Search Results (582)

A learning management system (LMS) plays a crucial role in supporting students’ educational activities by centralized platforms for course delivery, communication, and student support. Recently, many universities have integrated chatbots into their LMS to assist students with various inquiries and tasks. However, existing chatbots often necessitate human interventions to manually respond to complex queries, resulting in limited scalability and efficiency. In this paper, we present a memory-augmented large language model (LLM) framework that enhances the reasoning and contextual continuity of LMS-based chatbots. The proposed framework first embeds user queries and retrieves semantically relevant entries from various LMS resources, including instructional documents and academic frequently asked questions. Retrieved entries are then filtered through a two-stage confidence filtering process that combines similarity thresholds and LLM-based semantic validation. Validated information, along with user queries, is processed by LLM for response generation. To maintain coherence in multi-turn interactions, the chatbot incorporates short-term, long-term, and temporal event memories, which track conversational flow and personalize responses based on user-specific information, such as recent activity history and individual preferences. To evaluate response quality, we employed a multi-layered evaluation strategy combining BERTScore-based quantitative measurement, an LLM-as-a-Judge approach for automated semantic assessment, and a user study under multi-turn scenarios. The evaluation results consistently confirm that the proposed framework improves the consistency, clarity, and usefulness of the responses. These findings highlight the potential of memory-augmented LLMs for scalable and intelligent learning support within university environments. Full article

Soil liquefaction poses a significant risk to foundations during earthquakes, prompting the need for simple, risk-aware screening tools that go beyond single deterministic boundaries. This study creates a probability-calibrated dual-threshold screening rule using a random forest (RF) classifier trained on 208 SPT case histories with quality-based weights (A/B/C = 1.0/0.70/0.40). The model is optimized with random search and calibrated through isotonic regression. Iso-probability contours from 1000 bootstrap samples produce paired thresholds for fines-corrected, overburden-normalized blow count N_1,60,CS and normalized cyclic stress ratio CSR_7.5,1 at target liquefaction probabilities P_liq = 5%, 20%, 50%, 80%, and 95%, with 90% confidence intervals. On an independent test set (n = 42), the calibrated model achieves AUC = 0.95, F₁ = 0.92, and a better Brier score than the uncalibrated RF. The screening rule classifies a site as susceptible when N_1,60,CS is at or below and CSR_7.5,1 is at or above the probability-specific thresholds. Designed for level ground, free field, and clean-to-silty sand sites, this tool maintains the familiarity of SPT-based charts while making risk assessment transparent and auditable for different facility importance levels. Sensitivity tests show its robustness to reasonable rescaling of quality weights. The framework offers transparent thresholds with uncertainty bands for routine preliminary assessments and to guide the need for more detailed, site-specific analyses. Full article

Necrotizing fasciitis (NF) is a rare but aggressive soft tissue infection with high rates of mortality and amputation, making early identification of key prognostic biomarkers essential for clinical management. However, the rarity and heterogeneity of NF mean clinical datasets are often small and non-normally distributed, limiting the effectiveness of standard parametric statistical approaches. To address this, we retrospectively analyzed 66 NF patients using a robust, distribution-free framework that combines the Nonparametric Combination (NPC) methodology and bootstrap resampling. We specifically assessed glycated hemoglobin (HBA1C) and serum albumin (ALBUMINA) as potential predictors of two outcomes: mortality (MORTO) and major amputation (AMPUTAZIONE). NPC enabled exact multivariate hypothesis testing while rigorously controlling the family-wise error rate (FWER), and bootstrap resampling generated 95% confidence intervals (CI) for critical biomarkers. HBA1C was an exceptionally significant predictor compared to the 7.0% clinical threshold (p = 1.04 × 10⁻¹⁵⁴, CI: 0.0830–0.0957), while ALBUMINA showed greater biological variability but no significant association with outcomes (2.8 g/dL; p = 0.267, CI: 2.551–2.866). We also developed a global severity ranking, integrating multiple variables to improve clinical risk stratification. Our results demonstrate that permutation-based and resampling methods provide reliable, actionable insights from challenging small-sample clinical datasets. Based on a small-sample dataset from necrotizing fasciitis patients, this framework provides a replicable model for robust, nonparametric statistical analysis in similarly rare and high-risk medical conditions. This study introduces a Nonparametric Combination (NPC) framework for risk scoring in necrotizing fasciitis using bootstrap resampling and permutation tests. Key predictors like HBA1C and Albumin were assessed, achieving an AUC of 0.89 and a Youden Index of 0.71. The model offers a robust, interpretable tool for clinical risk stratification in small-sample rare disease settings. Full article

Objectives: To investigate the correlation between intraoperative conventional ultrasound, SWE, and SMI ultrasound manifestations of glioma and the expression of immunohistochemical markers. Methods: Patients with single superficial supratentorial glioma scheduled for brain tumor resection in our neurosurgery department from October 2020 to October 2022 were prospectively included. High-grade glioma (HGG) and low-grade glioma (LGG) were classified by pathological histological grading, and the differences in conventional ultrasound, SWE Young’s modulus, and SMI intratumoral and peritumoral blood flow architecture between HGG and LGG were analyzed, and the SWE diagnostic cut-off value was calculated by the Youdon index. Logistic regression models were used to analyze the independent predictive ultrasound signs associated with the diagnosis of HGG. HGG and LGG were classified by pathological histological grading. IDH1 expression was measured by immunohistochemical methods to analyze the correlation between IDH1 expression in glioma and clinical and ultrasound characteristics. Results: Forty-eight patients with glioma admitted to our hospital from October 2020 to October 2022 were included in this study, including 30 (62.5%) with HGG and 18 (37.5%) with LGG. For conventional ultrasound, HGG was often associated with severe peritumoral edema compared with LGG (p = 0.048). The sensitivity of HGG was 88.9%, the specificity was 86.7%, and the AUC was 0.855 (95% confidence interval: 0.741–0.968, p = 0.001) using Young’s mode 13.90 kPa as the threshold. Logistic analysis showed that SWE Young’s modulus values, and peritumoral and intratumoral SMI blood flow structures, were associated with the diagnosis of HGG. Among the 48 gliomas, 22 (45.8%) were IDH1-positive and 26 (54.2%) were IDH1-negative, with no statistical difference in age between the two groups and a statistical difference in histological grading (p < 0.05). There was a statistical difference between IDH1 mutant and wild type in terms of peritumoral edema and SMI intratumoral and peritumoral tissue vascular architecture. Logistic regression models showed that intratumoral and peritumoral tissue SMI vascular architecture was a valid predictor of IDH1 positivity, with a classification accuracy of 81.3%, sensitivity of 90.9%, and specificity of 73.1%. Further group analysis of mutant Young’s modulus values in LGG were higher than wild-type Young’s modulus values (p = 0.031). Conclusions: Peritumoral and intratumoral tissue SMI vascular architecture was a valid predictor of IDH1 positivity. Based on intraoperative ultrasound multimodality images, we can preoperatively determine the expression of molecular markers of lesions, which is of clinical significance for optimizing surgical strategies and predicting prognosis. Full article

Background: Left ventricular myocardial work (MW) derived from non-invasive pressure–strain loops has emerged as a load-adjusted index of contractile performance. Its value for risk stratification in advanced heart failure (HF) remains uncertain. Methods: We retrospectively studied 151 consecutive patients with advanced HF undergoing comprehensive evaluation at our tertiary centre between January 2016 and December 2022. MW parameters—left ventricular global work index (LVGWI), global constructive work (LVGCW), global wasted work (LVGWW) and global work efficiency (LVGWE)—were derived from speckle-tracking echocardiography integrated with brachial blood pressure. Cardiopulmonary exercise testing (CPET), right heart catheterisation (RHC) and biochemical markers were obtained. Patients were stratified according to an LVGWI threshold of 600 mmHg%, identified by receiver operating characteristic (ROC) analysis for predicting the combined end point of cardiovascular mortality or HF hospitalisation. Correlations between MW and traditional indices were assessed, and event-free survival was analysed by Kaplan–Meier curves. Results: LVGWI correlated modestly with pVO₂ (r = 0.35, p = 0.01) and left ventricular ejection fraction (r = 0.42, p < 0.001) and inversely with NT-proBNP (r = −0.30, p = 0.03). LVGWI displayed the largest area under the curve (AUC 0.76 [95% confidence interval 0.65–0.85]) for predicting the combined end point compared with pVO₂ (AUC 0.73) and LVEF (AUC 0.67). Dichotomisation by LVGWI ≤ 600 mmHg% identified a high-risk group (Group A) with worse NYHA class, lower systolic blood pressure and reduced exercise capacity. After a median follow-up of 24 months, Group A exhibited significantly lower event-free survival (log-rank p = 0.02). Multivariable analysis was not performed owing to the limited sample size; therefore, findings should be interpreted with caution. Conclusions: In patients with advanced HF, left ventricular myocardial work, particularly LVGWI, provides incremental prognostic information beyond conventional markers. An LVGWI cut-off of 600 mmHg% derived from ROC analysis identified patients at increased risk of cardiovascular events and may inform timely referral for mechanical circulatory support or transplantation. Larger prospective studies are warranted to confirm these observations and to establish standardised thresholds across vendors. Full article

Background: Postoperative wound complications following total joint arthroplasty (TJA) significantly impact patient outcomes and healthcare costs. Reliable preoperative biomarkers for identifying patients at increased risk are critical for optimizing patient management and reducing complication rates. This study evaluated the predictive utility of the C-reactive protein to albumin ratio (CAR) and the prognostic nutritional index (PNI) for periprosthetic joint infection (PJI) and postoperative wound complications in patients undergoing total hip arthroplasty (THA) and total knee arthroplasty (TKA). Methods: We retrospectively studied patients who underwent primary THA and TKA in our department from March 2019 to April 2024. The study included a total of 842 patients (568 knees and 274 hips). Preoperative blood samples were assessed for serum CRP, albumin, and total lymphocyte count, facilitating the calculation of CAR and PNI values. Patient outcomes were monitored, identifying PJI and aseptic wound complications such as persistent wound drainage, hematoma, seroma, skin erosion, and wound dehiscence within 2 weeks post-surgery. Results: The average follow-up time for patients was 39.2 months (range 13–73 months). PJI was significantly linked with elevated admission CAR and diminished PNI ratio (p < 0.001 and p < 0.001). ROC analysis demonstrated optimal predictive cut-off values for CAR at 3.1 (Area under curve [AUC]: 0.92, specificity 97.4%, sensitivity 92.3%) and PNI at 49.4 (AUC: 0.93, specificity 94.7%, sensitivity 91.7%). Furthermore, both CAR (Odds ratio [OR]: 3.84, 95% confidence interval [CI]: 1.6–9.1, p = 0.002) and PNI (OR: 21.8, 95% CI: 9–48.6, p < 0.001) were identified as two independent risk factors associated with the development of PJI following THA or TKA. Further subgroup analysis revealed distinct predictive thresholds for CAR and PNI according to surgical procedure type (TKA and THA), enhancing diagnostic accuracy. Conclusions: Preoperative admission elevated CAR and decreased PNI effectively predict PJI and postoperative wound complications in THA and TKA, supporting their utility as simple, cost-effective biomarkers in clinical practice. Incorporating CAR and PNI evaluations into preoperative assessments can enhance patient stratification and preventive strategies, thus mitigating risks and improving surgical outcomes. Full article

Shiga toxin-producing Escherichia coli (STEC) are genetically diverse foodborne pathogens of major global public health concerns. Serogroup-level identification is critical for effective surveillance and outbreak control; however, it is often challenged by STEC’s genome plasticity and frequent recombination. In this study, we employed a standardized pangenomic pipeline integrating Roary ILP Bacterial Core Annotation Pipeline (RIBAP) and Panaroo to analyze 160 complete, high-quality STEC genomes representing eight major serogroups at a 95% sequence identity threshold. Candidate serogroup-specific markers were identified using gene presence/absence profiles from RIBAP and Panaroo. Our analysis revealed several high-confidence markers, including metabolic genes (dgcE, fcl_2, dmsA, hisC) and surface polysaccharide-related genes (capD, rfbX, wzzB). Comparative pangenomic evaluation showed that RIBAP predicted a larger pangenome size than Panaroo. Additionally, some genomes from the O104:H1, O145:H28, and O45:H2 serotypes clustered outside their expected clades, indicating sporadic serotype misplacements in phylogenetic reconstructions. Functional annotation suggested that most candidate markers are involved in critical processes such as glucose metabolism, lipopolysaccharide biosynthesis, and cell surface assembly. Notably, approximately 22.9% of the identified proteins were annotated as hypothetical. Overall, this study highlights the utility of pangenomic analysis for potential identification of clinically relevant STEC serogroups markers and phylogenetic interpretation. We also note that pangenome analysis could guide the development of more accurate diagnostic and surveillance tools. Full article

Background: Artificial Intelligence (AI)-driven Decision Support Systems (DSSs) promise improvements in diagnostic accuracy and clinical workflow efficiency, but their adoption is hindered by inadequate confidence calibration, limited transparency, and poor alignment with real-world decision processes, which limit clinician trust and lead to high override rates. Methods: We developed and validated a dynamic scoring framework to enhance trust in AI-generated diagnoses by integrating AI confidence scores, semantic similarity measures, and transparency weighting into the override decision process using 6689 cardiovascular cases from the MIMIC-III dataset. Override thresholds were calibrated and validated across varying transparency and confidence levels, with override rate as the primary acceptance measure. Results: The implementation of this framework reduced the override rate to 33.29%, with high-confidence predictions (90–99%) overridden at a rate of only 1.7%, and low-confidence predictions (70–79%) at a rate of 99.3%. Minimal transparency diagnoses had a 73.9% override rate compared to 49.3% for moderate transparency. Statistical analyses confirmed significant associations between confidence, transparency, and override rates (p < 0.001). Conclusions: These findings suggest that enhanced transparency and confidence calibration can substantially reduce override rates and promote clinician acceptance of AI diagnostics. Future work should focus on clinical validation to optimize patient safety, diagnostic accuracy, and efficiency. Full article

Background: This study aimed to determine whether mental health status contributes to speed eating in adolescents and young adults. Methods: This study enrolled 106 subjects (53 males and 53 females), ranging in age from 12 to 24 years. After a self-administered lifestyle questionnaire and the 12-item General Health Questionnaire (GHQ-12) were administered, a swallowing threshold test was performed. The swallowing threshold was determined based on the concentration of dissolved glucose obtained from the gummy jellies. Low swallowing threshold was characterized by glucose levels falling within the bottom 20th percentile. GHQ-12 was categorized into poor (score 4–12) and normal (score 0–3). Following the univariate analysis, a multivariate binary logistic regression analysis was conducted to determine the factors linked to a low swallowing threshold. Results: Binomial logistic regression analysis revealed that the factors associated with a low swallowing threshold included poor mental health (odds ratio [OR] = 8.47, p = 0.007, confidence interval [CI] = 2.437–32.934) and no physical activity (OR = 5.604, p = 0.008, CI = 1.562–22.675). Conclusions: Speed eating is closely associated with risk behaviors for poor mental health in adolescents and young adults. Full article

Background: Chronic obstructive pulmonary disease (COPD) is a progressive condition leading to declining lung function, dyspnea, and reduced quality of life. Pulmonary rehabilitation (PR) remains a cornerstone in COPD management; however, access remains limited, with less than 3% of eligible patients participating. Inspiratory muscle training (IMT), especially through novel methods like the Test of Incremental Respiratory Endurance (TIRE), offers a potential home-based alternative to traditional rehabilitation services. Despite growing interest, a key knowledge gap persists: few randomized trials have directly compared TIRE with threshold loading IMT over extended, largely unsupervised home-based periods while concurrently evaluating inspiratory muscle endurance and adherence. This randomized controlled trial aimed to evaluate the long-term efficacy of TIRE IMT compared to traditional threshold IMT and sham training in COPD patients. The study also assessed adherence to these home-based interventions, focusing on unsupervised periods without additional motivational support. Methods: A total of 52 COPD patients were randomly assigned to one of three groups: TIRE IMT, Threshold IMT, or Sham IMT. The study consisted of an 8-week supervised Phase I followed by a 24-week unsupervised Phase II. Training details: TIRE—session template set to 50% of the day’s maximal sustained effort; 6 levels × 6 inspirations (total 36) with preset inter-breath recoveries decreasing from 60 s to 10 s. Threshold IMT—spring-loaded valve set to 50% MIP (re-set at week 4); 36 inspirations completed within ≤30 min. Sham—valve set to minimal resistance (9 cmH₂O); 36 inspirations within ≤30 min. Primary outcomes included changes in maximal inspiratory pressure (MIP) and sustained maximal inspiratory pressure. Secondary outcomes focused on adherence rates and correlations with functional capacity. Results: Of the 52 participants, 36 completed the study. Participant details: TIRE n = 12 (mean age 60.9 ± 12.9 years), Threshold n = 12 (67.4 ± 6.9 years), Sham n = 12 (67.3 ± 8.7 years); overall 21/36 (58%) men; mean BMI 30.0 ± 7.5 kg/m². The TIRE IMT group demonstrated significantly greater improvements in MIP (31.7%) and SMIP compared to both the Threshold and Sham groups at 24 weeks (p < 0.05). Despite a decline in adherence during the unsupervised phase, the TIRE group maintained superior outcomes. No adverse events were reported during the intervention period. Conclusions: In this randomized trial, TIRE IMT was associated with greater improvements in inspiratory muscle performance than threshold and sham IMT. While adherence was higher in the TIRE group, it declined during the unsupervised phase. The clinical interpretation of these findings should consider the relatively wide confidence intervals and modest sample size. Nevertheless, the mean change in MIP in the TIRE arm exceeded a recently proposed minimal important difference for COPD, suggesting potential clinical relevance; however, no universally accepted minimal important difference exists yet for SMIP. Further adequately powered trials are warranted. Full article

Background: Melioidosis, caused by Burkholderia pseudomallei, is a serious infectious disease in Southeast Asia and northern Australia. Methods: We systematically reviewed observational studies measuring C-reactive protein (CRP) in laboratory-confirmed melioidosis for diagnosis, severity assessment, or outcome evaluation. PubMed, Embase, and Scopus were searched up to May 2025. Data were pooled using a random-effects model; heterogeneity was quantified (I²). Results: Seven studies (n = 451) were included. The pooled mean CRP level in melioidosis was 74.37 mg/L (95% Confidence Interval [CI], 32.76–168.83; I² = 99.1%), considerably higher than healthy reference values (<10 mg/L). Conclusions: CRP is consistently raised in melioidosis and may aid in early diagnosis and severity monitoring, although high heterogeneity limits the precision of pooled estimates. Integration of CRP into multimodal prediction tools, rather than use in isolation, is recommended. Further prospective studies should define optimal diagnostic thresholds. Full article

This study aims to reconstruct NDI45 missing values due to cloud cover while outlining the importance of vegetation health for the climate–carbon cycle and the benefits of the NDI45 index for high canopy area indices. The methods include a novel hybrid framework that combines a deterministic Kalman filter (KF) and a clustering-based LSTM network to generate gap-free NDI45 series with 20 m spatial and 5-day temporal resolution. The innovation of the applied method relies on achieving a single-sensor workflow, provides a pixel-level uncertainty map, and minimizes LSTM overfitting through clustering based on a correlation threshold. In the northern Pampas (South America), this hybrid approach reduces the MAE by 22–35% on average and narrows the 95% confidence interval by 25–40% compared to the Kalman filter or LSTM alone. The three-dimensional spatio-temporal analysis demonstrates that the KF–LSTM hybrid provides better spatial homogeneity and reliability across the entire study area. The proposed framework can generate gap-free, high-resolution NDI45 time series with quantified uncertainties, enabling more reliable detection of vegetation stress, yield fluctuations, and long-term resilience trends. These capabilities make the method directly applicable to operational drought monitoring, crop insurance modeling, and climate risk assessment in agricultural systems, particularly in regions prone to frequent cloud cover. The framework can be further extended by including radar backscatter and multi-model ensembles, thus providing a promising basis for the reconstruction of global, high-resolution vegetation time series. Full article

Background: Robotic-assisted unicompartmental knee arthroplasty (UKA) enhances the precision of component alignment compared to conventional techniques. Although various robotic systems exist, direct comparisons assessing their relative clinical performance remain limited. The purpose of this study is to provide a comparison between image-based and imageless robotic UKA. Methods: A systematic review was conducted in accordance with PRISMA guidelines. Five databases were searched: PubMed (via MEDLINE), Epistemonikos, Cochrane Library, Web of Science, and Scopus. Inclusion criteria were (1) studies comparing rUKA and cUKA with radiologic parameters and revision rates (prospective or retrospective), (2) human subjects, (3) meta-analyses for cross-referencing, and (4) English language. Data collected included (1) pre- and postoperative radiologic parameters, (2) radiologic outliers, and (3) revisions and their causes. A random-effects meta-analysis was employed to enable a generalizable comparison. Mean differences (MDs) with 95% confidence intervals (CIs) were calculated for continuous variables, and log odds ratios (LORs) with 95% CIs for binary outcomes. Results: Image-based robotic UKA was associated with fewer joint line height outliers (LOR = 3.5, 95% CI: 0.69–6.30, p = 0.015) using a 2° threshold. HKA outliers (thresholds 2–3°) were also reduced (LOR = 0.6, 95% CI: 0.09–1.19, p = 0.024). Posterior tibial and posterior femoral implant fit were significantly lower with image-based systems (LOR = 1.7, 95% CI: 1.37–2.03, respectively, LOR = 1.7, 95% CI: 1.29–1.91; p < 0.001 for both). No significant differences in revision rates were observed. Conclusions: Image-based robotic systems may result in fewer outliers in key radiologic parameters, including hip–knee angle, joint-line height, posterior tibial, and posterior femoral fit, though reporting remains highly heterogeneous. Full article

With the rapid expansion of multi-platform remote sensing applications, cloud contamination significantly impedes cross-platform data utilization. Current cloud detection methods face critical technical challenges in cross-platform settings, including neglect of atmospheric radiative transfer mechanisms, inadequate multi-scale structural decoupling, high intra-class variability coupled with inter-class similarity, cloud boundary ambiguity, cross-modal feature inconsistency, and noise propagation in pseudo-labels within semi-supervised frameworks. To address these issues, we introduce a Physics-Guided Multi-Representation Network (PGMRN) that adopts a student–teacher architecture and fuses tri-modal representations—Pseudo-NDVI, structural, and textural features—via atmospheric priors and intrinsic image decomposition. Specifically, PGMRN first incorporates an InfoNCE contrastive loss to enhance intra-class compactness and inter-class discrimination while preserving physical consistency; subsequently, a boundary-aware regional adaptive weighted cross-entropy loss integrates PA-CAM confidence with distance transforms to refine edge accuracy; furthermore, an Uncertainty-Aware Quadruple Consistency Propagation (UAQCP) enforces alignment across structural, textural, RGB, and physical modalities; and finally, a dynamic confidence-screening mechanism that couples PA-CAM with information entropy and percentile-based thresholding robustly refines pseudo-labels. Extensive experiments on four benchmark datasets demonstrate that PGMRN achieves state-of-the-art performance, with Mean IoU values of 70.8% on TCDD, 79.0% on HRC_WHU, and 83.8% on SWIMSEG, outperforming existing methods. Full article

Background and Objectives: Postoperative physical recovery, particularly the acquisition of independent ambulation, is a critical milestone in rehabilitation following living donor liver transplantation (LDLT). Although liver function markers are conventionally used to assess hepatic physiology, emerging evidence has suggested their potential role as prognostic indicators of physical performance. Materials and Methods: This study investigated the association between liver function parameters at the initiation of postoperative physical therapy (total bilirubin [T-Bil], aspartate aminotransferase [AST], and alanine aminotransferase [ALT]) and the independent walking ability of 63 patients who underwent LDLT. A logistic regression model was constructed using these variables, and a receiver-operating characteristic (ROC) curve analysis was performed to evaluate its discriminative performance. Predicted probabilities of each patient were calculated, and the optimal cutoff value was determined based on the Youden Index. Results: The multivariate logistic regression model demonstrated a statistically significant association between liver function markers and the ambulation status of a cohort of 63 patients. The ROC curve analysis yielded an area under the ROC curve (AUC) of 0.8416 (95% confidence interval [CI]: 0.715–0.968), indicating strong predictive performance. The optimal cutoff value was 0.865, with sensitivity and specificity of 74.1% and 88.9%, respectively. The bootstrap CI for sensitivity at this threshold ranged from 0.6111 to 0.8519. The Hosmer–Lemeshow test indicated good model fit (p = 0.363), and the correct classification rate was 87.3%. Conclusions: Liver function test results may be indicators of hepatic dysfunction as well as functional biomarkers that could predict ambulatory outcomes following LDLT. This predictive model may enhance early clinical decision-making regarding rehabilitation and discharge planning. Future prospective studies should be performed to validate the generalizability of these results to broader clinical contexts. Full article