Search Results (11,249)

In sensor-centric fields like healthcare, environmental monitoring, and industry, image classification is key to turning visual sensor data into actionable insights. Sensor-generated dynamic streaming data poses significant challenges for traditional static image classification models due to the continuous emergence of new categories, distribution shifts, and limited edge storage. With sensor streaming data facing label scarcity and high annotation costs, semi-supervised continual learning is essential, leveraging unlabeled data for incremental learning and reducing reliance on costly annotations. However, current semi-supervised continual learning methods rely on labeled data to generate pseudo-labels, leading to confirmation and relational biases. To mitigate these dual biases, we propose a Bias Calibration method based on nearest-neighbor semi-supervised continual learning, which integrates and adapts Confidence-Enhanced Learning (originally introduced for static datasets) and Guided Contrastive Learning. Specifically, the Confidence-Enhanced Learning aims to reduce competition among similar classes and penalizes low-confidence predictions, thereby generating high-confidence pseudo-labels for unlabeled data and mitigating confirmation bias. Guided Contrastive Learning constructs a pseudo-label graph and a feature representation graph, using the pseudo-label graph to optimize the feature representation graph, thereby improving class discrimination and reducing feature bias. Experiments on CIFAR-10, CIFAR-100, and ImageNet-100 show that our method significantly outperforms existing approaches, enhancing classification performance with partial labeling. Full article

Background: Pregnancy represents a major identity transition, yet most perinatal assessments focus primarily on emotional symptoms rather than on how women integrate the maternal role into their broader identity and life context. Difficulties in maternal role integration may constitute an early vulnerability factor for psychological distress. This study aimed to develop and validate the Maternal Role Integration Questionnaire—pregnancy version (MRIQ-P), a brief instrument designed to assess maternal identity and role balance during pregnancy, and to examine its clinical relevance for perinatal mental health. Methods: A sequential mixed-methods design was employed. Phase 1 involved focus groups with pregnant women (n = 17) and cognitive debriefing to generate and refine items. Phase 2 included expert evaluation of content validity. In Phase 3, the MRIQ-P was psychometrically validated in a sample of pregnant women (n = 256), randomly divided into exploratory (n = 83) and confirmatory (n = 173) subsamples. Exploratory and confirmatory factor analyses were conducted, along with reliability analyses, tests of convergent, discriminant, incremental, and measurement invariance validity. Results: Analyses supported a bifactor structure comprising a general factor of maternal role integration and two specific dimensions: Maternal Identity and Balance of the Maternal Role. The final 8-item version demonstrated excellent internal consistency for the total score (α = 0.96) and subscales (α = 0.98 for Maternal Identity and α = 0.98 for Balance of the Maternal Role), as well as measurement invariance across primiparous and multiparous women. Higher maternal role integration was associated with greater self-esteem, positive affect, and life satisfaction, and with lower anxiety, depression, prenatal distress, and maternal ambivalence. Importantly, MRIQ dimensions explained additional variance in antenatal depression and dispositional guilt beyond established psychological predictors, supporting its incremental and potential clinical utility. Conclusions: The MRIQ is a brief, psychometrically robust, and clinically relevant instrument for assessing maternal role integration during pregnancy. By capturing identity- and role-related processes that are not directly addressed by symptom-based screening tools, it may contribute to early identification of vulnerability and to more comprehensive perinatal psychological assessment in healthcare settings. Full article

Objective: This study aimed to develop a nomogram prediction model for predicting 14-day in-hospital mortality in patients with acute myocardial infarction (AMI) and ventricular septal rupture (VSR). Methods: Clinical data of 86 hospitalized patients (44 survivors and 42 non-survivors within 14 days) were retrospectively collected in Nanjing First Hospital from 1 March 2015 to 7 August 2025. Lasso regression and multivariable logistic regression were used to identify predictors, which were subsequently incorporated into the nomogram development. The model performance was assessed using area under the receiver operating characteristic curve (AUC), calibration plots, decision curve analysis (DCA), and clinical impact curves, with internal validation via 1000 bootstrap resamples. Results: Analysis of lasso regression and multivariable logistic regression analysis identified WBC count (OR = 1.31, 95% CI: 1.01–1.28, p = 0.040), D-dimer level (OR = 1.18, 95% CI: 1.01–1.38, p = 0.043), early revascularization (OR = 0.22, 95% CI: 0.06–0.88, p = 0.032), ventilatory support (OR = 3.48, 95% CI: 1.07–11.29, p = 0.038), and infection (OR = 3.97, 95% CI: 1.02–15.42, p = 0.047) as independent predictors of 14-day mortality for patients. Based on the results, a prediction nomogram model was constructed. The model achieved an area under the receiver operating characteristic curve (AUC) of 0.866 (95% CI: 0.785–0.946), with sensitivity of 0.857 (95% CI: 0.751–0.963) and specificity of 0.818 (95% CI: 0.704–0.932). Calibration plots demonstrated acceptable agreement between predicted and observed probabilities; decision curve analysis (DCA) and clinical impact curve further confirmed its net benefit and clinical utility. By 1000 bootstrap resampling iterations, the model demonstrated an apparent AUC of 0.864, 95% CI: 0.776–0.938, confirming reasonable discriminative performance. Conclusions: In summary, this study developed a clinical interpretable nomogram to estimate short-term (14-day) in-hospital mortality risk in patients with AMI-VSR; it provides a robust and interpretable tool for predicting short-term in-hospital mortality. Full article

Objective biomarkers for neurodevelopmental disorders remain an unmet clinical need. The electroretinogram (ERG), a non-invasive recording of the retinal response to light, has shown promise as a physiological marker for autism spectrum disorder (ASD) and attention deficit/hyperactivity disorder (ADHD), yet existing classification approaches based on time-domain and time–frequency features achieve limited accuracy in clinically relevant multi-group scenarios. This study introduces ERG-Graph, a novel graph signal processing (GSP) framework that transforms each ERG waveform into a weighted, undirected graph through amplitude quantization and temporal-adjacency connectivity. Nine topological and spectral features, including total load centrality, clique number, algebraic connectivity, and clustering coefficient, were extracted from each graph to characterize the structural dynamics of the signal. Using light-adapted ERG recordings from 278 participants (ASD = 77, ADHD = 43, ASD + ADHD = 21, Control = 137), we evaluated these features across binary, three-group, and four-group classification scenarios using seven machine learning classifiers with 10-fold subject-wise cross-validation. The proposed ERG-Graph features achieved balanced accuracies of 0.91 (ASD vs. control, males) and 0.88 (ADHD vs. control, females). Critically, fusing ERG-Graph with time-domain features yielded a balanced accuracy of 0.81 for three-group classification (ASD vs. ADHD vs. control), representing an 11-percentage-point improvement over the previous benchmark of 0.70. Statistical analysis confirmed significant topological differences between groups (Kruskal–Wallis, p < 0.001; Cliff’s delta: large effect sizes), and SHAP analysis revealed that graph-theoretic features dominated the top-ranked predictors. These results demonstrate that graph-based topological features capture discriminative information in the ERG waveform that is inaccessible to conventional signal analysis methods, advancing the development of objective biomarkers for neurodevelopmental disorder screening. Full article

Background: Growth differentiation factor-15 (GDF-15) is a stress-responsive cytokine associated with inflammation, metabolic dysfunction, and cardiovascular disease. Its role as a biomarker of microvascular complications in type 2 diabetes (T2D) remains incompletely defined. Objective: To evaluate circulating GDF-15 levels and their association with microvascular complications in patients with T2D. Methods: This cross-sectional study included 160 participants divided into three groups: T2D (n = 93), obesity without carbohydrate disorders (n = 36), and healthy controls (n = 31). Microvascular complications (neuropathy, nephropathy, retinopathy) were assessed. Multivariable logistic regression and receiver operating characteristic (ROC) analysis were performed. Results: GDF-15 levels were significantly higher in T2D compared with non-diabetic individuals (267.5 ± 168.9 vs. 118.3 ± 55.5 pg/mL, p < 0.001). Higher GDF-15 was associated with neuropathy (odds ratio (OR) 1.985; 95% confidence interval (CI) 1.431–2.753) and nephropathy (OR 1.673; 95% CI 1.243–2.254) in unadjusted models. After adjustment, only nephropathy remained independently associated (OR 1.405; 95% CI 1.026–1.923). ROC analysis showed moderate discriminative ability for nephropathy (area under the curve (AUC) = 0.763). Conclusions: Circulating GDF-15 levels are significantly elevated in patients with T2D and are associated with microvascular complications, with the strongest independent association observed for diabetic nephropathy. These findings suggest that GDF-15 may represent a promising biomarker reflecting metabolic stress and risk of diabetic complications. Full article

Background/Objectives: This study examined whether dynamic systemic inflammation- and nutrition-based scores measured at baseline (T0) and during follow-up (T1: days 7–10) are associated with treatment response and surgical requirement in first-episode primary spontaneous pneumothorax (PSP). Methods: A total of 216 consecutive patients with first-episode PSP, treated between January 2020 and December 2024, were retrospectively analyzed. All patients initially underwent tube thoracostomy. During follow-up, 117 patients recovered with drainage therapy, whereas 99 required VATS because of a prolonged air leak. The CAR, SIII, SIRI, PIII, NLR, PLR, and PNI, measured at T0 and T1, were analyzed. Δ-values (T1–T0 differences) were evaluated, and diagnostic performance was assessed using ROC curve analysis. Results: At T0, inflammation- and nutrition-based indices did not differ significantly between groups. In contrast, at T1, CAR, SIII, SIRI, PIII, NLR, and PLR values were significantly higher in the VATS group than in the drainage group (all p < 0.05). Over time, inflammatory indices increased markedly in the VATS group, whereas changes in the drainage group remained limited. PNI decreased significantly at T1 in both groups. ROC analysis demonstrated that CAR, SIII, and NLR showed moderate discriminative performance for identifying patients who required VATS (area under the curve ≈ 0.65). Conclusions: Dynamic assessment of systemic inflammation-based biomarkers provides clinically relevant insight for surgical risk stratification in first-episode PSP. While baseline measurements alone are insufficient, follow-up values and temporal changes—particularly in CAR, SIII, and NLR—may reflect progression toward a surgical phenotype and could serve as adjunctive, non-directive decision-support indicators in PSP management. Full article

In the context of highly interconnected global markets and the rapid dissemination of multilingual information, traditional risk prediction methods that rely on single numerical sequences or monolingual text are insufficient for achieving early perception of cross-market risks. To address this issue, a cross-market risk early warning framework based on multilingual large language models and multimodal sensing fusion is proposed. The proposed approach is centered on a unified risk semantic space, where cross-lingual semantic alignment is employed to reduce semantic discrepancies across languages. Furthermore, a semantic–volatility coupling attention mechanism is introduced to capture the dynamic relationship between textual semantic evolution and market fluctuations. In addition, cross-market knowledge transfer and low-resource enhancement strategies are incorporated to improve the model’s generalization capability across multilingual and multi-market environments, thereby establishing an intelligent perception and early warning system for complex sensing scenarios. Experimental results demonstrate that the proposed method significantly outperforms multiple baseline models in multilingual cross-market risk prediction tasks. In the main experiment, the model achieves a root mean squared error (RMSE) of $0.1127$, an mean absolute error (MAE) of $0.0846$, and an area under the curve (AUC) of $0.8879$, while the early warning gain is improved to $5.2$ days, which is substantially better than the Transformer model (RMSE $0.1365$, AUC $0.8042$) and the multilingual BERT-based fusion model (AUC $0.8395$). In terms of classification performance, higher accuracy, precision, and recall are consistently achieved, with overall accuracy exceeding $0.88$, and both precision and recall are maintained above $0.85$, indicating strong discriminative capability in risk identification tasks. Cross-lingual generalization experiments further verify the robustness of the proposed framework. When trained solely on the English market, the model achieves AUC values of $0.8624$ and $0.8471$ on the Chinese and European markets, respectively, with RMSE reduced to $0.1185$, significantly outperforming competing methods. Overall, the proposed approach achieves substantial improvements in prediction accuracy, cross-lingual generalization, and early warning performance, providing an effective solution for artificial intelligence-driven sensing and risk early warning. Full article

Background: Perioperative cardiovascular risk assessment remains challenging in non-cardiac surgery, particularly in older patients and those with multiple comorbidities. Traditional models rely largely on clinical history and may not fully reflect current cardiovascular functional status. This study aimed to derive and assess the apparent performance of a new composite score, PERFORM-CV, integrating clinical, laboratory, and echocardiographic data. Methods: We conducted a prospective two-center cohort study including 503 non-cardiac surgical patients with cardiovascular comorbidity. The Revised Cardiac Risk Index (Lee/RCRI) and the AUB-HAS2 index were calculated according to their original published definitions as raw point totals ranging from 0 to 6; without additional normalization. The PERFORM-CV score was derived from univariable and multivariable analyses, with continuous predictors dichotomized using ROC-derived thresholds. Results: Emergency admission, chronic heart failure, and elevated serum creatinine remained independently associated with in-hospital mortality. Lower left ventricular ejection fraction, lower mitral annular plane systolic excursion (MAPSE), lower hemoglobin, and atrial fibrillation also contributed to the final composite score. ROC analysis showed good discrimination for PERFORM-CV (AUC 0.852; 95% CI 0.806–0.897; p < 0.001), comparable to Lee/RCRI (AUC 0.860; 95% CI 0.818–0.901; p < 0.001) and higher than AUB-HAS2 (AUC 0.779; 95% CI 0.731–0.826; p < 0.001). Conclusions: PERFORM-CV showed good apparent discrimination in the derivation cohort and may complement established bedside risk tools by incorporating echocardiographic and laboratory data. The ROC-derived thresholds should be interpreted as data-driven derivation cut-offs; resampling-based internal validation and external validation are required before broader clinical use. Full article

Acute Lymphoblastic Leukemia (ALL) is a highly prevalent hematological malignancy, especially in children, for whom precise and prompt subtype identification is essential to establish suitable treatment protocols. Current deep learning-based computer-aided diagnosis (CAD) methods for identifying ALL are hindered by numerous drawbacks, such as a dependence on solely spatial feature depictions, elevated feature dimensions, computationally extensive deep learning architectures, inadequate multi-layer feature utilization, and poor interpretability. This paper introduces LUMINA-Net, a custom, lightweight, and interpretable deep learning CAD for the automated identification and subtype diagnosis of ALL using microscopic blood smear pictures. LUMINA-Net makes four principal contributions: first, it integrates a self-attention module within a lightweight custom Convolution Neural Network (CNN) to effectively capture long-range spatial relationships across clinically pertinent cytological patterns while preserving a compact design. Second, it employs a Discrete Wavelet Transform (DWT)-based wavelet pooling layer that decreases feature dimensions by up to 96.875% while enhancing the obtained depictions with spatial-spectral information. Third, it utilizes a multi-layer feature fusion strategy that combines wavelet-pooled features from two deep layers with a third fully connected layer to create a discriminating multi-scale feature vector. Fourth, it incorporates Gradient-weighted Class Activation Mapping as a dedicated explainability process to furnish clinicians with apparent visual explanations for each classification decision. Withoit the need for image enhancement or segmentation preprocessing, LUMINA-Net outperforms the competing state-of-the-art methods on the same dataset, achieving a peak accuracy of 99.51%, specificity of 99.84%, and sensitivity of 99.51% on the publicly available Kaggle ALL dataset. This demonstrates that LUMINA-Net has the potential to be a dependable, effective, and clinically interpretable CAD tool for ALL diagnosis. Full article

Background and Objectives: Acute exacerbation of COPD (AECOPD) poses a major burden on healthcare systems, with critically ill AECOPD patients having increased morbidity and mortality. Since adverse outcomes are due both to respiratory failure and the systemic inflammatory response, prognostic markers accounting for these patterns are needed. Our aim was to investigate the predictive power of derived hematological indices for intensive care unit (ICU) mortality in patients with non-infectious versus infectious AECOPD. Materials and Methods: This is a retrospective, observational, monocentric cohort study on 88 AECOPD patients admitted to the ICU between 2018 and 2023. Descriptive statistics were performed for the entire cohort, and for predefined subgroups (non-infectious, infectious and bacterial AECOPD). Receiver Operating Characteristics (ROC) analysis was performed to test the predictive power of the studied indices. Cut-off values were identified using the Youden index. Kaplan–Meier analysis was conducted to test the association with ICU mortality. Results: Overall ICU mortality was 44%. For the whole cohort, neutrophil-to-lymphocyte ratio (NLR), neutrophil-to-platelets ratio (NPR) and systemic inflammation response index (SIRI) showed moderate predictive power for ICU mortality (areas under the curve (AUCs) of 0.71–0.73). Non-infectious and infectious subgroups were comparable in terms of size, demographics, comorbidities and baseline COPD characteristics (p > 0.05). Mortality was significantly higher in infectious AECOPD (64.6% versus 20%, p < 0.001). For non-infectious AECOPD, monocyte-to-lymphocyte ratio (MLR) and SIRI had very good predictive power (AUCs between 0.82 and 0.855), while NPR and systemic inflammation index (SII) showed moderate AUC values (between 0.7 and 0.79). In infectious AECOPD, only NPR retained fair predictive power (AUC 0.691), which improved in bacterial AECOPD (AUC 0.781). Conclusions: Derived hematological indices have different predictive values for ICU mortality. MLR and SIRI exhibited very good predictive power in non-infectious AECOPD, while NPR was the best discriminator in bacterial AECOPD. These stress the importance of individualized prognostication in AECOPD. Full article

The middle and lower reaches of the Yangtze River Basin occupy a strategically pivotal position in regional development; yet extreme drought-runoff events pose severe threats to water supply and ecological security. Despite this, systematic research gaps persist, including the lack of a unified definition, standardized identification criteria, and clear understanding of formation mechanisms for extreme drought-runoff. To address these limitations, this study focused on extreme drought-runoff in the basin, utilizing 1956–2024 discharge data from four mainstream hydrological stations and meteorological data from 171 stations. Quantitative discrimination criteria were established via Pearson-III frequency analysis; meteorological characteristics were analyzed using the Meteorological Drought Comprehensive Index; and formation mechanisms were explored through partial correlation analysis and multiple linear regression. This study innovatively proposed a basin-wide three-level quantitative discrimination criterion for drought-runoff based on the June–November flow frequency of key mainstream stations, which is distinguished from single-indicator drought identification methods (SPI/SPEI/SSI) by integrating basin-scale hydrological coherence and seasonal drought characteristics. The results revealed basin-wide extreme drought-runoff in 2006 and 2022, severe drought-runoff in 1972 and 2011, and relatively severe drought-runoff in 1959, 1992, and 2024. Typical extreme drought-runoff events were characterized by sustained low precipitation and high temperatures. Meteorological factors emerged as the primary driver during June–September, while reservoir operation and riverine water intake played secondary roles. Notably, the large-scale reservoir group in the Yangtze River Basin (53 key control reservoirs) helped alleviate drought-runoff impacts from December to May (non-flood season) via water supplementation. These findings provide a robust scientific basis for precise drought-runoff prediction and the development of targeted adaptation strategies in the Yangtze River Basin. Full article

Varroa destructor (Bee mite) and Deformed Wing Virus are primary causes of honeybee colony collapse. This study developed an automated AI-based vision inspection system for detecting bee mites and deformed bees using the YOLO algorithm. The system integrates an RGB camera, a beecomb rotation motor, and an image transmission module to enable automated dual-sided image acquisition of the beecomb. The image characteristics of normal bees, bee mites, and deformed bees were analyzed, and YOLO-based object detection models were developed to classify them. Six YOLO models—based on YOLOv8 and YOLOv11 architectures across three model sizes (nano, small, and large)—were evaluated on 405 test images (6441 objects). The proposed system reduced the inspection time from 240 s required for manual method to 20 s per beecomb, achieving 12-fold efficiency improvement. Comparative analysis showed model-task specialization: YOLOv8l excelled in detecting small bee mites (F1: 92.5%, mAP[0.5]: 92.1%), while YOLOv11s achieved the highest performance for morphologically diverse deformed bees (F1: 95.1%). Error analysis indicated that detection performance was influenced by morphological characteristics. Deformed bee detection errors correlated with overlap in wing-to-body ratio: DB Type II exhibited 18.6% miss rate, while DB Type III achieved perfect detection. In bee mite detection, a sensitivity–specificity trade-off was observed: YOLOv11l had the lowest false negatives (2.5%) but highest false positives, while YOLOv8l demonstrated superior discrimination. These results demonstrate the practical potential of the proposed system for field deployment in apiaries, supporting early pest diagnosis and improved colony health management. The model-task specialization framework provides guidance for architecture selection based on object characteristics. Future work will focus on multi-location validation and real-time monitoring integration. Full article

Transmission tower climbing is a critical specialized operation in ultra-high-voltage power maintenance and communication infrastructure servicing. However, existing lower-limb exoskeletons used for tower climbing still suffer from insufficient motion intention recognition accuracy under complex operational environments. To address this issue, this study proposes an inertial measurement unit (IMU)-based bidirectional temporal deep learning method for motion intention recognition. First, a one-dimensional convolutional neural network (1D-CNN) is employed to extract local temporal features from multi-channel IMU signals. Subsequently, a bidirectional long short-term memory network (Bi-LSTM) is introduced to model the forward and backward temporal dependencies of motion sequences. Furthermore, a temporal attention mechanism is incorporated to emphasize discriminative features at critical movement phases, enabling the precise recognition of short-duration and transitional motions. Experimental results demonstrate that the proposed method outperforms traditional machine learning approaches and unidirectional temporal models in terms of accuracy, F1-score, and other evaluation metrics. In particular, this method demonstrates significant advantages in identifying the flexion/extension phases and transitional states. This study provides an offline method for analyzing movement intentions in lower-limb exoskeleton control for power transmission tower climbing scenarios and offers a reference for developing assistive control strategies for assisted climbing tasks in this specific context. Full article

Background: Emergency triage systems using machine learning traditionally rely on structured tabular data (vital signs), creating a “contextual blind spot” that ignores diagnostic information embedded in unstructured clinical narratives. Hybrid AI models that fuse tabular and text data may improve predictive discrimination, but the magnitude and conditions under which fusion adds value remain unclear. Methods: Five databases (PubMed, Scopus, Web of Science, IEEE Xplore, ACM Digital Library) were searched from 1 January 2015 to 15 December 2025. Eligible studies employed Hybrid AI models integrating structured and unstructured emergency department data with quantitative baseline comparisons. Twenty-five studies (N ≈ 4.8 million encounters) met inclusion criteria. We extracted marginal performance gains (ΔAUC), calibration metrics, and demographic reporting. Synthesis followed SWiM principles with subgroup meta-regression testing our novel “Complexity Gradient” hypothesis. Results: Hybrid models demonstrated superior discrimination compared to tabular baselines, with effect magnitude dependent on clinical task complexity. Low-complexity tasks (tachycardia prediction) showed minimal gains (median ΔAUC + 0.036, IQR: 0.02–0.05), while high-complexity tasks (hypoxia, sepsis) demonstrated substantial improvement (median ΔAUC + 0.111, IQR: 0.09–0.13). Meta-regression confirmed complexity significantly moderated effect size (R² = 0.42, p = 0.003). Only 12% (3/25) of studies reported calibration metrics (Brier scores: 0.089–0.142). Zero studies stratified performance by race/ethnicity; 88% (22/25) failed to report training data demographics. Discussion: The complexity gradient framework explains when multimodal fusion adds predictive value: tasks where diagnostic signal resides in narrative features (temporality, negation) rather than physiological measurements. However, systematic absence of calibration reporting and fairness auditing prevents clinical deployment. Seventy-two percent of studies had high risk of bias in the analysis domain due to retrospective designs without temporal validation. Conclusions: Hybrid triage models show promise for complex diagnostic tasks but require mandatory calibration reporting and demographic performance stratification before clinical implementation. We propose minimum reporting standards including Brier scores, race-stratified metrics, and temporal validation protocols. Full article

Feature selection is essential for high-dimensional multi-source feature analysis, particularly in Internet of Things (IoT) environments characterized by data heterogeneity, redundancy, and noise. To address the need to balance classification performance, dimensionality reduction, and selection stability, this study proposes a residual-based conditional mutual information and feedback fusion (RCMF) feature-selection method. Inspired by the idea of conditional mutual information, the proposed method first introduces a residual-based indicator to characterize the incremental discriminative information retained by a candidate feature under given conditional constraints. On this basis, model-driven predictive contribution and stability score are further incorporated, and the weights of different evaluation components are iteratively updated during the feature-selection process to achieve adaptive fusion. In this way, the method jointly considers conditional discriminative information, task relevance, and selection consistency within a unified feature-evaluation procedure. Experiments on multiple publicly available benchmark and IoT-related datasets validate the rationality and effectiveness of the proposed method. Full article