Selecting a machine learning model for higher-education quality assurance is a multi-criteria decision problem that cannot be reduced to a single leaderboard metric. This paper presents a two-step disclosure protocol in which users first make an unaided model choice and then revise it after structured multi-criteria disclosure including criterion weights, normalized values, and a ranked recommendation. The overall study used a three-step experimental protocol, with the disclosure manipulation itself implemented as a two-stage intervention within Step 2. The protocol was evaluated with 38 participants under a stable Dean-oriented advisory framing across three institutional prediction tasks, yielding 228 confirmatory scenarios after predefined quality filters. Decision quality was operationalized as regret reduction relative to a frozen governance-oriented multi-criteria scoring policy. Results show that structured disclosure significantly improved policy-aligned decision quality (Wilcoxon p = 2.30 × 10⁻11, rank-biserial r = 0.864) and increased self-reported decision confidence (p = 7.62 × 10⁻12, r = 0.646). Importantly, significant improvement was already observed in the information-only stage before any explicit recommendation was shown (p = 5.18 × 10⁻5, r = 0.629). Post-decision trust changes were small and did not reach significance in the confirmatory analysis, and are therefore treated as exploratory. The findings provide protocol-level evidence that structured multi-criteria disclosure can improve alignment with a predefined governance-oriented model selection policy in educational QA settings.

Carbapenem-resistant Klebsiella pneumoniae (CRKP) is an emerging global threat. This study aimed to determine the prevalence of CRKP, the genetic basis of antimicrobial resistance, including beta-lactamase production, efficacy of novel beta-lactam/beta-lactamase inhibitor (BLBLI) combinations, hypervirulence, and genetic diversity of circulating clones in a Serbian hospital setting. From 2022-2023, 2001 K. pneumoniae isolates were collected from 16 hospitals across Serbia. The prevalence of CRKP was 53.4% (N = 1069). Among these, 191 randomly selected CRKP isolates were subjected to expanded antimicrobial susceptibility testing, string test, and carbapenemase production, with 150 further randomly chosen for whole-genome sequencing. Resistance rates to ceftazidime-avibactam, imipenem-relebactam, and meropenem-vaborbactam were 50.8% (N = 97), 84.8% (N = 162), and 90.6% (N = 173), respectively. The majority of CRKP isolates (N = 146; 97.3%) harboured carbapenemase-encoding genes: blaNDM-1 (N = 65; 44.5%), blaOXA-48 (N = 64; 43.8%), and blaKPC-2 (N = 11; 7.5%). Additionally, six CRKP isolates co-harbored blaNDM-1 and blaOXA-48 (4.1%). This study revealed ten sequence types (STs) and six clonal complexes (CCs), with ST147/CC147/blaNDM-1 being the most prevalent (N = 44; 29.3%) followed by ST101/CC101/blaOXA-48 (N = 40; 26.7%). One CRKP isolate, ST101/blaNDM-1/blaSHV-1/blaCTX-M15 was resistant to cefiderocol. The predominant hypervirulence-associated genes were ybt (N = 138; 92%) and iuc (N = 80; 53.3%). According to the genotypic analysis, 75 out of 150 (50.0%) CRKP isolates had iuc and rmpA2/rmpADC genes, whereas 26 (13.6%) strains exhibited the hypermucoviscous phenotype. The emergence of hypervirulent, K. pneumoniae clone ST147/blaNDM-1, suggests a high potential for regional spread of a high-risk clone resistant to last-line antibiotics.

Immune cells mediate acute and chronic renal failure in native and transplanted kidneys, initiating auto- and allo-immunity, and acting as effectors in other diseases such as diabetes, hypertension, and cancer. Many drugs already in use or in development for these diseases target the putative immune mechanisms at play, based on in vitro cell experiments and animal models. Here, we review how recent and upcoming advanced tissue imaging techniques-many of them applicable to human kidney samples as well as animal models-could further improve drug development by providing insights into immune cell types, activation states, and behaviours in kidney disease. We illustrate an innovative cross-scale multimodal imaging pipeline and its application to the investigation of immune cells in human kidney samples.

Immunotherapy has revolutionized cancer treatment, yet only a minority of individuals respond clinically, necessitating alternative strategies that can benefit these patients. Novel immuno-oncology targets may achieve this through bypassing resistance mechanisms to standard therapies. We introduce Mining Immunotherapy Drug tArgetS (MIDAS), a multimodal graph neural network system for immuno-oncology target discovery. MIDAS leverages gene interactions, multi-omic patient profiles, immune cell biology, antigen processing, disease associations and phenotypic consequences of genetic perturbations. It generalizes to time-sliced data, outcompetes state-of-the-art baselines (including OpenTargets) and ranks approved targets above those in clinical development. Moreover, MIDAS recovers immunotherapy-response-associated genes in unseen patients, thereby capturing immunotherapy response determinants. Interpretability analyses reveal a reliance on autoimmunity, regulatory networks and immuno-oncology pathways. Functionally perturbing oncostatin M–oncostatin M receptor signalling, a proposed MIDAS target, in TRACERx melanoma-patient-derived explants yielded reduced dysfunctional CD8+ T cells, which associate with immunotherapy response, and reduced CCL4 levels. Furthermore, oncostatin M and oncostatin M receptor expression is associated with altered T cell and macrophage profiles in bulk transcriptomic data from patient samples. These data are consistent with a role for oncostatin M–oncostatin M in modulating the tumour microenvironment towards immunosuppressive, tumour-promoting phenotypes. Our results present a machine learning framework for analysing multimodal data for immuno-oncology target discovery. Augustine et al. present a multimodal graph neural network that identifies cancer immunotherapy targets. It distinguishes approved and prospective targets, and promising candidates are validated using a clinically relevant patient-derived platform.

Selecting a machine-learning model for higher-education quality assurance (QA) is a governance decision that requires defensible trade-offs. This paper examines post-exposure acceptance, trust, and intended reliance in a staged multi-criteria AI decision-support workflow after repeated within-session use across six scenarios. Using a primary analytic subset from a controlled study (N = 38 participants, 228 completed scenarios), we distinguish perceived helpfulness, trust, policy acceptance, and intended reliance rather than collapsing them into a single attitude measure. The workflow was viewed as helpful (M = 4.05/5) and moderately trustworthy, while self-reported intended reliance remained moderate (M = 3.34/5). Scenario-level counterfactual trust indicated a heterogeneous pattern: 50.0% unchanged, 36.8% lower without the workflow, and 13.2% higher. An exploratory participant-level correlation pattern links reliance-related outcomes to evidence inspection rather than to recommendation uptake alone. A broader completed-all comparison set (N = 90) showed a directionally consistent profile. The observed pattern is more consistent with selective reliance than with undifferentiated recommendation uptake, although the correlational design does not permit strong causal claims. Analysis scripts and anonymized data for the locked snapshot are available in a public reproducibility package [15].

Titanium dioxide nanoparticles (TiO₂NPs) are widely produced engineered nanomaterials with ongoing human exposure through consumer and occupational uses. Conventional in vitro assays often focus on cytotoxicity and may therefore overlook early or sublethal cellular perturbations. Here, we applied Cell Painting-based phenomics to resolve size-dependent sub-lethal phenotypic signatures of TiO2NP exposure in human HepG2 hepatocytes. Two TiO2NPs (<25 nm and <100 nm) were characterized by field emission scanning electron microscopy and evaluated following 24-hour exposure at five concentrations: 6.25, 12.5, 25, 50, and 100 µg/mL. Cell viability was assessed using the alamarBlue assay, and high-dimensional phenotypic profiles were generated using Cell Painting-based phenomics, including automated high-content imaging and CellProfiler-based feature extraction. TiO2NP exposure induced modest reductions in viability at the highest concentration, indicating limited acute cytotoxicity. In contrast, phenomic profiling revealed clear, concentration-dependent phenotypic perturbations for both size fractions, with markedly stronger and more consistent effects for the < 100 nm TiO2NPs. At 100 µg/mL, the < 100 nm TiO2NPs altered 50.9% of the measured phenotypic features, compared with 28.9% for the < 25 nm particles, with prominent contributions from endoplasmic reticulum-, actin/Golgi/plasma membrane-, mitochondria-, and RNA-associated features. Dimensionality reduction and correlation analyses confirmed reproducible, concentration-dependent phenotypic trajectories. Importantly, the TiO2NP-induced phenotypes were distinct from those induced by the reference chemical CA-074Me, which produced broad perturbations and served as a reference chemical to verify assay sensitivity and dynamic range. Overall, Cell Painting phenomics sensitively captures size-dependent, sublethal cellular phenotypes induced by TiO2NPs, supporting its value as a New Approach Methodology for nanosafety assessment beyond conventional viability endpoints.

In conditions where information and communication technologies (ICT) dictate the “rules” of the market, the strong promotion and development of innovation-oriented small and medium-sized enterprises (SMEs) are essential. The transition from a traditional, linear system of waste management and fleet management in utility companies to a digital and circular-oriented system represents not only a significant challenge

Highlights What are the main findings? Quantitative 2D and 3D morphometric descriptors, combined with LLM-based interpretation, enable a structured and informed assessment of pediatric foreign-body aspiration risk. Object morphology, including shape, sharpness, and orientation, substantially influences both where an aspirated foreign body lodges in the pediatric airway and the severity of the resulting injury, beyond traditional size-based criteria. What are the implications of the main findings? For practicing pediatricians, incorporating shape-related metrics into hazard assessment may complement existing size-based safety standards and improve the bedside identification of high-risk objects. The proposed framework offers a basis for developing interpretable tools to support prevention strategies, product design, caregiver education, and clinical risk assessment. Abstract Background/Objectives: Foreign-body aspiration (FBA) is a common and largely preventable pediatric emergency, yet current safety standards and risk assessments rely predominantly on object size and on anecdotal descriptions and bronchoscopy findings. We propose a clinically oriented proof-of-concept workflow that combines high-resolution three-dimensional (3D) scanning and calibrated two-dimensional (2D) imaging of retrieved objects with radiomic shape descriptors and large language model (LLM) reasoning to support aspiration risk assessment and guide prevention. Methods: Objects were obtained from the Susy Safe registry and historical series from the University Clinical Centre Tuzla. Each object was digitized with 3D scanning and photographed with a ruler. Morphometric descriptors—including volume, surface area, sphericity, elongation, flatness, curvature and convexity—were computed from stereolithography (STL) meshes; silhouette area, perimeter and Feret diameters were extracted from 2D photographs. Normative airway dimensions from radiographic and computed tomography (CT) studies provided anatomical context. A sharp, irregular metallic object recovered from a child’s laryngo-tracheal tract served as an illustrative case. Results: The object’s major axis approximated the anteroposterior glottic diameter, suggesting potential traversal when longitudinally oriented, whereas its irregular shape increased the likelihood of mucosal laceration and lodging. LLM-based synthesis provided a structured narrative interpretation consistent with a high-risk profile and highlighted preventive implications. Conclusions: Combining 2D/3D morphometry with LLM reasoning provides objective assessment of FBA hazards and may support safer product design, injury-prevention policies, and caregiver education.

Abstract Objectives Rescue stenting (RS) has emerged as a bailout strategy after failed reperfusion during endovascular treatment (EVT). Optimal blood pressure (BP) management after RS remains unclear. Our aim is to evaluate the association of BP levels and blood pressure variability (BPV) during the first 24 h after RS with short-term and long-term patient outcomes. Methods We performed a retrospective analysis of an international registry where data from adult patients who underwent either RS or rescue angioplasty after failed EVT were collected. Patients who received RS with large vessel occlusion and at least 4 BP measurements in the first 24 h were included. Results RS was performed in 437 patients (40.5% female, mean age 67.1 ± 13 years). Admission median National Institutes of Health Stroke Scale score was 12 (IQR 7–18) and history of hypertension was present in 74.2% of patients. Μean Systolic BP (SBP) in the first 24 h was 137.4 ± 14.6 mmHg. Higher values of BPV (coefficient of variation, standard deviation, average real variability and successive variation) were associated with lower odds for Modified Rankin Scale score 0–2 at 90 days (adjusted odds ratio ranging 0.55 [0.38, 0.79] to 0.99 [0.98, 0.99] per 10 units increase). No associations were found between any SBP measure and death, sICH as well as neurological deterioration at 24 h. Conclusion In our study, higher BPV was associated with worse clinical outcomes in stroke patients treated with RS as bailout therapy after failed reperfusion. No association was shown between mean, maximum, minimum and delta SBP and clinical outcomes.