Search Results (336)

This study proposes a Variational AutoEncoder (VAE)-based Quality Controlled (QC) generative augmentation framework for North Atlantic right whale (NARW) upcall detection. Existing generative augmentation methods can generate synthetic samples; however, they do not provide a sample-level criterion for determining whether each generated sample is a positive sample that contributes to improved detector performance or a synthetic outlier that should be removed. This study learns manually extracted upcall frequency contours using a 1D-VAE and evaluates generated contour candidates in a 10-dimensional acoustic morphology feature space. The QC score is computed with respect to the reference distribution of manually extracted real upcall contours, and stochastic acceptance probabilities for borderline samples around the hard threshold are calibrated using same-call manual re-extraction variability. QC-passed contours are converted into detector training spectrograms using smooth amplitude modulation and Gaussian noise injection based on SNR statistics. Using 30,000 acoustic segments from the Kaggle NARW dataset, Original, Denoising Diffusion Probabilistic Models (DDPM), Contour-VAE without QC, and VAE-QC conditions were compared under the same detector and augmentation budget. VAE-QC with α = 0.97 achieved the highest mean AUC of 0.901, outperforming Original training (0.802), DDPM (0.845), and Contour-VAE without QC (0.810). Feature distribution and QC-score analyses further showed that VAE-QC suppresses morphology outlier tails observed in unfiltered generation. These results indicate that the key factor in generative augmentation is the QC process that defines feature boundaries useful for detector learning and selects synthetic positive samples accordingly. Full article

The Hassanat Distance (HasD) is a bounded, non-convex metric widely used in k-nearest-neighbor (KNN) classification for its robustness to noise, outliers, and heterogeneous feature scales. Its definition, however, breaks a natural symmetry: through a sign-dependent shift it assigns different distances to mirror-image pairs such as $(1,2)$ and $(-1,-2)$, distorting neighborhoods exactly in the value ranges that modern preprocessing (z-scoring, principal component analysis (PCA), learned embeddings) produces. We introduce the Sign-Symmetric Hassanat Distance (SHasD), a single branch-free formula $D(a,b)=|a-b|/(1+\max (\left|a\right|,\left|b\right|\left)\right)$ that is invariant under the reflection $x\mapsto -x$, coincides exactly with HasD on non-negative data, and removes the conditional shift entirely. We prove SHasD is a metric, and we derive a range-normalized companion, SHasD-R, that additionally restores ray monotonicity and the $[0,1)$ per-dimension bound. On 23 datasets across three normalization regimes and ten distance measures, SHasD improves significantly on HasD on data containing negative values (mean gain $+1.1$ percentage points, up to $+7.4$; Wilcoxon $p=0.0026$, Holm-corrected) and attains the best mean rank of the compared measures on signed, heavy-tailed, outlier-rich data, while preserving HasD’s robustness. An additive per-dimension decomposition yields a built-in interpretation of every prediction. Full article

Background: Uterine fibroids are the most common benign tumors of the female reproductive system and frequently affect women’s health-related quality of life (HRQoL). The increasing use of online surveys in clinical research raises concerns regarding the comparability and reliability of data collected through different recruitment methods. This study aimed to compare the clinical and methodological comparability of data obtained from an online survey and a clinic-based survey assessing symptom severity and HRQoL among women diagnosed with uterine fibroids. Methods: A cross-sectional observational study was conducted among 167 women with confirmed uterine fibroids. The clinic-based cohort included 82 patients recruited in a gynecological outpatient clinic, while the online cohort consisted of 85 women participating through an internet-based survey. Both cohorts were assessed using identical inclusion and exclusion criteria, the same questionnaire structure, and the validated Uterine Fibroid Symptom and Health-Related Quality of Life (UFS-QoL) instrument. Symptom severity scores and HRQoL outcomes were compared descriptively and methodologically. Results: The two independently recruited cohorts showed broadly similar distributions of several demographic characteristics and patient-reported outcomes, although important differences between the source populations should be considered when interpreting these findings. Mean symptom severity scores were 48.55 ± 21.61 in the clinic-based cohort and 46.33 ± 21.40 in the online cohort, while mean HRQoL scores were 57.23 ± 22.81 and 62.50 ± 19.32, respectively. Formal comparisons revealed no significant differences in symptom severity (p = 0.506) or overall HRQoL (p = 0.110). Among the six UFS-QoL domains, only the Concern domain differed significantly between cohorts (p = 0.005), whereas all other domains showed broadly similar distributions. Separate multivariable regression models were fitted for each cohort. Although direct comparison of the models was limited by differences in variable coding, both models identified symptom severity as the strongest independent predictor of HRQoL (β = −0.715 and β = −0.752) and showed similar overall model performance (R² = 0.595 and 0.578). Regression diagnostics confirmed normality of residuals, homoscedasticity, absence of influential outliers, and low variance inflation factors in both datasets. Conclusions: Despite differences in recruitment setting and source population, similar patterns of patient-reported HRQoL and symptom severity were observed across the two independently recruited cohorts. Similarities were observed across overall HRQoL scores, symptom severity measures, and multivariable regression models, suggesting that the relationships between symptom burden and quality of life remained consistent across recruitment methods. Although differences were identified in disease-related concerns and reproductive history, these findings did not substantially alter the overall pattern of results. Taken together, the findings demonstrate that similar patterns of patient-reported HRQoL and symptom severity were observed across two independently recruited cohorts of women with uterine fibroids. These findings support the feasibility of collecting disease-specific patient-reported outcomes within uterine fibroid-specific online communities but should not be interpreted as evidence that online and clinic-based recruitment methods generate equivalent study populations. Full article

To address the high dimensionality, redundancy, and noise interference present in wind turbine Supervisory Control and Data Acquisition (SCADA) data, as well as the limitations of conventional K-means algorithms—including excessive reliance on manual parameter tuning and weak anti-noise performance—this paper proposes a Variable Grid Optimized K-means (VGO-K-means) preprocessing algorithm. Z-score standardization is adopted to unify feature magnitudes. Meanwhile, an adaptive grid density calculation strategy is developed to dynamically adjust grid resolution according to the value range of each dimension, enabling accurate characterization of the spatial distribution of monitoring samples. Furthermore, a multi-index voting mechanism integrating the silhouette coefficient, Davies–Bouldin index, and inertia metric is established to adaptively determine the optimal cluster number without manual intervention. Utilizing the density discrepancy between dense normal samples and sparse outliers, the proposed method identifies abnormal samples through a clustering density threshold. Validated on real wind farm SCADA data containing 892 manually labeled abnormal samples, the VGO-K-means algorithm achieves a precision of 96.8% and an F1-score (the harmonic mean of precision and recall) of 0.89. Under identical test conditions, it outperforms traditional K-means, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and fixed-grid K-means methods. The entire workflow consumes only 3.9 s, achieving an excellent balance between detection accuracy and computational cost. The proposed framework provides a reliable and practical preprocessing solution for wind turbine condition monitoring data. Full article

Point clouds acquired by LiDAR typically contain a large number of outliers, which can substantially degrade downstream processing. Existing outlier removal methods suffer from significant performance degradation on point clouds with high outlier ratios. In this study, we propose a multi-stage outlier removal method called MORPH for point clouds with high outlier ratios. The proposed method first employs local reachability density combined with density-peak clustering to coarsely eliminate prominent outliers while preserving the global structural skeleton. Subsequently, a structure-adaptive refinement stage is introduced to generate incomplete point clouds and candidate points. Finally, a completion strategy that integrates pairwise scale consistency with local plane residual score is proposed to identify inlier points from the candidate points. Outlier removal is completed by using the obtained inlier points to fill the incomplete point clouds. Experimental results on synthetic high-contamination point clouds and real-world scanned point clouds demonstrate that MORPH achieves robust outlier removal performance while better preserving valid geometric structures. Full article

This study examines the characteristics of Group Multi-Attribute Decision Making (GMADM), including highly ambiguous information, divergent expert opinions, and bounded rationality among decision-makers. From the perspective of symmetry modeling and bias control, we propose an adaptive decision-making framework based on Spherical Fuzzy Sets (SFS). First, a spherical fuzzy quantification system for online reviews is constructed to map multi-source asymmetric information within reviews to Spherical Fuzzy Numbers. Second, an adaptive expert weighting mechanism is developed that integrates individual expert performance with the level of group consensus, dynamically adjusting weights to suppress the asymmetric interference of outlier opinions. Subsequently, we design the Credibility-based Spherical Weighted Arithmetic Mean (CSWAM) to preserve the dominance of expert judgments in a nonlinear manner and construct the Spherical Fuzzy Score function with Adaptive Hesitation Penalty (HP-SC) to ensure robustness and non-negativity in the defuzzification process. Furthermore, we extend DEMATEL and TODIM to the SFS environment, constructing a comprehensive evaluation model that captures causal relationships among attributes and asymmetric information, such as decision-makers’ loss aversion. Finally, empirical results from online movie review usefulness rankings demonstrate that this model can accurately identify and mitigate asymmetric information biases while maintaining decision symmetry equilibrium and exhibiting higher ranking stability. Full article

Background/Objectives: The gut–brain axis and systemic metabolic dysregulation are increasingly implicated in Alzheimer’s disease (AD) pathogenesis. This study aimed to characterize gut microbiota and plasma metabolomic profiles associated with amyloid pathology and cognitive impairment in patients with mild cognitive impairment (MCI). Methods: A cross-sectional multi-omics baseline analysis was performed in 47 MCI patients enrolled in a randomized, double-blind, crossover dietary intervention trial (NCT05029765). Gut microbiota composition was assessed by 16S rRNA sequencing (n = 47), and plasma metabolomics by untargeted LC-MS/MS (n = 45 after exclusion of two PCA-defined metabolomic outliers). Patients were stratified according to plasma amyloid-beta 42/40 ratio (BA42/40) and ADAScog11 score, representing complementary biomarkers of amyloid burden and cognitive impairment, respectively. Results: Higher amyloid burden and worse cognitive performance were associated with significant gut microbiota alterations, including increased alpha diversity and distinct beta diversity profiles. Differential abundance analyses consistently showed enrichment of Bacteroides-associated taxa and Akkermansia, alongside depletion of short-chain fatty acid-producing genera such as Faecalibacterium, Blautia, and Phascolarctobacterium. Plasma metabolomics identified a coherent signature associated with elevated BA42/40, characterized by accumulation of secondary bile acid sulfates and depletion of sphingolipids, neuroactive steroids, and anti-inflammatory lipid mediators, including pregnenolone sulfate, resolvin E1, and anandamide. A valid OPLS-DA discriminant model was obtained for BA42/40, whereas no predictive model was achieved for ADAScog11. Critically, this dissociation, characterized by significant microbiota differences but no metabolomic separation for ADAScog11, is itself an informative finding, suggesting that gut microbiota dysbiosis and plasma metabolomic alterations are not equally coupled to both dimensions of MCI pathophysiology. Conclusions: MCI patients with greater amyloid pathology and cognitive impairment exhibited gut microbiota dysbiosis. However, metabolic associations were observed only for BA42/40, but not for ADAScog11. These findings provide a mechanistic framework for evaluating the impact of Mediterranean diet and probiotic interventions in the longitudinal phase of the trial. Full article

Marine pollution monitoring increasingly relies on Biological Early Warning Systems (BEWSs), which use living organisms as continuous, integrative sentinels of water quality. The moon jellyfish Aurelia aurita is a sensitive but under-exploited candidate for this role. We present a computer-vision BEWS pipeline that is unsupervised at inference time and operates without labelled pollution-response data, converting side-view aquarium video of single A. aurita medusae into a binary pollution alarm. Per-frame YOLO bounding-box detections are reduced to a continuous bell-area signal and a centroid trajectory, from which eleven pulsation, kinematic, and detection-quality features are extracted on 60 s sliding windows. A per-animal baseline is fitted on a clean-water baseline (recommended ≥15 min), and a two-layer detector—fast outlier detection on the mean absolute z-score with a k-of-N rule, plus one-sided CUSUM (cumulative sum) accumulation—flags any sustained deviation. Validation on six adult medusae exposed to diesel-WAF detected all six animals (95% CI 54–100%) and produced no false alarms in 203 clean-window opportunities (exact 95% upper bound 1.8%; rule-of-three estimate ≈1.5%). First-alarm latencies ranged from 1.0 to 23.7 min, and the observed responses were described as three descriptive patterns in this pilot dataset: sharp step-change, slow drift, and mixed. The deployed anomaly scoring step contains no neural-network weights, runs in under 300 lines of Python, and is designed for field-portable use in settings where a stationary side-view camera can be positioned alongside an aquarium, although field validation remains required. Per-animal anomaly detection accommodates the strong inter-individual variability of the diesel-WAF response that limits supervised clean-versus-polluted classification at this sample size. Full article

This study investigates the effect of stabilizing axial load transfer to the bit on mechanical-specific energy during horizontal well drilling. To analyze drilling regimes, field data on WOB, ROP, RPM, TOB, and MSE were used and processed using z-score standardization, outlier removal based on the IQR criterion, and k-means clustering. The selected number of clusters was determined using the elbow method and silhouette analysis; after filtering, the dataset included 1750 observations, and the six-cluster solution was selected for exploratory interpretation. It was found that the lowest MSE values were characteristic of clusters 0, 1, and 5, among which cluster 5 demonstrated the most favorable combination of WOB, ROP, and energy efficiency. Clustering the modeled data confirmed the possibility of identifying drilling regimes with reduced MSE under a stabilized-load scenario. The predicted MSE reduction should be interpreted as a scenario-based, idealized model-based estimate obtained under quasi-steady assumptions rather than as a field-validated performance gain. The results may be used to support the selection of rational drilling parameters and to guide the further development of drilling control systems for horizontal well drilling. Full article

Outlier detection in time series is traditionally formulated as the identification of rare or extreme observations with respect to global statistical properties. While effective for stationary processes, this perspective becomes insufficient in complex and non-stationary systems, where atypical behavior may manifest as disruptions of stable relationships rather than numerical extremeness. This paper proposes a rule-based framework for detecting structural outliers in non-stationary time series. Regular system behavior is represented by an interpretable set of deterministic IF–THEN rules describing stable relational patterns between features. Each rule defines a logical context and an admissible range of a diagnostic quantity, estimated nonparametrically from historical observations satisfying the rule condition. For a given observation, the set of active rules is identified and a structural inconsistency score is computed as the fraction of violated rule consequences. Additionally, observations lacking support from high-frequency contexts are treated as candidates for structural atypicality. The method is deterministic and avoids the need for explicit probabilistic modeling or iterative parameter learning, which simplifies interpretation and implementation. The framework is illustrated on daily EUR/USD data (2010–2022) using technical indicators (EMA, RSI) and absolute log-returns as the diagnostic measure. Results provide evidence that structurally atypical events can be identified even when global statistical thresholds remain unviolated, suggesting the practical relevance of relational analysis for non-stationary time series monitoring contexts. Full article

Background/Objectives: This study describes the process of harmonising data from food consumption screeners (FCSs) and food frequency questionnaires (FFQs) in pregnant women, highlighting challenges and strategies. Methods: It is a methodological, descriptive study on the harmonisation of individual food intake data. The data were divided into two datasets: FCS and FFQ. FCS responses were categorised as “never/almost never”, “1–4 days per week”, and “≥5 days per week”. FFQ data were harmonised by deriving variables in grams per day. Outliers were identified using z-scores for total harmonised caloric intake exceeding ±2 standard deviations. The distribution and heterogeneity of the derived variables were assessed using multilevel models. Results: Data were drawn from 12 studies conducted in Brazil, part of the Brazilian Maternal and Child Nutrition Consortium (BMCNC). The sample included pregnant women aged 18 years or older, at any stage of pregnancy. The final harmonised datasets comprised eight studies (n = 5484) with FCS data and four studies (n = 1759) with FFQ data. Most food categories in the FCS dataset had comparable frequencies across studies, with differences observed for natural juices, soft drinks, and sweetened beverages. In the FFQ data, the largest variations in daily consumption were found for leafy vegetables, sweetened beverages, and soft drinks. Heterogeneity ranged from less than 0.01% for beans (FCS) to 15.5% for fruits and natural juices (FFQ). Conclusions: By enabling standardised analyses across diverse Brazilian populations, the harmonised BMCNC datasets provide a valuable resource for investigating nutritional inequities and supporting future research to improve maternal and child nutrition. Full article

Background/Objectives: For patients with profound deafness resulting from auditory nerve pathology, as in Neurofibromatosis type 2, auditory brainstem implantation (ABI) can restore meaningful acoustic input. The literature reporting real-world results for ABI users is limited, especially regarding patients with bilateral implants. Here, we provide an updated report on the audiologic outcomes among all ABI patients treated at a tertiary institution, including high-performing bilateral ABI users. Methods: In this updated and expanded retrospective case series, audiologic outcomes were reviewed in sixteen consecutive patients who underwent ABI placement by a single neurosurgeon-neurotologist team at our center since 2018. Implantation in four of these patients was on their second side after having undergone first side implantation prior to receiving care at our hospital. Main outcome measures were sound awareness (sound-field threshold testing) and speech understanding (pattern perception, spondee, open-set speech testing). Results: Sound awareness was achieved in 100% of patients (16/16) using an average of 12 electrodes (range 7–20). Persistent non-auditory sensations were reported by 12.5% (2/16). Postoperative speech differentiation (with or without lip-reading) was experienced in 87.5% (14/16). Two second-sided ABI recipients experienced exceptional outcomes as high-performing outliers: one achieved 57% audio only and 86% audio + visual hearing in noise test (HINT) sentence scores; the second bilateral user scored 92% with auditory-only input. Conclusions: ABI represents a viable option for patients who are at risk of developing bilateral profound deafness resulting from auditory nerve disruption. Second sided device implantation is safe and has the potential to significantly improve auditory outcomes. Full article

Early stage mineral exploration requires the reliable identification of anomalous gold zones from drillhole geochemistry in data-limited environments. This study applies a machine learning (ML) classification framework to detect anomalous gold zones (Au ≥ 0.68 ppm; 90th percentile) from bulk XRF multielement drillhole geochemistry in a Paleoproterozoic Birimian greenstone belt sulphide-hosted orogenic gold system, West African Craton. A total of 53,126 one-metre diamond core samples from 301 drillholes were preprocessed within a compositional data analysis (CoDA) framework, with Au being explicitly excluded from the centred log-ratio (CLR) transformation to eliminate target–predictor circularity. After Minimum Covariance Determinant (MCD) outlier filtering, 40,385 samples were retained to construct a 19-feature matrix of 10 CLR-transformed elements, 1 rock-type feature, and 8 sulphide–lithology interaction features. Drillhole-based block cross-validation (DH-block CV), validated by an experimental along-hole variogram (practical autocorrelation range ≈ 20 m), ensured spatially honest performance estimates. Four nonlinear classifiers—Random Forest (RF), XGBoost, LightGBM, and Multi-Layer Perceptron (MLP)—were benchmarked against a Logistic Regression (LR) linear baseline. All nonlinear classifiers achieved validation AUC of 0.936–0.938, outperforming LR (AUC = 0.931) with F1-score improvements of +0.09 to +0.11 and precision gains of up to +35 percentage points—directly reducing wasted drill holes in applied exploration. MLP recorded the highest F1-score (0.666) and precision (0.765), and XGBoost the highest recall (0.787). Permutation importance identified S-Ti (ΔAUC = 0.028), S-Fe (0.021), and S-Al (0.013) as the top-ranked features, confirming that sulphide enrichment relative to lithological background is the primary discriminating signal. Partial dependence analysis revealed a threshold-driven non-monotonic Fe dependence at CLR(Fe) ≈ 3, marking the transition from lithological dilutant to sulphide co-indicator—a nonlinear pattern inaccessible to linear classifiers. Full article

Radiomics has the potential to improve risk stratification in head and neck squamous cell carcinoma (HNSCC), but clinical adoption is limited by inconsistent performance across institutions. A key source of variability is how radiomic features are generated, preprocessed, and selected prior to model development. This multicenter study evaluated how radiomics parameterization and feature selection strategies affect external model performance, feature stability, and time-to-event risk stratification. We studied pre-treatment CT scans from 752 patients with primary HNSCC from three hospitals. For each scan, 1648 radiomic features were computed using 20 different preparation methods that varied in scaling, outlier removal, and gray-level bin width. We compared five feature selection methods: Graph-FS with connected components, Boruta, Lasso, RFE-RF, and mRMR. The classification models used were Random Forest, XGBoost, CatBoost, and Logistic Regression. We measured performance using external ROC-AUC, bootstrap confidence intervals, Brier score, and RobustScore. Stability of feature selection was assessed using the Kuncheva and Jaccard indices. Cox proportional hazards models confirmed time-to-event results, and consensus SHAP analysis helped explain the models. Radiomics parameterization influenced model performance, and no single configuration was optimal across all analyses. Radiomics-only models outperformed clinical-only models, while clinical–radiomics models achieved the highest overall performance. mRMR and Lasso produced the highest average external AUCs, while Graph-FS showed the greatest stability. The best classification model achieved an external AUC of 0.817. In Cox validation, the best clinical–radiomics configuration achieved an external C-index of 0.662 and separated high- and low-risk patients in the external cohort. Full article

Accurate obesity risk classification is often hindered by the asymmetric and heteroscedastic nature of health data, where traditional mean-based machine learning models fail to capture critical distribution tails. This study addresses this gap by proposing a robust Expectile Random Forest (ERF) model, a novel ensemble architecture that integrates an expectile learning framework via the Asymmetric Least Squares (ALS) loss function for seven-level (multilevel) classification. Utilizing a dataset of 2111 empirical records, the sensitivity analysis identifies $\tau =0.7$ as the optimal configuration, achieving an overall Accuracy of 94.6 ± 0.7% and a Macro F1-Score of 94.5 ± 0.7%. This performance represents a significant quantitative improvement over state-of-the-art benchmarks, outperforming XGBoost by 1.8% and standard Random Forest by 3.9%. Feature importance analysis identifies body weight, age, and sedentary factors as primary predictors, while the ERF model demonstrates exceptional ordinal consistency and robustness against clinical outliers. These findings provide a superior methodological framework for developing precise medical decision support systems, shifting the paradigm from central-tendency predictions to tail-sensitive health risk mapping. Full article