Treatment with glucagon-like peptide-1 receptor agonists such as semaglutide is generally subject to strict reimbursement criteria. This leads to a growing off-label market outside of the regular healthcare system without strict quality controls, which may jeopardize patient safety. Two cases are reported of patients without diabetes or obesity, who had injected semaglutide subcutaneously that was obtained without a prescription. Both presented in a coma with severe hypoglycemia, hypokalemia and hypothermia, which was caused by an intoxication with synthetic insulin. Monotherapy with semaglutide is not associated with significant hypoglycemia in healthy individuals. Up to two days of continuous glucose infusion was required to keep patients euglykemic. The use of illegally obtained semaglutide may lead to life-threatening intoxications, due to insufficient safety precautions for the provided medication. Timely notification of cases to the national health regulatory authority is essential to mitigate further distribution of falsified drugs.
There are no studies comparing the primary therapeutic options for laryngopharyngeal reflux disease (LPRD). To compare the effectiveness of antireflux diet and 3 medication classes in patients with LPRD. This is a retrospective analysis of consecutive patients with LPRD who were prospectively recruited from April 2018 to February 2024 from 2 medical hospitals in Belgium and France. Data were analyzed in November 2025. Strict antireflux diet and stress reduction activities, alginates, magaldrates, or proton pump inhibitors. Age, sex, pretreatment to posttreatment reflux symptom score (RSS), and reflux sign assessment (RSA) were prospectively collected in patients with an objective LPRD diagnosis at the 24-hour hypopharyngeal-esophageal multichannel intraluminal impedance-pH testing. The pretreatment to posttreatment clinical features (RSS, RSA, and therapeutic response rates) were compared across patient groups regarding treatment regimens. Of 145 included patients, 84 (57.9%) were female, and the median (IQR) age was 53 (37.0-63.3) years. Patients treated with proton pump inhibitors (n = 32), antacids (n = 27), alginates (n = 38), and strict diet (n = 48) were comparable for median age, sex proportions, and baseline RSS and RSA. RSS and RSA demonstrated significant reduction in all groups. Linear mixed model analysis revealed a significant main effect of time on both RSS (F1,274 = 19.82; P < .001) and RSA (F1,219 = 26.47; P < .001), confirming symptom and sign improvement across all groups. For RSS, treatment group was significantly associated with RSS scores (F3,274 = 3.53; P = .02). The diet group had significantly lower estimated marginal mean RSS scores than the antacid group (mean difference, 38.73; 95% CI, 4.24-73.23). The change in RSS scores over time did not differ between groups. For RSA, neither the group effect nor the interaction reached significance. The diet group had the highest proportion of responders (diet, 39 of 48 [81.2%]; proton pump inhibitor, 18 of 32 [56.3%]; alginates, 22 of 38 [57.9%]; antacids, 20 of 27 [74.1%]), with proportion differences of 24.9 percentage points (95% CI, 4.6-45.4) compared with proton pump inhibitors, 23.4 percentage points (95% CI, 4.2-42.5) compared with alginates, and 7.2 percentage points (95% CI, -12.7 to 27.1) compared with antacids. In this cohort study, adherence to an antireflux diet and stress-reduction activities may be associated with comparable or greater symptom relief at 3 months posttreatment compared with conventional medical therapies. Randomized clinical trials comparing the 4 regimens are needed.
While deep learning has improved motor imagery (MI) brain-computer interfaces (BCIs), its "black-box" nature lacks physiological interpretability. Building upon our previous findings that cortical state transitions are governed by non-linear network dynamics, this study aims to elucidate subject-specific functional network delays during MI and propose a physiologically transparent BCI architecture incorporating these functional network temporal delays. 
Approach: We analyzed 4-class MI EEG data (sensorimotor μ and β rhythms, 8-30 Hz) from the full cohort of 109 subjects in the PhysioNet dataset. To effectively mitigate instantaneous volume conduction effects, we utilized partial correlation-based True Transfer Entropy (True-TE) to extract the optimal functional causal latency (τopt) of information between the supplementary motor area and the primary motor cortex. We then proposed a Tangent Space Fusion (TSF-PDER) framework, independently projecting the current and delayed spatial covariance matrices into the Riemannian tangent space before fusion to prevent topological degradation.
Main results: Under a strict, leakage-free nested cross-validation where τopt was estimated exclusively within the training folds, the extracted personalized latencies exhibited a wide functional distribution (median: 374.0 ms). Incorporating TSF-PDER significantly outperformed the spatial-only Riemannian baseline (mean accuracy: 47.24% vs. 45.70%, Wilcoxon signed-rank p = 1.577e-04), while a deep learning baseline (EEGNet) achieved only 28.53% under strictly limited data conditions. Furthermore, bidirectional control analysis revealed significantly stronger feedback information flow than feedforward flow. External validation on the BCI Competition IV-2a dataset demonstrated consistent improvements, with TSF-PDER achieving an average accuracy of 61.92% (vs. baseline 59.07%).
Significance: MI execution involves personalized, long-range functional network loops. Fusing these personalized functional delays within the Riemannian tangent space provides a robust decoding boundary without topological degradation. Consequently, TSF-PDER offers a computationally lightweight proof-of-concept for an interpretable BCI, paving the way for personalized neurorehabilitation tailored to patient-specific cortical network dynamics.
We aimed to explore whether the admission serum amylase levels may help inform the selection of medical management in patients with AAST Grade I pancreatic injuries. We retrospectively reviewed the records of 20 patients treated for pancreatic trauma between 2010 and 2024. They were categorized into medical management or intervention groups according to the initial treatment strategy. We used a receiver operating characteristic (ROC) analysis to evaluate the predictive value of the admission serum amylase levels for successful medical management. Eight patients were treated with strict medical management, whereas 12 required additional interventions. All patients in the medical management group had AAST Grade I injuries, and none required conversion to interventional treatment. The admission serum amylase levels were significantly lower in the medical management group (median 106 vs. 350 IU/L; p = 0.0228), whereas there were no significant intergroup differences in inflammatory markers. The ROC analysis yielded an area under the curve of 0.83 (95% CI, 0.59- 1.00). The cutoff value was 158 IU/L, with a sensitivity of 87.5% and a negative predictive value of 90.9%. Lower admission serum amylase levels in patients with AAST Grade I pancreatic injury may be associated with suitability for strict medical management; however, these findings are exploratory and require validation in larger multicenter studies.
Idiopathic congenital clubfoot (congenital talipes equinovarus) is a common congenital musculoskeletal deformity with an incidence of approximately 1-2 per 1000 live births. The Ponseti method is the current standard of treatment and requires prolonged parental involvement, repeated hospital visits, and strict adherence to bracing protocols. These demands may impose psychological stress on caregivers. This study aimed to determine the prevalence of anxiety among parents of children with idiopathic congenital clubfoot and identify factors associated with increased anxiety levels. A cross-sectional study was conducted at the clubfoot clinic of the Department of Pediatric Orthopedics, King George's Medical University, Lucknow. Parents of children undergoing treatment for idiopathic clubfoot using the Ponseti method were enrolled after obtaining informed consent. Sociodemographic characteristics and clinical variables were recorded. Anxiety was assessed using the anxiety subscale of the Depression Anxiety Stress Scale-21 (DASS-21). Perceived social support and stressful life events were evaluated using validated questionnaires. Bivariate analysis was performed to identify factors associated with increased anxiety scores. A total of 190 parents were included in the analysis. The mean age of participants was 28.51 ± 4.72 years, with mothers constituting 76.3% of respondents. Thirty-eight (20%) parents had anxiety scores above the normal range on the DASS-21 anxiety subscale. More than half of the participants (52.1%) belonged to rural areas. On bivariate analysis, female gender, education up to grade 12, possession of a Below Poverty Line card, lower perceived social support scores, and higher stressful life event scores were significantly associated with elevated anxiety levels. A considerable proportion of parents of children undergoing clubfoot treatment experience anxiety. Identifying vulnerable caregivers and incorporating psychological support and counseling into clubfoot care programs may improve caregiver well-being and support treatment adherence. Résumé Contexte:Le pied bot varus équin congénital idiopathique (talipes equinovarus congénital) constitue une malformation musculosquelettique congénitale fréquente, avec une incidence estimée à 1–2 cas pour 1 000 naissances vivantes. La méthode de Ponseti représente actuellement le traitement de référence et nécessite une implication parentale prolongée, des consultations hospitalières répétées ainsi qu’une observance rigoureuse du protocole d’orthèse. Ces exigences peuvent être à l’origine d’une détresse psychologique chez les aidants. La présente étude avait pour objectif de déterminer la prévalence de l’anxiété chez les parents d’enfants atteints de pied bot congénital idiopathique et d’identifier les facteurs associés à une augmentation des niveaux d’anxiété.Matériels et Méthodes:Une étude transversale a été menée au sein de la clinique spécialisée du pied bot du Département d’orthopédie pédiatrique de la King George’s Medical University, à Lucknow. Les parents d’enfants traités pour un pied bot congénital idiopathique selon la méthode de Ponseti ont été inclus après obtention d’un consentement éclairé. Les caractéristiques sociodémographiques ainsi que les variables cliniques ont été recueillies. L’anxiété a été évaluée à l’aide de la sous-échelle d’anxiété de la Depression Anxiety Stress Scale-21 (DASS-21). Le soutien social perçu et les événements de vie stressants ont été évalués au moyen de questionnaires validés. Une analyse bivariée a été réalisée afin d’identifier les facteurs associés à une augmentation des scores d’anxiété.Résultats:Au total, 190 parents ont été inclus dans l’analyse. L’âge moyen des participants était de 28,51 ± 4,72 ans, et les mères représentaient 76,3 % des répondants. Trente-huit parents (20 %) présentaient un score d’anxiété supérieur à la normale selon la sous-échelle d’anxiété du DASS-21. Plus de la moitié des participants (52,1 %) résidaient en milieu rural. L’analyse bivariée a montré que le sexe féminin, un niveau d’instruction limité à la classe de terminale (12e année), la détention d’une carte attestant un statut inférieur au seuil de pauvreté (Below Poverty Line), des scores plus faibles de soutien social perçu ainsi que des scores plus élevés d’événements de vie stressants étaient significativement associés à des niveaux d’anxiété plus élevés.Conclusion:Une proportion importante de parents d’enfants suivant un traitement pour pied bot présente des symptômes d’anxiété. L’identification des aidants les plus vulnérables et l’intégration d’un accompagnement psychologique et de conseils adaptés dans les programmes de prise en charge du pied bot pourraient améliorer le bien-être des aidants et favoriser l’adhésion au traitement.
Contraceptive counselling is a key preventive intervention for reducing unintended pregnancies and unsafe abortions. In a setting such as Vietnam, where contraceptive use is prevalent, the quality and consistency of counselling provided by future physicians are critical. Although this is a core competency in medical education, evidence suggests that medical students demonstrate deficits in applied knowledge, skills and confidence regarding counselling. Evidence regarding medical students' intentions to provide contraceptive counselling in Vietnam remains limited and has not been systematically examined. A nationwide cross-sectional study will be conducted from October 2025 to October 2026 at 12 medical universities representing three distinct regions. Self-administered online questionnaires (via KoboToolbox) will be distributed to first-year to sixth-year medical students. This protocol describes a two-phase study. Phase 1 (scale development and validation; n=600, including pilot and psychometric evaluation) aims to develop and validate the Extended Theory of Planned Behaviour (E-TPB) scales (Knowledge, Attitude, Subjective Norms, Perceived Behavioural Control, Intention). This phase uses the Content Validity Index, Classical Test Theory and Item Response Theory Two-Parameter Logistic to generate latent trait scores for Knowledge, followed by exploratory factor analysis, confirmatory factor analysis (CFA) and assessment of convergent and discriminant validity. Phase 2 (main survey; n=1300) describes the current status by academic year, gender and institution. Simultaneously, the E-TPB model will be tested using Covariance-Based Structural Equation Modelling to estimate standardised path coefficients and mediation effects (5000 bootstrap samples). A multigroup SEM will be conducted to test measurement/structural invariance and to compare key groups (gender; preclinical vs clinical). Missing-data handling will prioritise full information maximum likelihood for CFA/SEM, with strict adherence to predefined quality control criteria. The study protocol was formally approved by the Institutional Ethical Review Board for Biomedical Research at Hanoi Medical University on 3 March 2026 (Approval No. 120/GCN-KHCN). Results will be published in international peer-reviewed journals and sent to participating medical universities.
Hard ticks (Acari: Ixodidae) are major vectors of zoonotic pathogens affecting humans and animals. Accurate species identification is essential for surveillance and disease risk assessment but is often hindered by morphological similarity and intraspecific variation. In Malaysia, tick studies have largely relied on morphology or single-gene barcoding, with limited taxonomic and geographic coverage. This study applied a multilocus mitochondrial approach using cytochrome c oxidase I (COI) and 16S ribosomal DNA (rDNA) to evaluate species boundaries among ixodid ticks from Peninsular Malaysia. Ticks were collected from wild and domestic hosts between 2022 and 2023. A total of 319 specimens representing 14 morphologically defined species were analysed using COI and 16S rDNA sequencing, with additional reference sequences retrieved from GenBank for comparative analyses. Phylogenetic reconstruction, barcode gap assessment, and four species delimitation methods (ASAP, ABGD, bPTP, and GMYC) were employed to assess genetic divergence and identification accuracy. Molecular analyses were largely congruent with morphological identification. Distance-based delimitation methods (ASAP and ABGD) recovered operational taxonomic units (OTUs) largely consistent with morphologically defined species, whereas tree-based approaches (bPTP and GMYC) inferred substantially higher numbers of OTUs, reflecting sensitivity to intraspecific mitochondrial structuring. These additional subdivisions are interpreted conservatively as population-level genetic differentiation rather than evidence of distinct species. Clear barcode gaps were observed for COI and the concatenated COI + 16S rDNA datasets, whereas partial overlap occurred with 16S rDNA alone. COI demonstrated the highest and most consistent performance for routine species identification, while concatenated datasets improved phylogenetic resolution but reduced assignment clarity under strict barcoding criteria. Overall, COI remains the most effective mitochondrial marker for routine species identification of ixodid ticks in Peninsular Malaysia. These findings highlight the value of integrative approaches combining morphology and molecular data to strengthen tick taxonomy and support surveillance of tick-borne pathogens in the region.
Celiac disease (CD), a chronic immune-mediated enteropathy triggered by gluten ingestion, is increasingly recognized as a systemic disorder with significant extraintestinal manifestations, including effects on the cardiovascular system. This review synthesizes recent epidemiological evidence demonstrating that individuals with CD face a modest but statistically significant increase in cardiovascular disease risk, including ischemic heart disease, myocardial infarction, and venous thromboembolism, despite a lower prevalence of traditional risk factors such as obesity, hypertension, and smoking. This so-called "risk factor paradox" underscores the significance of nontraditional, disease-specific mechanisms, including chronic systemic inflammation, immune dysregulation, nutritional deficiencies, and the metabolic consequences of a gluten-free diet. Emerging data also highlight the early onset of nontraditional cardiovascular risk in pediatric CD populations and the reversibility of some cardiac manifestations, such as arrhythmias and cardiomyopathy, with strict gluten-free diet adherence. Clinical recommendations now emphasize annual cardiovascular screening, comprehensive nutritional assessment, and multidisciplinary management in patients with CD. However, significant research gaps remain, including the need for CD-specific risk prediction tools, randomized trials on dietary quality, and greater understanding of genetic and autoimmune contributions to cardiovascular disease in CD. Addressing these gaps is essential for optimizing prevention and management strategies in this growing patient population.
Microplastics (MPs) and nanoplastics (NPs) have recently been detected in several human biological matrices; however, evidence in children remains limited. This exploratory study aimed to investigate the presence and concentration of urinary MPs and NPs (MNPs) in primary school children residing in Cyprus. First-morning urine samples from 29 children were analyzed using scanning electron microscopy coupled with energy-dispersive X-ray analysis (SEM/EDX), applying strict contamination-control measures and focusing on particles < 10 μm. Pyrolysis-GC/MS was additionally used to characterize polymer composition, identifying polyethylene (PE) and polypropylene (PP) as the predominant polymers. MNPs < 10 μm were detected in all samples, with concentrations ranging from 393 to 8050 particles/ml (median: 1217 particles/ml; IQR: 800-2030). Particle diameters ranged from 0.88 μm to 3.44 μm (median: 1.69 μm; IQR: 1.25-2.25 μm; minimum: 0.77 μm; maximum: 4.88 μm). No statistically significant associations were observed between MNP concentrations and body mass index (BMI)-for-age categories or hand-to-mouth behavior. Although direct comparisons with previous studies are limited due to methodological variability among studies and the lack of standardized protocols for MNP quantification in human urine, these findings provide preliminary evidence of urinary MNPs in children. Further large-scale studies using harmonized analytical approaches are needed to better characterize exposure patterns in pediatric populations.
Mismatch repair-deficient tumors exhibit high sensitivity to immune checkpoint inhibitors. While "watch and wait" organ preservation strategies following clinical complete response are increasingly established in rectal cancer, their application to non-rectal malignancies remains experimental. This report summarizes emerging evidence regarding non-operative management, specifically recent pivotal data on tumor-agnostic responses, and contextualizes these findings through a representative case of MSH6-deficient upper tract urothelial carcinoma. Recent prospective data suggest early substantial clinical complete response rates in selected non-rectal mismatch repair-deficient tumors (approximately 65%), enabling initial non-operative management in selected patients. We present a case of high-grade upper tract urothelial carcinoma in a patient with Muir-Torré syndrome treated with definitive pembrolizumab. The patient achieved a deep clinical response characterized by complete metabolic and endoscopic resolution, with no evidence of invasive carcinoma on biopsy, resulting in successful kidney preservation without surgery. Immunotherapy-based organ preservation may represent a viable strategy for highly selected patients with mismatch repair deficiency. This approach offers a potential selective alternative to immediate radical nephroureterectomy in deep responders, provided that strict and intensive surveillance protocols are feasible.
Transfusion transmissible infections (TTIs) such as Human Immunodeficiency Virus (HIV), hepatitis B virus (HBV), hepatitis C virus (HCV), and syphilis, remain a major public health concern in low-income countries due to inadequate blood donation systems. This highlights the urgent need for continuous surveillance and a deeper understanding of the factors associated with TTIs in Rwanda. A retrospective cross-sectional study analyzed voluntary blood donor records from the National Center for Blood Transfusion (NCBT) in Kigali, Rwanda, spanning January 2019 to December 2023. A census of all available records was used to select 38,345 eligible donors aged 18-60 years. Data were extracted from electronic databases and analyzed using SPSS version 25, with chi-square tests and multivariate logistic regression to identify predictors of transfusion-transmissible infections. The overall prevalence of transfusion-transmissible infections declined from 2019 to 2023, with HBV decreasing from 0.76% to 0.41%, HCV from 0.30% to 0.18%, HIV remaining stable at 0.04%-0.08%, and syphilis slightly increasing to 0.26%. Multivariate analysis showed that male gender was independently associated with HBV (aOR = 2.60, 95% CI: 1.54-4.38, p < 0.001) and HIV (aOR = 3.84, 95% CI: 1.63-9.05, p = 0.002). First-time donor status was independently associated with higher odds of HBV (aOR = 4.09, 95% CI: 2.26-7.39, p < 0.001) and HIV (aOR = 2.95, 95% CI: 0.77-11.34, p < 0.001). Gasabo district was independently associated with lower odds of HBV (aOR = 0.56, 95% CI: 0.34-0.90, p = 0.017) and higher odds of HCV (aOR = 2.34, 95% CI: 1.25-4.40, p = 0.008), relative to Nyarugenge. TTI prevalence among Kigali blood donors remained low with declining HBV and HCV trends from 2019 to 2023, while HIV rates stayed consistently low. Male gender, older age, first-time donor status, and specific occupations were independently associated with TTI positivity. Rwanda's blood screening protocols appear effective, though targeted interventions for high-risk groups and continued surveillance are recommended. Not applicable.
Telemedicine adoption in gastroenterology has accelerated rapidly in recent years. Virtual interventions offer disease activity control comparable to standard in-person care. However, factors determining patient satisfaction and personal perceptions of virtual clinics remain poorly explored. This study identifies the clinical, demographic, and socioeconomic predictors of clinic format preference among patients with inflammatory bowel disease. We conducted a cross-sectional observational study at King Fahad Medical City in Riyadh, Saudi Arabia. The study included adult patients diagnosed with Crohn's disease or ulcerative colitis. Authors collected demographic parameters, clinical covariates, and patient experience ratings using a structured questionnaire. A multivariable binary logistic regression model was employed to isolate independent predictors of virtual clinic preference. The cohort comprised 1,005 patients. Overall, 55.6% preferred in-person appointments and 44.4% preferred virtual clinics. Prior telehealth exposure served as the strongest predictor for future virtual care (odds ratio [OR] = 4.10; 95% confidence interval [CI]: 2.90, 5.82). Patients utilizing subcutaneous injections (OR = 2.02) or oral medications (OR = 1.81) demonstrated significantly higher odds of selecting virtual care. Conversely, an ulcerative colitis diagnosis predicted a preference for physical visits (OR = 0.63). Prioritizing ease of access (OR = 1.47) and the use of technology (OR = 2.14) underpinned virtual choices. Strict privacy concerns (OR = 0.57) and inadequate physician communication (OR = 0.05) deterred patients from the remote clinic. Prior telehealth experience and the use of home-based therapies drive virtual clinic adoption. Disease phenotype, reliance on intravenous treatments, and privacy concerns may necessitate accessible in-person care. Health care systems must abandon uniform digital strategies and implement tailored, hybrid care models to optimize resource allocation.
Bound states in the continuum (BIC) leverage symmetry-protected resonant modes for exceptional light confinement, yet their leaky modes are almost underutilized. Meanwhile, multiple quantum well (MQW) structures face limited optical absorption due to strict transition selection rules. We demonstrate the regulation of the leaky mode of quasi-BIC (QBIC) by analyzing MQW-vertical field coupling, revealing that increasing asymmetric parameters enhances the transverse leakage of wave vector and optical field nonlinearly. This drives a nonlinear photoresponse as increasing asymmetry parameter, while linear scenario with incident angle and external bias voltage. We then develop an optoelectrical fusion neuromorphic processor, implementing QBIC-MQWs into an artificial neural network for machine vision applications. Graphical abstract. This work presents a proof-of-concept BIC-MQW device, in which the photocurrent characteristics are leveraged to realize image processing functionalities.
Research interest in positive body image has increased in recent years, marking a shift away from a traditional focus on negative body image. We aimed to examine the psychometric properties of the Body Appreciation Scale 2 (BAS-2) and its short form (BAS-2SF) in a Danish sample aged 15-25 years and explore their level of measurement invariance across age and gender groups. In this cross-sectional study, respondents completed an online survey that was disseminated through social media, newsletters, and schools. The survey included the BAS-2, BAS-2SF, and related scales to explore convergent validity. We conducted a Confirmatory Factor Analysis (CFA) to test unidimensionality, we tested internal consistency, and measurement invariance across genders (men, women, other) and age groups (15-20, 21-25). We analyzed 2005 complete responses. Of these, 471 were men, 1435 were women, and 99 were individuals of other gender identities. The mean age of participants was 21.7 years (SD = 2.5). CFA supported unidimensionality for BAS-2 and BAS-2SF, and internal consistency was also supported. Fit indices supported scalar to strict measurement invariance across age and gender groups. Correlations with related scales supported convergent validity. Group comparison showed that men reported the highest body appreciation, followed by women, and individuals of other gender identities. The differences between age groups were insignificant. The BAS-2 and BAS-2SF exhibited good psychometric properties in a Danish sample aged 15-25 years.
Public leaderboards such as the Therapeutics Data Commons (TDC) ADMET benchmark are widely treated as a ranking of state-of-the-art models. However, a high leaderboard position is only meaningful if the corresponding model can actually be reproduced and deployed by an independent researcher. In this work, we audit whether the top-ranked TDC ADMET models meet that bar. We assessed the top-ranked models of all 22 TDC ADMET leaderboards from the perspective of an end user with access only to the publicly released artifacts of each model─its publication, code repository, and installation instructions. For every end point, the top three models were screened with a unified protocol including an execution environment reproducibility check, a data-leakage assessment, verification of the hyperparameter-optimization procedure, and a reevaluation against the current leaderboard. Only three models (CaliciBoost, MapLight, and MapLight + GNN) passed all stages and reproduced their reported performance. The remaining models failed because of unavailable code, nonreproducible environments, runtime incompatibilities, or methodological flaws. We traced direct or indirect data leakage in the MiniMol, GradientBoost, and XGBoost models, and used deliberately overfitted variants of our own Mol2Vec-based models to show that tuning on the public test set─whether accidental or intentional─can substantially inflate both metrics and leaderboard rank. These results indicate that current TDC leaderboard positions cannot be read as a direct measure of model quality and practical applicability and emphasize the urgent need for better public ADMET benchmarks based on the hidden test sets, strict data set versioning and model submission with standardized inference environments.
Toxicity and safety prediction models are now common in virtual screening and drug discovery, but conventional benchmark splits can overstate deployment reliability. Under scaffold shift, strict out-of-distribution (OOD) shift, or transfer across curated datasets, a model may assign high confidence to predictions that should be reviewed. We propose MARS, a multi-anchor reasoning framework for reliability-centered toxicity prediction under distribution shift. MARS combines a molecular encoder with evidence retrieved from multiple training anchors and trains a dedicated shift-aware reliability head, separating the toxicity score from the trust signal. For each target molecule, the framework uses target representations, anchor similarities, anchor labels, local disagreement, and novelty summaries to reason about both prediction and confidence. We evaluate MARS on seven public toxicity and safety tasks covering 24,256 molecules, using scaffold split, three strict OOD split families, nine cross-dataset transfer pairs, post-hoc calibration, conformal prediction, risk-control, and curated molecule-level cases. Ours-Full reaches 0.776 area under the receiver operating characteristic curve (AUROC) and 0.131 expected calibration error (ECE) in scaffold summaries. Across the retained transfer pairs, Ours-Anchor reduces ECE from 0.176 to 0.127 relative to random-forest Morgan-fingerprint (RF-Morgan). Within the anchor model, the learned reliability head raises error-detection AUROC from 0.672 to 0.719 and reduces risk-coverage area under the curve (AUC) from 0.176 to 0.153 compared with probability-margin confidence. These results suggest that toxicity models should be judged not only by discrimination, but also by whether confidence remains usable when chemistry shifts.
Accurate segmentation of the left atrial appendage (LAA) is essential for device occlusion planning in atrial fibrillation patients who cannot receive anticoagulation. Yet 3D echocardiography suffers from low signal-to-noise ratio, anisotropy and marked morphological variability, increasing overfitting risk and reliance on operator-dependent post-processing. Existing U-Net variants capture local detail but often miss long-range dependencies, while transformers improve context at the cost of boundary precision; semi-automated pipelines still require experts. We present a fully automated, AERO-optimised DAT-DAD U-Net with SE-augmented skip fusion. Deformable attention transformers (DAT) provide content-adaptive global context, while a dual attention with deformable convolution (DAD) block refines rims and addresses shape irregularities; spatial-channel squeeze-and-excitation improves multi-scale fusion. Hyperparameters are selected by AERO, a surrogate- and multi-fidelity-driven optimiser balancing exploration and exploitation under limited data in practice. Validation used a 22-patient 3D echocardiography cohort from King's College Hospital. Volumes were reformatted into axial 2D slices, trained with on-the-fly anatomy-preserving augmentation and evaluated using strict patient-wise splits. The model achieved Dice 0.8925 ± 0.0144, IoU 0.8026 ± 0.0156 and HD95 9.14 ± 1.96 mm. Ablations confirmed additive gains from DAT, DAD and SE, with faster convergence and a lower error floor, supporting operator-light, time-sensitive LAA workflows.
To develop and evaluate deep learning models for upper abdominal ultrasound standard section recognition and downstream multi-organ segmentation. This retrospective study used real-world clinical ultrasound data from a single center. A total of 521 upper abdominal ultrasound examination videos were collected, of which 465 met the inclusion criteria. Standard section frames were extracted and reviewed by sonographers, yielding 5535 images representing 12 clinically defined upper abdominal standard sections across the liver, gallbladder, pancreas, and spleen. A two-stage sequential framework was implemented, in which standard section recognition using a convolutional neural network served as a prerequisite screening step, followed by organ segmentation using a deep learning-based segmentation network applied only to recognized standard sections. Data were divided into training, validation, and test sets with strict case-level separation. The standard section recognition model demonstrated stable performance across datasets, achieving Precision 97.08%, Recall 97.00%, Specificity 99.70%, and Accuracy 97.00% on the test set. Organ segmentation performed under standard section constraints yielded strong results for the liver, gallbladder, and right kidney, whereas performance was lower for the spleen and particularly the pancreas (test IoU 53.76%). Performance patterns were consistent across the validation and test subsets. Deep learning models for standard section recognition and multi-organ segmentation are feasible within this single-center real-world dataset and address two complementary tasks in upper abdominal ultrasound; external validation is still needed to determine broader generalizability.
Homalonotidae is a morphologically distinctive trilobite family well known in systematics studies, yet phylogenetic studies remain scarce. This study investigated internal relationships among taxa within the Homalonotidae family and reassessed the classification of problematic species. A phylogenetic analysis was conducted using a matrix composed of 36 taxa and 56 morphological characters of the trilobite exoskeleton. Analysis under the TBR algorithm recovered eight most parsimonious trees, from which a strict consensus tree was obtained. The results revealed substantial inconsistencies in the diagnosis of several genera and species, particularly due to the presence of paraphyletic and polyphyletic groupings. Some taxa, such as Burmeisteria fontinalis, Dipleura dekayi, and Burmeisterella braziliensis, remained phylogenetically stable and well-supported. Conversely, certain species traditionally assigned to Burmeisteria, including B. oiara and B. antarcticus, behaved as wild-cards within the trees, leading to their reclassification as Burmeisteria sp. Moreover, strong phylogenetic correlations were identified between species from different genera. These findings challenge the current taxonomic framework and allow for the proposal of new taxonomic arrangements. Furthermore, the relationships among the taxa indicate that Devonian Homalonotidae reflect a complex palaeobiogeographic system and further suggests that the Gondwana basins functioned as a dynamically connected network during the Devonian period.
Monolithic clinical decision support architectures entangle IT infrastructure with medical logic, compounding regulatory validation overhead (e.g., IEC 82304-1). We present an event-driven microservice architecture designed to streamline the clinical validation process through structural fault containment. Orchestrated by a Kafka message broker and gated by a Drools rule engine, the system intercepts structurally invalid data before it reaches the deterministic medical logic core comprised of ArdenSuite. Empirical testing demonstrated zero state loss during node failures and confirmed that shedding malformed data at the architectural perimeter actively prevents computational bottlenecks, effectively reducing the mean evaluation time per valid input during high-stress scenarios. By enforcing strict architectural boundaries, the pipeline decouples IT reliability from clinical safety. This confines the rigorous clinical validation burden entirely to the isolated clinical interpretation engine. The resulting secure containment boundary accelerates the integration of diverse data ingestion modalities and the continuous adaptation of clinical logic, equipping healthcare systems with an agile, compliant foundation for evolving decision support.