The current practice of pre-operative imaging in breast cancer is highly varied throughout Europe. Therefore, the European Society of Breast Imaging (EUSOBI) launched a call to all other European scientific societies involved in breast care to provide expert advice for pre-operative imaging in breast cancer and come to a common understanding of the types of evidence required for clinical practice guidelines for diagnostic tests. A panel comprising 13 experts (invited based on their level of expertise and representation in medical societies) voted on statements and questions encompassing all aspects of pre-operative staging. Consensus was reached in 67.4% of statements and questions, a majority in 28.3%, and no decision in 4.3%. Based on these findings, the panel developed a practical toolbox, based on currently available evidence and panel expertise, for the optimal breast cancer staging pathway considering all current imaging methods. KEY POINTS: Question What is the role of breast MRI and other imaging methods in pre-operative staging in the light of current evidence? Findings Consensus was reached among panelists in 67.4% of statements and questions, a majority in 28.3%, and no decision in 4.3%. Clinical relevance A practical working toolbox for optimal breast cancer staging is provided based on currently available evidence and panel expertise.
Though subgroup performance reporting helps ensure the safety of artificial intelligence (AI) products, the extent of this reporting remains unclear. This scoping review identifies studies validating commercially available AI-based products and reports the trends in performance reporting across sex, age, and race/ethnicity demographic subgroups. Peer-reviewed validation studies of commercially available products published after 2010 were collected from the Health AI Register and PubMed on 29 November 2024. Study trends in the reporting of sex, age, and race/ethnicity were mapped with regression analysis. We apply the Wilson confidence interval equation to estimate which tuberculosis detection studies are underpowered for subgroup meta-analysis. Three hundred ninety-two of 545 studies validating 252 products reported subgroup demographic data for any of the three groups. Only 77 of these presented subgroup performance results. Skeletal (20/88) and lung (30/139) studies, and those utilizing chest (24/79) or bone (19/63) radiographs, most often presented subgroup performance data. We found no evidence that more recent studies (OR: 1.039 [95% CI: 0.959-1.127]) or company sponsorship (OR: 1.010 [95% CI: 0.492-1.920]) led to increased subgroup reporting. We show that 14/21 tuberculosis datasets may be underpowered for post-hoc subgroup meta-analysis. This scoping review quantifies how fragmented the commercial validation landscape is, showing that reporting for both the demographics and per-subgroup performance is inadequate for estimating subgroup bias. This systemic problem requires effort from all stakeholders, from researchers to regulatory agencies, encouraging thorough reporting and commercial product validation to support physician and patient trust in medical AI products. Question The number of studies validating the performance of each commercially available radiology AI product for minority subgroup bias is unclear. Findings The currently available commercial AI validation studies often neglect to describe demographic subgroup data, and fewer provide performance results per subgroup, prohibiting algorithmic bias meta-analysis. Clinical relevance Physician and patient trust in the medical AI already used clinically must be built on peer-reviewed literature and meta-analysis. The current literature is insufficient for determining the safety and performance of these products for demographic minorities.
Pediatric neurointerventional procedures are increasingly performed, yet antiplatelet management remains largely empirical or extrapolated from adult practice because of limited pediatric-specific evidence. This guideline provides scenario-based recommendations for antiplatelet therapy in children undergoing neurointerventional procedures; anticoagulant therapy will be addressed separately. Institutional protocols from experienced pediatric neurointerventional specialists were integrated with a comprehensive scoping literature review through an iterative consensus process involving the European Society of Minimally Invasive Neurological Therapy (ESMINT), European Society of Neuroradiology (ESNR), International Pediatric Stroke Organization (IPSO), Society of NeuroInterventional Surgery (SNIS), and World Federation of Interventional and Therapeutic Neuroradiology (WFITN). Recommendations are organized around common clinical scenarios, including elective and emergency aneurysm treatment and the management of thromboembolic complications. Dosing, administration, and monitoring considerations are provided for aspirin, clopidogrel, prasugrel, ticagrelor, cangrelor, and glycoprotein IIb/IIIa inhibitors, with additional attention to age-dependent pharmacokinetics, renal function, available formulations, and platelet-function testing. These expert-consensus recommendations are intended to support safer and more consistent clinical practice while recognizing the need for individualized clinical judgment and prospective pediatric studies.
Endometrial cancer is a growing global health concern, and magnetic resonance imaging (MRI) is central to its diagnosis, staging, and management. This study used a descriptive bibliometric approach to map the research landscape, identify emerging themes, and characterize global collaboration in MRI‑related endometrial cancer research over the past four decades. This study searched Web of Science Core Collection for English-language original articles on MRI in endometrial cancer published from 1984 to 2024. Using Microsoft Excel, VOSviewer, CiteSpace, and Bibliometric.com, we described publication trends, main contributors, collaboration networks, and research hotspots. Only network- and frequency-based analyses were performed; no inferential statistics or causal analyses were done. A total of 1,165 publications by 6,067 authors from 3,676 institutions in 256 countries/regions were identified, showing sustained growth in research activity. China, Japan, and the USA were the most productive countries. Key journals included Radiology, European Radiology, and Gynecologic Oncology. Co-authorship and institutional analyses indicated extensive international cooperation, particularly among centers such as the University of Bergen and Kyoto University. Keyword co-occurrence and burst analyses identified four main thematic clusters: (1) diagnostic imaging and pathology, (2) multi-modality imaging and disease management, (3) clinical outcomes and therapeutic approaches, and (4) prognostic assessment and staging. Recent studies have increasingly focused on advanced imaging analytics, with emerging hotspots in "radiomics," "nomogram," and "risk assessment". MRI research in endometrial cancer has grown and become more diverse, with increasing international collaboration. New work on radiomics, artificial intelligence (AI), and prognostic models points toward more personalized imaging but is still mainly experimental. As a descriptive bibliometric study, this work does not judge the quality or clinical readiness of specific methods. Standardized MRI protocols, better access in low-resource settings, and prospective validation of radiomics and AI tools are needed to turn these trends into real benefits for patients.
Medical artificial intelligence (AI) tools are increasingly used to manage the growing imaging workload in radiology. Although highly relevant to daily clinical practice, AI-driven workflow tools remain underrepresented in research. Inconsistencies in DICOM metadata are a major obstacle to workflow optimization, AI integration, and inter-institutional data sharing. Automatic DICOM metadata standardization is an important step towards addressing these challenges. This study evaluates the labeling accuracy of a commercially available AI-aided hybrid software tool and its impact on radiologists' reading times. A retrospective cohort study assessed the labeling accuracy of the DICOM standardization tool. A retrospective before-and-after design evaluated its impact on reading times. The tool was applied by a radiology provider between 2022 and 2024. Standardized DICOM labels (modality, body part, laterality, plane, contrast protocol) across 422 CR images and 1503 CT series were manually reviewed (gold standard). In a separate analysis, reading times before (10,966 cases) and after (10,342 cases) DICOM standardization were compared. Labeling accuracy ranged from 83 to 100% for body part, 91 to 100% for plane, and 88-100% for protocol classification. Following implementation, average reading times significantly decreased for CT Abdomen (-2.9 min), Total Body (-2.2 min), Head (-0.73 min), and Temporal Bone (-2.5 min) (p ≤ 0.02), with relative efficiency gains of 8-22%. Extrapolated annually, this equals 270 h saved. No significant changes were observed for CT Chest and Sinus/Orbits. This study suggests that an AI-aided tool can adequately standardize DICOM labels and may be associated with statistically significant reductions in radiologists' reading times following implementation. Question Inconsistent DICOM metadata pose long-standing challenges in imaging data management, hindering processes such as workflow efficiency, AI integration, and inter-institutional image sharing. What is the role of AI solutions to address DICOM inconsistencies? Findings An AI-aided hybrid software tool adequately standardized DICOM labels and may be associated with statistically significant reductions in radiologists' reading times following implementation. Clinical relevance Standardizing DICOM metadata may help streamline radiology workflows and image accessibility, benefiting both radiologists and other healthcare professionals.
To highlight the potential of large language models (LLMs) in radiology and to stimulate discussion on their integration into clinical practice, including associated benefits and challenges. This editorial was authored with the assistance of an advanced artificial intelligence (AI) language model (ChatGPT-4, OpenAI), under direct human oversight. The content was critically reviewed, edited, and refined by the author, and all references were verified by the editors. LLMs demonstrate substantial potential to enhance radiological practice, including improvements in workflow efficiency, reporting, education, and clinical decision support. However, their implementation raises important considerations related to accuracy, reliability, ethical use, and appropriate clinical oversight. The integration of LLMs into radiology requires a balanced approach that embraces their innovative capabilities while ensuring ethical, safe, and practical implementation. With appropriate safeguards and continued evaluation, LLMs may contribute to improved patient care and outcomes.
Background and Objectives: Sarcomas are rare malignancies of mesenchymal origin comprising less than 1% of all adult solid tumors, exhibiting marked histological heterogeneity and variable clinical behavior. Data from Eastern European tertiary oncology centers remain scarce. This study characterized the clinicopathological features, treatment modalities, and survival outcomes of patients with sarcomas of the extremities and trunk treated at the "Prof. Dr. Alexandru Trestioreanu" Institute of Oncology from Bucharest over a ten-year period. Materials and Methods: We conducted a retrospective analysis of 164 patients diagnosed with sarcomas of the extremities and trunk between 2010 and 2020 at "Prof. Dr. Alexandru Trestioreanu" Institute of Oncology. Variables included age, sex, tumor localization, histological subtype, immunohistochemical profile, treatment modalities, recurrence, metastatic spread, and overall survival (OS). Kaplan-Meier curves estimated survival; log-rank tests were applied for subgroup comparisons. Results: The cohort comprised 82 males and 82 females (50.0% each), with a mean age of 48.8 ± 18.3 years. The lower limb was the most frequent site (n = 96, 58.5%), particularly the thigh/femur (34.1%). The most common subtypes were undifferentiated pleomorphic sarcoma (14.6%), osteosarcoma (12.2%), and fibrosarcoma (11.0%). Surgery was performed in 75.6%, chemotherapy in 80.5%, and radiotherapy in 59.8%. Local recurrence occurred in 35.4% and distant metastases in 41.5%. The median OS was 96.0 months (vital status known for 160/164 patients; 90 deceased, 70 alive; OS duration available in 126 patients). Metastatic disease was associated with shorter observed survival in descriptive Kaplan-Meier analysis (log-rank p < 0.001); this comparison is exploratory given the time-dependent nature of the variable. Survival ranged from 11.5 months (leiomyosarcoma) to 162.5 months (dermatofibrosarcoma protuberans) by histotype. Conclusions: This study provides clinically relevant epidemiological and survival data from Romania. The findings illustrate real-world heterogeneity of sarcoma presentations and outcomes at an Eastern European tertiary center and highlight the need for improved diagnostic standardization, prospective data collection, and integration within specialized sarcoma networks.
To identify subgroups of patients with prostate cancer (PCa) after radical prostatectomy (RP) based on clinical and magnetic resonance imaging (MRI) radiomics features and evaluate the prognostic value in predicting 5-year progression-free survival (PFS). Preoperative MRI and clinical data from 400 patients (185 with recurrence) were collected from three centers (one training and two external validation groups). Radiomics features were extracted from index lesions. PFS-associated clinical and radiomics features were selected by least absolute shrinkage and selection operator (LASSO)-Cox analysis. The K-means clustering method was used to identify subgroups and construct a Radiomic-Clinical model. PFS differences across subgroups were assessed using Kaplan-Meier survival analyses. The predictive performance of the Radiomic-Clinical model was compared with the European Association of Urology (EAU), University of California, San Francisco (UCSF) Cancer of the Prostate Risk Assessment (CAPRA), and PIPEN models using the concordance index (C-index). A total of 5 clinical and 13 radiomics features were selected, and three distinct prognostic subgroups were identified within the Radiomic-Clinical model. The Radiomic-Clinical model demonstrated superior predictive accuracy with C-indices of 0.82 (training group), 0.78 (validation group 1), and 0.79 (validation group 2), outperforming the EAU (0.68, 0.70, and 0.65), CAPRA (0.71, 0.67, and 0.70), and PIPEN models (0.71, 0.70, and 0.68) (p < 0.05). Unsupervised learning using radiomics and clinical data effectively identifies distinct prognostic subgroups in PCa patients after RP, offering superior predictive performance over existing models for 5-year PFS.
Left atrial volume index (LAVI) has been recognized as a significant indicator of left heart remodeling and diastolic dysfunction. This study aimed to investigate the association between LAVI assessed by coronary computed tomography angiography (CCTA) and major adverse cardiovascular events (MACEs) in patients with severe aortic stenosis (AS) after transcatheter aortic valve replacement (TAVR). This retrospective single-center study enrolled patients with severe AS undergoing TAVR between February 2020 and June 2024. All patients underwent CCTA examination prior to TAVR. Left atrial (LA) volume was automatically quantified from CCTA images, and the LAVI was computed by indexing to body surface area (BSA). Univariate and Firth-penalized Cox proportional hazards regression analyses were used to determine the predictors of MACE. Additionally, restricted cubic spline analysis was performed to explore the relationship between LAVI and MACE. A total of 206 patients (113 males, 93 females; mean age 68.11±8.01 years) were included in the final analysis. The incidence of MACE was 13.1% over a median follow-up time of 581 [interquartile range (IQR), 378-990] days. The LAVI was significantly higher in patients with MACE than in those without MACE [70.14 (54.81-84.12) vs. 51.19 (39.67-64.45) mL/m2, P<0.001]. After adjustment for clinical confounders and European System for Cardiac Operative Risk Evaluation (EuroSCORE II), LAVI ≥53.32 mL/m2 [hazard ratio (HR) =4.249, 95% confidence interval (CI): 1.585-11.394, P=0.001] and male sex (HR =2.864, 95% CI: 1.164-7.051, P=0.012) still independently predicted MACE. Restricted cubic spline results showed a nonlinear relationship between LAVI and MACE (P for nonlinearity =0.032). Preprocedural CCTA-derived LAVI may aid risk stratification for MACE after TAVR in severe AS. Male patients warrant closer postprocedural attention.
Improving the quality and equity of oncology care is a strategic priority in Europe. The Organisation of European Cancer Institutes (OECI) Accreditation and Designation Programme provides a unified framework integrating care, research, education, and governance. Its standards emphasise multidisciplinary coordination, digital infrastructures, and patient-centred outcomes. Head and neck cancer remains a major global challenge, with rising incidence and heterogeneous results, highlighting the need for standardised pathways and interoperable data systems. This study describes how a Head and Neck Cancer Unit was redesigned and digitalised to comply with OECI requirements, focusing on elements of the care domain. A structured organisational redesign was undertaken using workflow mapping, a RACI matrix, and Plan-Do-Act-Check cycles. Governance involved clinical leaders, specialised nurses, quality officers, and IT and engineering staff. Guidelines were updated, fast‑track pathways refined, and synoptic forms integrated into the electronic health record using terminologies compatible with OMOP. PROMs and PREMs were deployed through the institutional portal. A modified Delphi survey assessed team consensus. Between January and October 2025, 353 patients were evaluated, 198 through fast‑track referral. Diagnostic and treatment intervals met targets. PROM completion reached 61%. PREMs highlighted strengths in professionalism and multidisciplinary coordination and identified communication and administrative burden as areas for improvement. The Delphi survey showed strong agreement (α = 0.71), particularly for Multidisciplinary participation and nursing collaboration. The redesign, aligned with OECI standards, strengthened governance, digitalisation, and patient‑centred care. The model demonstrates how accreditation can drive sustainable, data‑driven improvement and support integrated cancer networks.
Incident Learning Systems (ILSs) are central to patient safety in radiotherapy, enabling learning from adverse events and near misses. Despite EU regulatory requirements, substantial variability persists across Europe in the implementation and effectiveness of ILSs in radiotherapy. This paper presents radiotherapy-specific recommendations derived from the MARLIN study to support harmonised, risk-informed implementation of ILSs. The 24-month MARLIN study, conducted under the SAMIRA Action Plan, employed a structured literature review, an online European survey of clinical facilities, competent authorities and professional societies, expert interviews, and a multi-stakeholder consensus workshop. Survey data from 172 respondents in 28 countries were analysed to identify current practices, barriers and enabling factors for ILS implementation in radiotherapy. Although all responding countries reported transposition of the Directive, substantial variability was observed in criteria for reporting significant radiotherapy events, feedback mechanisms and use of international databases. Fear of punitive actions, limited resources, lack of training in incident analysis, and insufficient dissemination of lessons learned were identified as key barriers. External-beam radiotherapy showed more mature ILS implementation than brachytherapy. Findings from the MARLIN study informed recommendations on category-based event classification, radiotherapy-specific taxonomies, multidisciplinary incident-learning committees, and collaboration between clinical facilities, competent authorities and professional societies. The MARLIN recommendations provide a practical framework to strengthen ILS implementation in radiotherapy, promote a just culture, enhance learning from incidents and support regulatory compliance, ultimately improving patient safety and quality of care across Europe, while the broader RP-208 report extends these principles to all medical fields using ionising radiation, supporting cross-disciplinary harmonisation.
Germline risk variants for oral squamous cell carcinoma (OSCC) in never-smoking non-drinking (NSND) young adults (YA) are poorly characterized and genome-wide association studies focusing on NSND YA OSCC are lacking. This study aims to evaluate if rare germline variants are associated with NSND YA OSCC. We conducted a retrospective, single-center Canadian cohort study of patients who underwent primary surgery for OSCC between 2010 and 2024. Genome-wide association analyses were performed for three comparisons: 1- NSND YA (n = 10) vs. other OSCC (n = 150), 2- all-ages NSND (n = 29) vs. ever-smoker and/or drinker (n = 131), and 3- any risk factor YA OSCC (n = 22) vs. age ≥ 45 (n = 138). A polygenic risk score for age at OSCC diagnosis was derived from clumping + thresholding. We included 160 patients of European ancestry with a diagnosis of OSCC: 22 YA and 29 NSND patients; 10 patients were both YA and NSND. YA cases predominantly involved the oral tongue. One rare variant, rs191595756, showed a suggestive association with NSND YA OSCC (p = 3.3e-06), but did not reach genome-wide significance. This variant is located in an intron of LINC00944 and in an exon of LINC02824. No rare variant signals were supported in the all-ages NSND and any risk factor YA OSCC. None of the identified variants is reported in ClinVar. An exploratory polygenic risk score composed of 13 index variants was constructed. It is independent of NSND or ever-smoker and/or drinker status. It is also independent of oral tongue or other oral cavity primary cancer site. This is the first study investigating genome-wide associations in NSND YA OSCC. No single germline variant of genome-wide significance associated with NSND YA OSCC occurrence was found, suggesting a polygenic architecture. Larger, multicentric studies are needed to validate a polygenic risk score specifically predictive of NSND YA OSCC risk.
Sarcopenia may adversely affect postoperative outcomes, but most studies in colorectal cancer (CRC) have relied solely on computed tomography (CT)-derived muscle mass. We evaluated the association between a comprehensive assessment based on the European Working Group on Sarcopenia in Older People 2 (EWGSOP2) criteria and short-term postoperative outcomes after curative CRC surgery. In this prospective observational study, 70 consecutive patients who underwent curative CRC resection between July 2024 and July 2025 were assessed for muscle strength (handgrip strength), physical performance (4-m gait speed), and muscle mass (L3 CT skeletal muscle index). According to EWGSOP2 criteria, patients were classified as non-sarcopenic or into the hierarchical cumulative stages of probable, confirmed, and severe sarcopenia. Postoperative outcomes included major complications and length of hospital stay. Of the 70 patients, 41 (58.6%) were non-sarcopenic, whereas 29 (41.4%) had at least probable sarcopenia; within this hierarchical group, 27 (38.6%) had confirmed sarcopenia and 18 (25.7%) had severe sarcopenia. Patients with at least probable sarcopenia were older and had lower body mass index and serum albumin levels (all p < 0.01). Across the hierarchical EWGSOP2 stages, the frequency of major postoperative complications increased compared with non-sarcopenic patients (probable 34.4%, confirmed 37.0%, severe 55.5% vs. 9.7%, respectively; p < 0.01), and hospital stay was significantly longer (9.3-10.6 vs. 6.1 days; p < 0.001). Tumor characteristics did not differ significantly between groups. In exploratory multivariable analysis, at least probable sarcopenia was associated with major postoperative complications (adjusted odds ratio [aOR] 11.75, 95% confidence interval [CI] 1.27-108.82; p = 0.030) and prolonged hospital stay (aOR 8.33, 95% CI 1.01-68.38; p = 0.049), although these estimates should be interpreted cautiously because of the small sample size and wide confidence intervals. EWGSOP2-based sarcopenia assessment was associated with worse short-term postoperative outcomes after curative CRC resection. Given the limited sample size and wide confidence intervals, these findings should be interpreted as exploratory and require external validation in larger prospective cohorts before clinical implementation. ClinicalTrials.gov Identifier: NCT06698289.
To compare mechanical axis image quality between the EOS imaging system and digital radiograph (DR). This retrospective study included 100 patients who underwent both EOS and DR mechanical axis imaging. It was conducted according to the Quality and Standards Department at the University of Sheffield. Three observers independently scored the DR and EOS images according to the modified European guidelines on quality criteria for diagnostic radiographic images in paediatrics: resolution, coverage of the area of interest, positioning, contrast, and motion. The range of scores was (Yes) for the fulfilled criterion and (No) for the not fulfilled criterion. For the motion criterion, the scores were (No) if the image quality criterion was not fulfilled (motion) and (Yes) if the image quality criterion was fulfilled (no motion). Inter-rater agreement was measured using the Gwet's agreement coefficient (AC1) with a 95% confidence interval. The accuracy of the EOS was assessed using sensitivity, specificity, and the area under curve (AUC). Variables between the EOS and DR were compared using McNemar's test. Both Gwet's AC1 and corresponding 95%CI were computed using irrCAC library in R4.41 software. The mean (range) age of patients when they underwent DR and EOS imaging was 11.2 (4-18) and 11.8 (5-17) years, respectively. EOS demonstrated almost perfect inter-rater reliability in the resolution criterion (AC1 = 0.99, 95% CI = 0.98,1.00, p < 0.001), while DR showed similar reliability in the motion artefacts (AC1 = 0.96, 95% CI = 0.92.00, p < 0.001). Intra-rater reliability was high for most criteria. EOS images demonstrated high sensitivity for all criteria and variable specificity, which equalled or exceeded that of DR images. Significant differences were found in some criteria, such as the coverage of the area of interest (chi-squared = 37.39, p < 0.001) and hip level (chi-squared statistic = 31.44, p < 0.001). The high AC1, sensitivity, and AUCs highlight the diagnostic image quality of the EOS imaging system for mechanical axis images, making it a valuable modality in clinical settings. Therefore, the EOS imaging system can be used instead of DR in this context, offering advantages such as lower irradiation, shorter scanning time, and the capability for 3D reconstruction.
Chronic thromboembolic pulmonary hypertension (CTEPH) is a treatable cause of pulmonary hypertension but remains under-recognized and is often diagnosed with delay. Limited access to lung ventilation-perfusion (V/Q) scintigraphy, especially outside tertiary centers, is one contributor. Dynamic chest radiography (DCR), with a pulmonary circulation analysis program, can provide a rapid, non-invasive, and widely deployable assessment of pulmonary perfusion. We describe the protocol of a multicenter reader study testing whether adding DCR-based analysis to standard initial work-up improves diagnostic accuracy for CTEPH among patients with echocardiographically suspected pulmonary hypertension. This investigator-initiated, multicenter, assessor-blinded, case-wise randomized superiority reader study compares standard initial work-up (blood tests, chest X-ray, ECG, pulmonary function tests, and transthoracic echocardiography per guidelines) with standard work-up plus DCR-based pulmonary circulation analysis. The primary endpoint is diagnostic accuracy for discrimination between CTEPH and non-CTEPH in the intention-to-treat set. The final diagnosis of CTEPH versus non-CTEPH will be defined as the reference standard according to the Japanese and European guidelines for pulmonary hypertension. Secondary endpoints include sensitivity, specificity, positive and negative predictive values; agreement with V/Q scintigraphy regarding regional perfusion defects using κ statistics; and STARD-conformant academic performance evaluation of DCR-based pulmonary circulation analysis and lung perfusion scintigraphy in relation to the site-level final diagnosis of CTEPH versus non-CTEPH in a full analysis set. Safety endpoints include adverse events during DCR acquisition and device malfunctions. The target sample size is 108 cases with 1:1 allocation. Recruitment started on 30/07/2025 and is expected to continue until 28/02/2027, with overall study completion planned for 31/05/2027. This multicenter reader study addresses a key limitation of current CTEPH diagnostic pathways-reliance on V/Q scintigraphy, which may be delayed or unavailable outside tertiary centers-by evaluating whether DCR-based pulmonary circulation analysis can improve early discrimination of CTEPH and support timely referral. Japan Registry of Clinical Trials (jRCT), jRCT2072250027.
Background: Hepatitis D virus (HDV) causes one of the most severe forms of chronic viral hepatitis. Despite its severity, universal screening of hepatitis B surface antigen (HBsAg)-positive individuals, as recommended by European guidelines, is not widely implemented. This study aimed to evaluate the yield of reflex HDV testing and to characterize HBV carriers who tested positive or negative for anti-HDV. Methods: A retrospective cohort study was conducted using the Clalit Health Services database in northern Israel (2014-2024). HBsAg-positive patients were categorized into two groups: those screened for HDV via reflex testing (2019-2024) and those tested based on clinical discretion (2014-2019). We compared these cohorts to evaluate the impact of reflex screening on coverage, diagnostic yield, and time to diagnosis. Results: Among 1336 HBsAg-positive individuals, HDV screening rates increased from 57.5% to 93.1% following reflex implementation. HDV seropositivity increased from 3.17% to 6.48% (p = 0.02). Ethiopian-born individuals had significantly higher positivity than others (10.4% vs. 3.9%, p = 0.0221). The average time from HBV diagnosis to HDV testing decreased from 38.1 ± 31 months (median 37.5) to 1.3 ± 6.1 months (median 0). Conclusions: Anti-HDV reflex testing significantly improved screening coverage, increased detection of anti-HDV seropositive cases and was associated with shorter time to serologic identification. These findings support the integration of reflex testing into national screening policies to enable earlier diagnosis and reduce the burden of infection.
The Nexus Duo endograft is a novel device designed for endovascular repair of aneurysmal disease of the aortic arch (EAAR). It features a dedicated indwelling branch for the brachiocephalic trunk and a customized retrograde branch for either the left common carotid artery (LCCA) or left subclavian artery (LSA). This study evaluates the early clinical outcomes of the Nexus Duo endograft in EAAR. This multicenter, observational, physician-sponsored registry retrospectively analyzed prospectively collected data from nine European centers. All consecutive patients who underwent EAAR with the Nexus Duo endograft between January 2023 and April 2025 were included in the study. The primary endpoint was 30-day mortality. Secondary endpoints included technical success, overall reinterventions, major adverse events (MAEs), and the incidence of endoleaks at 30 days. Thirty patients (53% female) were treated. Indications for EAAR included non-dissecting aneurysm (n=7, 23.3%), post-type-B-dissection aneurysm (n=7, 23.3%), post-type-A-dissection aneurysm (n=14, 46.7%), false aneurysm (n=1, 3.3%), and penetrating ulcer (n=1, 3.3%). The LSA was the target vessel in 17 patients (56.7%). Seventeen patients underwent simultaneous LCCA-LSA bypass (58.6% using polytetrafluoroethylene grafts). Technical success was 100%. At 30 days, the mortality rate was 3.3% (n=1), the reintervention rate was 16.7%, and the MAE rate was 6.7%. No disabling strokes were observed. One transient ischemic attack occurred and one type Ia endoleak (3.3%) were observed at 30 days. In this real-world, multicentre study, the Nexus Duo endograft demonstrated a favorable early safety profile, with high technical success, no disabling strokes, and encouraging 30-day outcomes.
ObjectiveTo assess the causal effects of specific eating habits on ischemic stroke risk and functional outcomes using Mendelian randomization.MethodsThis two-sample Mendelian randomization study used genetic variants associated with 32 eating habits as instrumental variables. Summary-level data were obtained from large-scale genome-wide association studies of individuals of European ancestry. The primary analysis used the inverse-variance weighted method, supplemented by sensitivity analyses to assess pleiotropy and multivariable Mendelian randomization to evaluate mediation via lipids and blood pressure.ResultsAfter multiple-testing correction, genetically predicted higher intake of cheese (odds ratio = 0.70, 95% confidence interval: 0.57-0.86), dried fruit (odds ratio = 0.57, 95% confidence interval: 0.41-0.80), and muesli (odds ratio = 0.20, 95% confidence interval: 0.07-0.54) showed potential protective associations with ischemic stroke risk. Sensitivity analyses supported the robustness of the findings, and multivariable Mendelian randomization indicated that the effects of cheese and muesli remained significant after adjustment for cardiovascular risk factors. No causal associations were observed for post-stroke recovery.ConclusionThis study provides genetically derived causal evidence suggesting that higher consumption of cheese, dried fruit, and muesli may reduce the risk of ischemic stroke. Further studies are warranted to validate these food-specific dietary recommendations.
Retrieval-augmented generation (RAG) is rapidly emerging as a transformative paradigm for large language models (LLMs), especially in high-stakes domains like oncology that demand precision, factual grounding, and up-to-date knowledge. By pairing LLMs with external knowledge repositories, RAG systems explicitly ground model outputs in relevant retrieved documents, helping to reduce hallucinations and ensure responses reflect current evidence. In oncology, where clinical knowledge evolves continually with new research and drug approvals, RAG offers a way to integrate the latest data (e.g., trial results, guidelines, genomic databases) into decision-making. This review synthesizes the technical foundations of RAG, including its architecture and key components, and examines current applications in oncology such as clinical decision support, patient education, radiology reporting, pathology analysis, and genomics-driven precision medicine. We highlight recent studies that demonstrate RAG's potential-for instance, improving treatment recommendations by incorporating genetic profiles and literature, and enhancing diagnostic accuracy by integrating guidelines. We also discuss emerging developments like multimodal RAG (combining text with imaging or other data), ensemble model approaches, and new explainability tools that trace model outputs to sources. Finally, we critically analyze the limitations and challenges of deploying RAG in healthcare, including computational costs, retrieval errors, noise or conflicts in retrieved information, and ethical and regulatory considerations. While RAG-based systems show promise in augmenting oncologists' expertise with timely knowledge, careful implementation, high-quality curation of knowledge bases, and human oversight will be crucial for safe and effective adoption in clinical practice.
To benchmark medical image-specific vision-language models (VLMs) against real-world radiologist-written reports, focusing on diagnostic quality, clinical acceptability, hallucinations, and language clarity. This retrospective study included adult patients who presented to the emergency department of a tertiary center between January 2022 and April 2025 and underwent same-day chest radiograph (CXR) and CT for febrile or respiratory symptoms. Reports from five VLMs (AIRead, Lingshu, MAIRA-2, MedGemma, and MedVersa) and radiologist-written reports were randomly presented and blindly evaluated by three thoracic radiologists using four criteria: RADPEER, clinical acceptability, hallucination, and language clarity. Comparative performance was assessed using generalized linear mixed models, with radiologist-written reports treated as the reference. Finding-level analyses were also performed with CT as a reference. A total of 478 patients (median age, 67 years [interquartile range, 50-78]; 282 males [59.0%]) were included. AIRead demonstrated the lowest RADPEER 3b rate (5.3% [76/1434] vs radiologists 13.9% [200/1434]; p < 0.001) and the highest clinical acceptability (84.5% [1212/1434] vs radiologists 74.3% [1065/1434]; p < 0.001), with hallucination rates comparable to radiologists (0.3% [4/1425]) vs 0.1% [1/1425]; p = 0.21). Other VLMs showed higher disagreement (16.8-43.0%; p < 0.05), lower acceptability (41.1-71.4%; p < 0.05), and more frequent hallucinations (5.4-17.4%; p < 0.05). Language clarity was higher for several VLMs (AIRead, Lingshu, and MedVersa) than for radiologist-written reports (82.9-88.4% [1189-1268/1434] vs 78.1% [1120/1434]; p < 0.05). Finding-level analyses showed substantial variability in sensitivity across VLMs for common thoracic findings. Medical VLMs for CXR report generation exhibited variable performance in report quality and diagnostic measures. Question How do medical VLMs perform compared with radiologist-written reports regarding diagnostic quality, clinical acceptability, hallucinations, and language clarity for CXRs? Findings One non-open-source VLM achieved the lowest RADPEER 3b rate and highest acceptability with radiologist-level hallucinations, whereas other models showed inferior performance. Clinical relevance VLMs may support automated preliminary CXR reporting and serve as adjunct tools to enhance workflow efficiency and consistency. Although performance varied widely, carefully developed models approached radiologist-level quality, supporting continued refinement and targeted clinical integration.