Search Results (90)

Background/Objectives: The Prognostic Nutritional Index (PNI), which integrates serum albumin and lymphocyte count, reflects both nutritional and inflammatory status. However, its role in early renal function decline among patients with type 2 diabetes (T2D), particularly in relation to glycemic control, remains unclear. This study aimed to: (1) characterize the dose–response relationship between PNI and early renal function decline in type 2 diabetes using restricted cubic splines; (2) identify whether glycemic control (HbA1c) modifies the PNI–renal decline association; and (3) evaluate the clinical utility of combining PNI and HbA1c for risk stratification. Methods: We analyzed data from 1711 community-based participants with T2D who had preserved renal function at baseline. The PNI was calculated as serum albumin (g/L) + 5 × lymphocyte count (×10⁹/L). The primary outcome was a composite of rapid estimated glomerular filtration rate (eGFR) decline (>3 mL/min/1.73 m² per year) or incident chronic kidney disease (CKD) stage 3. Restricted cubic spline models, multivariable regression, and Johnson–Neyman analyses were used to examine non-linearity and effect modification by glycated hemoglobin (HbA1c). Results: A consistent inverse linear association was observed between PNI and the rate of eGFR decline (P for non-linearity > 0.05). Johnson–Neyman analysis further demonstrated that the protective association of PNI was statistically significant within an HbA1c range of 7.24% to 8.71%. Stratification by clinical cut-offs revealed a significant effect modification by glycemic status. The inverse linear association between PNI and renal risk was most pronounced under hyperglycemic stress, as evidenced by the markedly elevated incidence (50.0%) among individuals with both poor glycemic control (HbA1c ≥ 8%) and low PNI (<50). Conversely, under good glycemic control (HbA1c < 8%), this inverse association was substantially attenuated, with a lower incidence observed in the low-PNI subgroup (6.7%) than in the high-PNI subgroup (15.9%). These findings indicate that the protective role of PNI is conditional upon the glycemic milieu. Conclusions: The PNI demonstrates a stable linear association with early renal function decline in T2D, with its protective effect most pronounced at suboptimal HbA1c levels. Combining PNI and HbA1c effectively identifies a high-risk subgroup characterized by synergistic risk, underscoring the need for integrated nutritional and glycemic management. Full article

In response to the critical demand for improved characterization of atmospheric stability effects in wind turbine wake prediction, this study proposes and systematically validates a new analytical wake model that incorporates atmospheric stability effects. In recent years, research on wake models with atmospheric stability effects has primarily followed two approaches: incorporating stability through high-fidelity numerical simulations or modifying classical analytical wake models. While the former offers clear mechanical insights, it incurs high computational costs, whereas the latter improves efficiency yet often suffers from near-wake prediction biases under stable stratification, lacks a unified framework covering the entire wake region, and relies heavily on case-specific calibration of key parameters. To overcome these limitations, this study introduces a stability-dependent turbulence expansion term with a square of a cosine function and the stability sign parameter, enabling the model to dynamically respond to varying atmospheric conditions and overcome the reliance of traditional models on neutral atmospheric assumptions. It achieves physically consistent descriptions of turbulence suppression under stable conditions and convective enhancement under unstable conditions. A newly developed far-field decay function effectively coordinates near-wake and far-wake evolution, maintaining computational efficiency while significantly improving prediction accuracy under complex stability conditions. The Present model has been validated against field measurements from the Scaled Wind Farm Technology (SWiFT) facility and the Alsvik wind farm, demonstrating superior performance in predicting wake velocity distributions on both vertical and horizontal planes. It also exhibits strong adaptability under neutral, stable, and unstable atmospheric conditions. This proposed framework provides a reliable tool for wind turbine layout optimization and power output forecasting under realistic atmospheric stability conditions. Full article

Background: Immune checkpoint inhibitors (ICIs) improve cancer outcomes but may cause cardiovascular toxicity, including early subclinical myocardial injury. Conventional echocardiography has limited sensitivity, whereas speckle-tracking echocardiography (STE) allows for early detection of myocardial deformation. Data on short-term ICI-related effects on biventricular mechanics are limited, and atrial function remains poorly characterized. This study evaluated the early impact of ICI therapy on biventricular and biatrial mechanics using STE in patients with advanced cancer. Methods: In this prospective, single-center study, 28 consecutive patients with advanced cancer undergoing ICI therapy were followed for 3 months. Clinical, laboratory, electrocardiographic, and echocardiographic assessments were performed at baseline, 1 month, and 3 months. STE was used to assess left ventricular global longitudinal strain (LV-GLS) and circumferential strain; right ventricular GLS (RV-GLS); and left and right atrial reservoir, conduit, and contractile strain parameters. Subclinical LV dysfunction was defined as a relative LV-GLS reduction >15%. Logistic and Cox regression analyses identified predictors of strain impairment and adverse clinical events. Results: Conventional echocardiographic parameters, including left ventricular ejection fraction, remained stable. In contrast, LV-GLS declined progressively from 20.7 ± 2.1% to 17.6 ± 2.7% at 3 months (p = 0.002), with subclinical LV dysfunction observed in 85.7% of patients. RV-GLS also deteriorated despite preserved TAPSE. Both left and right atrial strain and strain-rate parameters showed an early and marked decline, accompanied by increased left atrial stiffness despite unchanged atrial volumes. Older age and higher neutrophil-to-lymphocyte ratio (NLR) were associated with LV-GLS impairment. Over a mean follow-up of 5.4 ± 3 months, baseline LV-GLS independently predicted adverse clinical events and mortality. Optimal cut-off values were 67 years for age, 4 for NLR, and 19.5% for LV-GLS. Conclusions: Short-term ICI therapy is associated with early, diffuse subclinical myocardial dysfunction involving both ventricles and atria, detectable only by STE. Comprehensive biventricular and biatrial strain assessment may enhance early cardio-oncology surveillance and risk stratification in ICI-treated patients. Full article

The Jiangsu Coastal Thermal Front (JCF), a persistent feature in Chinese marginal seas, plays a significant role in modulating phytoplankton dynamics and carbon cycling. However, the multi-scale spatiotemporal variability of the persistent JCF and the underlying mechanisms driving its ecological effects remain limited. Using satellite observations and reanalysis data, this study systematically investigates the JCF’s distribution and its regulatory impact on phytoplankton chlorophyll-a (Chla) and particulate organic carbon (POC). Results show the persistent JCF is most active in summer and winter, primarily in Haizhou Bay and the Jiangsu Shoal. In summer, stratification-induced nutrient limitation within the Haizhou Bay thermal front decreases Chla and POC (by ~−20% and ~−40%, respectively), whereas nutrient-replete non-frontal waters support higher biomass. In the Jiangsu Shoal, the thermal front blocks the southward transport of POC, helping to maintain stable POC levels in the nearshore non-frontal region; meanwhile, the shift from southward to northward transport leaves the offshore non-frontal area without sufficient replenishment, resulting in a ~35% decrease in POC. In winter, the Haizhou Bay thermal frontal barrier effect restricts suspended particulate matter, alleviating light limitation inside the front and enhancing Chla (up to 15%) while reducing POC due to diminished resuspension. We elucidate that the JCF shapes ecological patterns through two primary pathways: by directly acting as a barrier to material transport and by interacting with ancillary processes like upwelling. These findings advance the mechanistic understanding of frontal impacts on coastal ecosystems and provide a mechanistic basis for understanding synergistic coastal carbon sinks. Full article

Background: Early administration of antibiotics is commonly recommended for pneumonia in intensive care unit (ICU) patients. However, the clinical benefit of very early empirical treatment remains uncertain and may reflect differences in illness severity, baseline risk, or care pathways, particularly in non-septic or hemodynamically stable ICU populations. Methods: We performed a retrospective cohort study using the Medical Information Mart for Intensive Care IV (v2.2) database to evaluate the observational association between antibiotic timing and in-hospital mortality among adult ICU patients with pneumonia. Patients were categorized as receiving early (<3 h) or delayed (≥3 h) antibiotic therapy after ICU admission. A multivariable logistic regression model adjusted only for age and sex. Given the absence of detailed severity-of-illness measures, no causal inference was intended, and all analyses were considered hypothesis-generating. Additional analyses exploring antibiotic class, dosing frequency, and combination therapy were conducted in an exploratory manner, given substantial variation in sample sizes and a high risk of confounding by indication, misclassification, immortal-time, and survivorship bias. Results: Among 7569 ICU patients with pneumonia, 56.5% received antibiotics within three hours of ICU admission. Early antibiotic initiation was associated with higher in-hospital mortality than delayed therapy (26.1% vs. 21.5%; OR 1.30, 95% CI 1.16–1.44; p < 0.001). Because validated severity-of-illness measures were unavailable, residual confounding and confounding by indication are likely and may largely explain this association. A potential signal of increased mortality was observed in patients receiving ≥3 doses of levofloxacin (OR 4.39, 95% CI 1.13–17.02); however, this subgroup was small and the finding is highly susceptible to survivorship and indication bias. Mortality appeared lower in patients receiving two or three antibiotics compared with monotherapy, but marked group imbalances, lack of restriction or stratification, and clinical selection effects limit interpretability. Regimens involving ≥4 agents were rare and primarily associated with prolonged ICU length of stay rather than a clear mortality difference. Conclusions: In this large retrospective ICU cohort, very early antibiotic administration for pneumonia was observationally associated with higher in-hospital mortality. Causality cannot be inferred, and early treatment likely represents a marker of higher baseline risk or clinical urgency rather than a harmful exposure. These findings challenge the assumption that earlier antibiotic initiation is uniformly beneficial and underscore the importance of individualized, stewardship-aligned, and context-dependent decision-making regarding antimicrobial timing and intensity in critically ill patients. Full article

Background: Chronic obstructive pulmonary disease (COPD) is a heterogeneous condition marked by airway inflammation, airflow limitation, and structural remodeling. Exosomal microRNAs (exo-miRNAs) are stable, cell-free biomarkers reflecting airway molecular changes. While serum and BALF exosomal miRNAs have been examined, sputum-derived profiles remain underexplored. Methods: Induced sputum was collected from 20 clinically stable COPD patients and 10 age-matched healthy controls. Exosomes were isolated by polymer-based precipitation and verified by transmission electron microscopy and Western blotting for CD9 and CD81. Nine candidate miRNAs (miR-21, miR-155, miR-34a, miR-126, miR-210, miR-146a, miR-199a-5p, miR-223, miR-1246) were quantified by RT-qPCR. Group comparisons used the Mann–Whitney U test, correlations Pearson’s r, and diagnostic accuracy ROC analysis. Results: Sputum-derived exosomes displayed characteristic morphology and canonical protein markers. COPD patients showed significant dysregulation of exosomal miRNAs, including upregulation of miR-21 (fold change = 3.4; 95% CI: 0.12–0.64 vs. 0.18–0.22; p < 0.001) and miR-223 (fold change = 2.1; 95% CI: 0.00–3.79 vs. 0.86–1.22; p = 0.004), and downregulation of miR-155 (fold change = 0.35; 95% CI: 0.43–0.67 vs. 0.86–1.22; p = 0.002), miR-126 (fold change = 0.42; 95% CI: 0.30–0.39 vs. 0.80–1.42; p = 0.009), and miR-146a (fold change = 0.28; 95% CI: 0.49–1.12 vs. 0.87–1.35; p = 0.006). miR-21 correlated with symptom burden (CAT; r = 0.445; p = 0.049). Among individual biomarkers, miR-155 exhibited the best diagnostic performance for COPD detection (AUC = 0.730; 95% CI: 0.53–0.93), which further improved when combined with miR-126 and miR-146a (AUC = 0.841; 95% CI: 0.69–0.98). For disease stratification, miR-126 most effectively discriminated mild from moderate-to-severe COPD (AUC = 0.728; 95% CI: 0.50–0.96). These results indicate that sputum-derived exosomal miRNAs—particularly miR-155, miR-126, and miR-146a—may serve as promising non-invasive biomarkers for COPD diagnosis and clinical phenotyping. Conclusions: Sputum exosomal miRNAs reveal a distinct COPD-specific signature reflecting inflammation, impaired repair, and immune dysregulation. Composite panels incorporating miR-155, miR-126, and miR-146a enhance diagnostic accuracy and could be integrated into non-invasive workflows for COPD detection and staging. Full article

Background: Although orbital fractures are common in trauma care, age-specific mechanisms and admission physiology-based risk stratification have not been systematically characterized. This study aimed to identify age–mechanism interaction patterns and develop an admission-based physiological risk score for orbital fracture patients. Methods: This retrospective cohort study analyzed 41,464 adult orbital fracture patients from the National Trauma Data Bank (2018–2020). A three-component physiological risk score was developed using admission vital signs: severe hypotension (<90 mmHg, 2 points), tachycardia (>100 bpm, 1 point), and severe traumatic brain injury (GCS ≤ 8, 1 point). Risk stratification performance was validated against composite adverse outcomes. Results: Distinct age–mechanism patterns emerged: 74.0% of elderly patients (≥65 years) sustained falls, while young adults demonstrated a bimodal distribution with motor vehicle crashes (31.2%) and violence (28.4%). Violence-related injuries occurred in younger patients (40.3 vs. 55.0 years) but had lower injury severity scores (10.0 vs. 14.4) and mortality (2.8% vs. 5.2%) than accidental mechanisms. High-/critical-risk patients (8.4% of the cohort) had 16.2% mortality versus 2.1% in stable patients. Complex facial injuries demonstrated 11-fold higher mortality (7.7% vs. 0.7%). The physiologic risk score achieved AUC 0.79 (95% CI: 0.78–0.80). Conclusions: Age–mechanism interactions revealed distinct bimodal injury patterns in young adults. Admission physiologic parameters effectively identify 8.4% of patients requiring intensive resources, while violence-related injuries paradoxically demonstrate better outcomes than accidental mechanisms. Full article

Breast cancer is the second leading cause of cancer-related mortality among women. Early detection enables timely treatment, improving survival outcomes. This paper presents a comparative evaluation of federated learning (FL) frameworks for multiclass breast cancer classification using ultrasound images drawn from three datasets: BUSI, BUS-UCLM, and BCMID, which include 600, 38, and 323 patients, respectively. Five state-of-the-art networks were tested, with MobileNet, ResNet and InceptionNet identified as the most effective for FL deployment. Two aggregation strategies, FedAvg and FedProx, were assessed under varying levels of data heterogeneity in two and three client settings. Results from experiments indicate that the FL models outperformed local and centralized training, bypassing the adverse impacts of data isolation and domain shift. In the two-client federations, FL achieving up to 8% higher accuracy and almost 6% higher macro-F1 scores on average that local and centralized training. FedProx on MobileNet maintained a stable performance in the three-client federation with best average accuracy of 73.31%, and macro-F1 of 67.3% despite stronger heterogeneity. Consequently, these results suggest that the proposed multiclass model has the potential to support clinical workflows by assisting in automated risk stratification. If deployed, such a system could allow radiologists to prioritize high-risk patients more effectively. The findings emphasize the potential of federated learning as a scalable, privacy-preserving infrastructure for collaborative medical imaging and breast cancer diagnosis. Full article

Background/Objectives: Coronary artery disease (CAD) is a leading cause of global mortality, necessitating effective risk stratification tools for optimal patient management. The electrical risk score (ERS) is a multi-parametric index incorporating various electrocardiographic (ECG) parameters, previously shown to predict unfavorable cardiovascular outcomes. However, the relationship between ERS and the presence and severity of CAD remains unclear. This study aimed to investigate the association of ERS with the presence and extent of CAD as assessed by coronary angiography. Methods: This retrospective study included 314 consecutive patients who underwent coronary angiography. ERS was calculated using six ECG parameters: heart rate > 75 bpm, left ventricular hypertrophy, delayed QRS transition zone, frontal QRS-T angle > 90°, prolonged QTc interval, and extended T peak to T end interval. Results: Of the study population (mean age 57.8 ± 11.4, 61.5% male), 158 were diagnosed with CAD, and 156 constituted the control group. The mean ERS was significantly higher in the CAD group than the control group (2.34 ± 1.35 vs. 1.78 ± 1.12, p = 0.006). Among ERS components, delayed QRS transition (p = 0.023), prolonged QTc (p = 0.004), and extended T peak to T end interval (p = 0.001) were notably more prevalent in the CAD group. ERS was independently associated with the presence of CAD on multivariate logistic regression analysis (p < 0.05). Conclusions: ERS is significantly associated with the presence and severity of CAD in stable patients. Elevated ERS, particularly due to delayed QRS transition, prolonged QTc, and extended T peak to T end interval, may serve as a valuable, non-invasive marker for prediction and early identification of CAD. Full article

This study conducts a diagnostic analysis of an extremely heavy rainfall event and its causative factors that occurred in Libya, North Africa on 10 September 2023. The Weather Research and Forecasting (WRF) model was also employed to perform some sensitivity experiments for this heavy rainfall event and further reveal its causes. Results indicate that the primary synoptic system responsible for this extreme precipitation event was an extratropical cyclone (storm) named “Daniel”. During the formation and development of this cyclone, the circulation at the 500 hPa level from the eastern Atlantic to western Asia exhibited a stable “two troughs and one ridge” pattern, with a upper-level cold vortex over the eastern Atlantic, a high-pressure ridge over central Europe, and a cut-off low over western Asia, collectively facilitating the formation and development of this cyclone. As this cyclone moved southward, it absorbed substantial energy from the Mediterranean Sea; following landfall, the intrusion of weak cold air enabled the cyclone to continue intensifying. Meanwhile, the northwest low-level jet stream to the west of the extratropical cyclone moved alongside the cyclone to the coastal regions of northeastern Libya, where it converged with water vapor transport belts originating from the Ionian Sea, the Aegean Sea, and the coastal waters of northeastern Libya. This convergence provided abundant water vapor for the rainstorm event, and under the combined effects of convergence and orographic lifting on the windward slopes of the coastal mountains, extreme precipitation was generated. In addition, the atmosphere over the coastal regions of northeastern Libya exhibited strong stratification instability, which was conducive to the occurrence of extreme heavy precipitation. Although WRF successfully reproduced the precipitation process, the precipitation amount was underestimated. Sensitivity experiments revealed that both the topography in the precipitation area and the sea surface temperature (SST) of the Mediterranean Sea contributed to this extreme heavy precipitation event. Full article

Background/Objectives: The most prevalent prolonged cardiac arrhythmia and a significant global health burden is atrial fibrillation (AF). Although its connection to AF is still unknown, lipoprotein(a) (Lp(a)), a genetically determined lipoprotein with pro-inflammatory and pro-atherogenic characteristics, has been linked to cardiovascular disease. The purpose of this study was to measure and assess the relationship between circulating Lp(a) levels and AF. Methods: In compliance with the PRISMA 2020 guidelines, a systematic review and meta-analysis were carried out using a protocol that was preregistered in PROSPERO (CRD420251153244). Comprehensive searches of PubMed/MEDLINE, Embase, Web of Science, Scopus, the Cochrane Library, and Google Scholar up until September 2025 were used to find observational studies comparing circulating Lp(a) levels in adults with and without AF. Results: Circulating Lp(a) concentrations were significantly higher in AF patients than in controls across 10 studies (pooled MD = 2.81; 95%CI: 1.58–4.05; p < 0.0001). In the subgroup analysis by geographical setting, studies conducted in Asia and studies from Europe/USA exhibited a statistically significant effect. Despite the extreme heterogeneity (I² = 99%), sensitivity analyses verified that the overall effect was stable. Conclusions: Our pooled analysis revealed a statistically significant association between Lp(a) and AF; however, the certainty of the evidence was rated as very low according to the GRADE methodology. To elucidate causality, enhance risk stratification, and investigate whether Lp(a)-lowering tactics could alter AF risk, large, prospective, multi-ethnic studies with standardized biomarker assessment are needed. Full article

“Wind theft”, the extraction of upstream wind resources by neighboring wind farms on account of wind farm or cluster wakes, is receiving wider popular attention. Cluster wakes need to be accounted for in wider planning strategies, for which measurements and wake models can be deployed to aid this process. To contribute to such planning measures, a flight campaign for investigating cluster waking and other phenomena in the North Sea was conducted in 2020 and 2021 to contribute extra flight data obtained during the first flight campaign of 2016 and 2017. We report the latest results of the 2020–2021 flight campaign following the work and methodology of Cañadillas et al. (2020), where, using the 2016–2017 flight measurements, wake lengths extending up to approximately 60 km in stable stratification were inferred, consistent with an explicit stability-dependent analytical model. Analysis of the recent 2020–2021 flight data is approximately consistent with the results of Cañadillas et al. (2020) in stable conditions, albeit with greater scatter. This is because Cañadillas et al. (2020) analyzed only flights in which the wind conditions remained nearly constant during the measurement period, whereas the current dataset includes more variable conditions. Comparisons with the analytical-based engineering model show good first-order agreement with the flight data, but higher-order effects, such as flow non-homogeneity, are not accounted for. The application of these results to the stability information for developing offshore wind energy regions such as the East Coast of the USA and Bass Strait, Australia gives an outline of the expected wake lengths there. Simple engineering models, such as that demonstrated here, though primarily designed for commercial applications, need to be further developed into advanced spatial planning frameworks for offshore wind energy areas. Full article

With increasing global energy demand and depletion of onshore oil–gas resources, deep-sea hydrocarbon exploration and development have become strategically vital. As core subsea transportation equipment, the performance of helico-axial multiphase pumps directly determines the efficiency and economic feasibility of deep-sea extraction. However, non-uniform inflow patterns caused by uneven gas–liquid distribution in pipelines degrade pressure-boosting capability and reduce pump efficiency under actual operating conditions. To address this, an optimization method employing helical static mixers was developed. A mixer with a 180° helical angle was designed and installed upstream of the pump inlet. Numerical simulations demonstrate that the mixer enhances gas-phase distribution uniformity in stratified flow, improving efficiency and head across varying gas void fractions (GVFs). At a stratification height ratio (Ψ) of 0.32, efficiency increased by 15.41% and head rose by 15.64 m, while turbulent kinetic energy (TKE) at the impeller outlet decreased by up to 50%. For slug flow conditions, the mixer effectively suppressed gas volume fraction fluctuations, consistently improving efficiency under different slug flow coefficients (φ) with a maximum head increase of 9.82%. The optimized flow field exhibits uniform gas–liquid velocity distribution, stable pressure boosting, and significantly reduced TKE intensity within impeller passages. Full article

Gravity waves play a crucial role in the evolution of convective systems. The unique thermal structure of elevated convection occurring above a stable boundary layer facilitates the generation and propagation of gravity waves. This study focuses on an elevated convection event over Central China on the night of 2–3 February 2024. The WRF model, combined with terrain sensitivity experiments, is employed to analyze the characteristics of gravity waves and the effects of terrain variability. The event consists of two elevated convection clusters; the first triggers gravity waves on its southwestern side, which subsequently initiates the second convection cluster. The gravity waves exhibit a horizontal wavelength of 25 km and a period of 17.5 min, propagating eastward. Below an altitude of 3 km, a stable wave duct layer is present, overlain by an unstable overreflective zone. This stratification creates an ideal channel for ducted gravity wave propagation, which is essential for maintaining the waves. Sensitivity experiments confirm that convective forcing alone is sufficient to generate the observed gravity waves. Although the terrain lies within the stable boundary layer, its ruggedness modulates the distribution of waves and indirectly influences the organization of elevated convection. Full article

Background/Objectives: Multiple sclerosis (MS) is a chronic demyelinating disease where early identification of patients at risk of conversion from clinically isolated syndrome (CIS) to clinically definite MS remains a critical unmet clinical need. Existing machine learning approaches often lack interpretability, limiting clinical trust and adoption. The objective of this research was to develop a novel two-stage machine learning framework with comprehensive explainability to predict CIS-to-MS conversion while addressing demographic bias and interpretability limitations. Methods: A cohort of 177 CIS patients from the National Institute of Neurology and Neurosurgery in Mexico City was analyzed using SeruNet-MS, a two-stage framework that separates demographic baseline risk from clinical risk modification. Stage 1 applied logistic regression to demographic features, while Stage 2 incorporated 25 clinical and symptom features, including MRI lesions, cerebrospinal fluid biomarkers, electrophysiological tests, and symptom characteristics. Patient-level interpretability was achieved through SHAP (SHapley Additive exPlanations) analysis, providing transparent attribution of each factor’s contribution to risk assessment. Results: The two-stage model achieved a ROC-AUC of 0.909, accuracy of 0.806, precision of 0.842, and recall of 0.800, outperforming baseline machine learning methods. Cross-validation confirmed stable performance (0.838 ± 0.095 AUC) with appropriate generalization. SHAP analysis identified periventricular lesions, oligoclonal bands, and symptom complexity as the strongest predictors, with clinical examples illustrating transparent patient-specific risk communication. Conclusions: The two-stage approach effectively mitigates demographic bias by separating non-modifiable factors from actionable clinical findings. SHAP explanations provide clinicians with clear, individualized insights into prediction drivers, enhancing trust and supporting decision making. This framework demonstrates that high predictive performance can be achieved without sacrificing interpretability, representing a significant step forward for explainable AI in MS risk stratification and real-world clinical adoption. Full article