded by a rim of ground-glass opacity that radiologists consider a hallmark of early invasive aspergillosis, appeared in 37 patients. Ground-glass opacities and pleural effusion, fluid around the lungs, each turned up in 16 patients.

The pattern of co-occurrence was equally informative. When children had nodules, those nodules most frequently appeared alongside the halo sign, with 33 patients showing both features simultaneously. This makes biological sense: the halo represents hemorrhage around a fungal lesion, and its frequent pairing with nodules reflects the angioinvasive nature of Aspergillus, which seeds blood vessels and starves surrounding tissue. For diagnostic purposes, the study confirms that CT remains a cornerstone of disease detection and characterization, giving clinicians a visual map of how far the infection has spread. What the scan could not do, however, was predict who would survive.

That was the study’s most striking twist. Among the 136 patients included in the survival analysis, 41 deaths occurred during follow-up. The researchers estimated survival at 96 percent at four weeks after diagnosis, but that figure fell to 66.3 percent by the end of follow-up, a sobering trajectory that illustrates the delayed toll the infection takes on fragile patients. When the team fed CT variables, including the presence of nodules, the total nodule burden, and nodule size, into multivariable Cox proportional hazards models alongside laboratory markers, none of the radiological features showed a significant association with survival. The pictures told doctors where the fungus was, but not where the patient was headed.

The blood tests told a different story. Higher serum galactomannan, a sugar molecule released by growing Aspergillus hyphae and detected through a widely used enzyme immunoassay, independently predicted mortality, with a hazard ratio of 1.06 and a 95 percent confidence interval of 1.02 to 1.10, statistically significant at p below 0.001. In practical terms, every incremental rise in galactomannan corresponded to a measurable increase in the risk of death, likely reflecting the sheer fungal load multiplying inside the lungs and spreading through the bloodstream. Because galactomannan is shed by viable fungus, serial measurements essentially track the living enemy in real time, a capability no static imaging study can match.

C-reactive protein, the classic inflammatory sentinel produced by the liver in response to systemic inflammation, proved to be an equally powerful prognostic marker. Each unit increase in CRP carried a hazard ratio of 1.02, with a confidence interval of 1.01 to 1.03 and p below 0.001. While a hazard ratio per single unit may sound modest, CRP values in sick children routinely range across hundreds of units, so the cumulative effect on risk is substantial. Elevated CRP in this setting signals an uncontrolled inflammatory storm accompanying the infection, a process known to drive tissue damage and organ failure independent of the pathogen itself. Together, the two biomarkers capture both sides of the lethal equation: how much fungus is present and how violently the body is responding.

The authors’ conclusion is carefully framed and clinically meaningful. Chest CT, they write, primarily supported disease detection and radiologic characterization, whereas serum galactomannan and CRP demonstrated greater prognostic value for mortality risk stratification within diagnosed invasive pulmonary aspergillosis. This distinction matters because diagnostic and prognostic tools serve different purposes at the bedside. A scan answers the question of whether the disease is present and what it looks like; biomarkers answer the question of how the patient is likely to fare. The study suggests that once a diagnosis is established, clinicians should look to repeated blood tests, not follow-up imaging, to gauge trajectory and intensity of therapy.

Several secondary analyses enriched the picture. The researchers examined predictors of nodules and nodule burden, probing which factors drove the most characteristic radiologic manifestation of the disease. While the source data abstracts these details, their inclusion signals the team’s effort to understand the biology behind imaging findings rather than merely cataloging them. Kaplan-Meier survival estimates anchored the mortality analysis, providing clear time-to-event curves that clinicians can interpret intuitively, and the Cox models adjusted for multiple variables simultaneously to isolate the independent contribution of each marker.

The study carries the usual caveats of retrospective, single-center research. All patients came from one oncology hospital in Shiraz, and practice patterns, host populations, and diagnostic thresholds elsewhere may differ. Not every child had a CT scan available for review, and biomarker values were drawn from routine clinical testing rather than a standardized protocol. Still, the cohort size of 141 children with proven or probable disease is substantial for a condition that remains rare and hard to diagnose, and the consistency of the statistical findings lends weight to the conclusions. For a disease in which early risk stratification can mean the difference between aggressive antifungal therapy and a fatal outcome, the message is potent: in children with invasive pulmonary aspergillosis, what is in the blood may matter more than what is on the scan.

Subject of Research: Prognostic value of chest CT findings and serum biomarkers for mortality in pediatric invasive pulmonary aspergillosis

Article Title: Chest CT findings and biomarker predictors of mortality in pediatric invasive pulmonary aspergillosis: a retrospective cohort study

Article References: Meftah, E., Mirbagheri, S. S., Reihani, Y., Jafarian, H., Hamzavi, S. S., Abbasi, A., Khajeh, S., Naseri, A., Nazemosadat, S. M. A., Habibzadeh, A., Haghshenas, A., Abdipour Mehrian, S. R., & Amanati, A. (2026). Chest CT findings and biomarker predictors of mortality in pediatric invasive pulmonary aspergillosis: a retrospective cohort study. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14512-y

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14512-y

Keywords: invasive pulmonary aspergillosis, pediatric oncology, chest CT, galactomannan, C-reactive protein, mortality, fungal infection, immunocompromised children, halo sign, prognostic biomarkers, retrospective cohort, Aspergillus

The study, led by clinicians within a lung cancer screening program directed by Julie Barta, MD, an associate professor at Sidney Kimmel Medical College, focused on women between the ages of 50 and 75 who were current smokers or had a significant smoking history. Current guidelines call for people in this group to undergo an annual chest computed tomography scan to look for early signs of lung disease, including lung cancer, which remains the leading cause of cancer death in the United States. But the same population carries elevated risk for other malignancies as well. Smoking is a well-established risk factor across many cancer types, and as Meghan Maceyko, MD, a fourth-year general surgery resident and one of the researchers, puts it plainly: smoking is a risk factor in all cancers. For older women with substantial smoking histories, that includes breast cancer, one of the most common diagnoses in this age group and one for which mammography screening has repeatedly been shown to save lives when tumors are caught early.

To understand how well these women were keeping up with breast cancer screening, the team took a retrospective look at their electronic health records. The question was simple: had each woman received a mammogram in the two years before her lung cancer screening scan, a window consistent with standard recommendations for breast cancer screening in this age range? The records allowed the researchers to go beyond a single adherence number and examine patient characteristics associated with being overdue. What emerged was a pattern that clinicians who work in preventive care will recognize with frustration. Women with a greater smoking history were more likely to be overdue for mammography, and so were women who had previously been diagnosed with cancer.

That second finding is the one that gives the study its urgency. Patients with a prior cancer diagnosis are, by definition, among the highest-risk people in any screening population. They have already faced one malignancy, they are monitored closely for recurrence, and many are at increased risk for secondary cancers, whether because of shared risk factors such as smoking, the late effects of treatment, or genetic susceptibility. Yet the Jefferson analysis suggests that these very patients were among the least likely to be current on breast cancer screening. In other words, the women who stood to benefit most from a routine mammogram were the ones least likely to be getting one. The researchers describe this as a critical gap in preventive care, and the phrase is not an exaggeration. Screening programs are designed to concentrate resources where risk is highest, so an inverse relationship between risk and adherence signals that something in the delivery system is failing the patients who need it most.

Why might the highest-risk patients be the least well screened? The study did not set out to answer that question directly, but several plausible mechanisms fit the data. Patients with extensive smoking histories often carry a heavier burden of chronic disease overall, from cardiovascular disease to chronic obstructive pulmonary disease, and their medical visits may be dominated by the management of immediate symptoms rather than preventive planning. A prior cancer diagnosis can complicate the picture further. Survivorship care is frequently fragmented between oncologists and primary care providers, and responsibility for routine screening can fall into the space between them. Some survivors may also experience avoidance or anxiety around testing that recalls their original diagnosis. Whatever the individual explanations, the aggregate result is the same: the patients with the most to gain from screening are the ones most likely to slip past it unnoticed.

The proposed remedy is deliberately modest in its technical demands. Rather than building new programs or requiring additional visits, the researchers suggest embedding a breast cancer screening prompt into the intake process that lung screening patients already complete. For medical centers that operate lung cancer screening programs, such as Jefferson Health, the infrastructure is already in place. Patients are scheduled, registered, asked structured questions about their health, and seen by nursing staff before their scan. The next step, Maceyko notes, is simply to choose an easy and effective prompt. That could take the form of a pop-up reminder embedded in the screening questionnaire, or a single added question posed by the intake nurse. Either way, the intervention targets the moment of maximum engagement, when a high-risk patient is physically present and the health system has her attention.

The elegance of this approach lies in what it does not require. It does not require new imaging equipment, additional staffing, or a separate outreach campaign with mailings and phone calls. It does not require patients to navigate a new appointment or a new department. It piggybacks on a visit that already exists for a population that is already identified as high-risk by virtue of enrollment in lung screening. In implementation science, this kind of design is often described as leveraging existing touchpoints, and it is one of the most reliable ways to improve adherence at scale. A prompt that takes seconds to deliver can convert an otherwise missed screening into a completed one, particularly when it is paired with a clear pathway for scheduling the mammogram before the patient leaves.

The implications extend well beyond the pairing of chest CT and mammography. The same logic applies to any setting in which patients at elevated risk for one condition are seen for the surveillance of another. A patient arriving for colonoscopy could be asked about cervical screening; a patient in a dermatology surveillance program could be prompted about breast or colorectal testing. The researchers point out that further work in this area may open additional avenues for improving adherence, including prompts about other cancer screenings and the creation of pathways for electronic health records to automatically flag overdue tests. Modern EHR systems already track due dates for preventive services, but turning that data into action at the point of care remains inconsistent. A flag that surfaces during check-in, or a question scripted into the intake workflow, closes the distance between knowing that a test is overdue and actually getting it done.

There is also a broader lesson in the study’s central paradox. Health systems tend to assume that patients who are engaged enough to participate in one specialized screening program are likely to be current on their other preventive care, but the Jefferson data show that this assumption cannot be relied upon. Enrollment in lung screening identifies a high-risk population, yet within that population, adherence to mammography varies in ways that track with smoking history and cancer history rather than with engagement. Screening status for one test is simply not a reliable proxy for screening status across the board. That is precisely why an explicit prompt matters: it replaces an assumption with a question, and a question asked at the right moment can be answered, acted upon, and documented.

For the women at the center of the study, the stakes are concrete. Lung cancer screening exists because early detection changes outcomes in a disease that is usually found too late; mammography exists for the same reason in breast cancer. A patient who receives both screenings on schedule is protected twice, at a combined cost of two imaging appointments a year. If a pop-up reminder or an extra intake question can move even a fraction of overdue patients into compliance, the intervention will have earned its place in the workflow. The Jefferson team’s contribution is to show where the gap lies, who falls into it, and how small the fix could be. The next phase of research will determine which prompts work best and how far the principle can be extended, but the direction is already clear: the most efficient place to catch a missed screening is at the appointment the patient was already going to attend.

Subject of Research: Breast cancer screening adherence among women enrolled in a lung cancer screening program

Article Title: Boosting cancer screenings in high-risk groups

Article References: Boosting cancer screenings in high-risk groups. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: lung cancer screening, mammography, breast cancer, smoking, preventive care, screening adherence, electronic health records, Thomas Jefferson University, high-risk patients, cancer survivors, chest CT, health care intervention