Search Results (233)

The classification of submarine topography and geomorphology is essential for marine resource exploitation and ocean engineering, with wide-ranging implications in marine geology, disaster assessment, resource exploration, and autonomous underwater navigation. Submarine landscapes are highly complex and diverse. Traditional visual interpretation methods are not only inefficient and subjective but also lack the precision required for high-accuracy classification. While many machine learning and deep learning models have achieved promising results in image classification, limited work has been performed on integrating backscatter and bathymetric data for multi-source processing. Existing approaches often suffer from high computational costs and excessive hyperparameter demands. In this study, we propose a novel approach that integrates pruning-enhanced ConDenseNet with label smoothing regularization to reduce misclassification, strengthen the cross-entropy loss function, and significantly lower model complexity. Our method improves classification accuracy by 2% to 10%, reduces the number of hyperparameters by 50% to 96%, and cuts computation time by 50% to 85.5% compared to state-of-the-art models, including AlexNet, VGG, ResNet, and Vision Transformer. These results demonstrate the effectiveness and efficiency of our model for multi-source submarine topography classification. Full article

Congenital cataracts (CCs) are a leading cause of preventable childhood blindness, with genetic factors playing a crucial role in their etiology. Nance–Horan syndrome (NHS) is a rare X-linked dominant disorder associated with CCs but is often underdiagnosed due to variable expressivity, particularly in female carriers. Objective: This study aimed to explore the genetic landscape of CCs in a Swiss cohort, focusing on two novel NHS and one novel GJA8 variants and their phenotypic presentation. Methods: Whole-exome sequencing (WES) was conducted on 20 unrelated Swiss families diagnosed with CCs. Variants were analyzed for pathogenicity using genetic databases, and segregation analysis was performed. Clinical data, including cataract phenotype and associated systemic anomalies, were assessed to establish genotype–phenotype correlations. Results: Potentially pathogenic DNA sequence variants were identified in 10 families, including three novel variants, one in GJA8 (c.584T>C) and two NHS variants (c.250_252insA and c.484del). Additional previously reported variants were detected in CRYBA1, CRYGC, CRYAA, MIP, EPHA2, and MAF, reflecting genetic heterogeneity in the cohort. Notably, NHS variants displayed significant phenotypic variability, suggesting dose-dependent effects and X-chromosome inactivation in female carriers. Conclusions: NHS remains underdiagnosed due to its variable expressivity and the late manifestation of systemic features, often leading to misclassification as isolated CC. This study highlights the importance of genetic testing in unexplained CC cases to improve early detection of syndromic forms. The identification of novel NHS and GJA8 variants provides new insights into the genetic complexity of CCs, emphasizing the need for further research on genotype–phenotype correlations. Full article

Adversarial attacks in Natural Language Processing (NLP) present a critical challenge, particularly in sentiment analysis, where subtle input modifications can significantly alter model predictions. In search of more robust defenses against adversarial attacks on sentimental analysis, this research work introduces two novel defense mechanisms: the Lexicon-Based Random Substitute Model (LRSM) and the Word-Variant Voting Model (WVVM). LRSM employs randomized substitutions from a dataset-specific lexicon to generate diverse input variations, disrupting adversarial strategies by introducing unpredictability. Unlike traditional defenses requiring synonym dictionaries or precomputed semantic relationships, LRSM directly substitutes words with random lexicon alternatives, reducing overhead while maintaining robustness. Notably, LRSM not only neutralizes adversarial perturbations but occasionally surpasses the original accuracy by correcting inherent model misclassifications. Building on LRSM, WVVM integrates LRSM, Frequency-Guided Word Substitution (FGWS), and Synonym Random Substitution and Voting (RS&V) in an ensemble framework that adaptively combines their outputs. Logistic Regression (LR) emerged as the optimal ensemble configuration, leveraging its regularization parameters to balance the contributions of individual defenses. WVVM consistently outperformed standalone defenses, demonstrating superior restored accuracy and F1 scores across adversarial scenarios. The proposed defenses were evaluated on two well-known sentiment analysis benchmarks: the IMDB Sentiment Dataset and the Yelp Polarity Dataset. The IMDB dataset, comprising 50,000 labeled movie reviews, and the Yelp Polarity dataset, containing labeled business reviews, provided diverse linguistic challenges for assessing adversarial robustness. Both datasets were tested using 4000 adversarial examples generated by established attacks, including Probability Weighted Word Saliency, TextFooler, and BERT-based Adversarial Examples. WVVM and LRSM demonstrated superior performance in restoring accuracy and F1 scores across both datasets, with WVVM excelling through its ensemble learning framework. LRSM improved restored accuracy from 75.66% to 83.7% when compared to the second-best individual model, RS&V, while the Support Vector Classifier WVVM variation further improved restored accuracy to 93.17%. Logistic Regression WVVM achieved an F1 score of 86.26% compared to 76.80% for RS&V. These findings establish LRSM and WVVM as robust frameworks for defending against adversarial text attacks in sentiment analysis. Full article

Given the importance of turbidity as a key indicator of water quality, this study investigates the use of a convolutional neural network (CNN) to classify water samples into five turbidity-based categories. These classes were defined using ranges inspired by Mexican environmental regulations and generated from 33 laboratory-prepared mixtures with varying concentrations of suspended clay particles. Red, green, and blue (RGB) images of each sample were captured under controlled optical conditions, and turbidity was measured using a calibrated turbidimeter. A transfer learning (TL) approach was applied using EfficientNet-B0, a deep yet computationally efficient CNN architecture. The model achieved an average accuracy of 99% across ten independent training runs, with minimal misclassifications. The use of a lightweight deep learning model, combined with a standardized image acquisition protocol, represents a novel and scalable alternative for rapid, low-cost water quality assessment in future environmental monitoring systems. Full article

This study investigates public perceptions of sustainable mobility within university environments, which are important trip generation hubs with the potential to influence and disseminate sustainable mobility behaviors. Using sentiment analysis on 120,236 tweets from São Paulo, Rio de Janeiro, Lisbon, and Porto, tweets were classified into positive, neutral, and negative sentiments to assess perceptions across transport modes. It was hypothesized that universities would exhibit more positive sentiment toward active and public transport modes compared to perceptions of these modes within the broader city environment. Results show that active modes and public transport consistently receive higher positive sentiment rates than individual motorized modes, and, considering the analyzed contexts, universities demonstrate either similar (São Paulo) or more positive perceptions compared to the overall sentiment observed in the city (Rio de Janeiro, Lisbon, and Porto). Chi-square tests confirmed significant associations between transport mode and sentiment distribution. An exploratory analysis using topic modeling revealed that perceptions around bicycle use are linked to themes of safety, cycling infrastructure, and bike sharing. The findings highlight opportunities to promote sustainable mobility in universities by leveraging user sentiment while acknowledging limitations such as demographic bias in social media data and potential misclassification. This study advances data-driven methods to support targeted strategies for increasing active and public transport in university settings. Full article

This study analyzes the use of a lightweight image-based deep learning model to classify dairy cows into low-, medium-, and high-milk-yield categories by automatically detecting the udder region of the cow. The implemented model was based on the YOLOv11 architecture, which enables efficient object detection and classification with real-time performance. The model is trained on a public dataset of cow images labeled with 305-day milk yield records. Thresholds were established to define the three yield classes, and a balanced subset of labeled images was selected for training, validation, and testing purposes. To assess the robustness and consistency of the proposed approach, the model was trained 30 times following the same experimental protocol. The system achieves precision, recall, and mean Average Precision (mAP@50) of 0.408 ± 0.044, 0.739 ± 0.095, and 0.492 ± 0.031, respectively, across all classes. The highest precision (0.445 ± 0.055), recall (0.766 ± 0.107), and mAP@50 (0.558 ± 0.036) were observed in the low-yield class. Qualitative analysis revealed that misclassifications mainly occurred near class boundaries, emphasizing the importance of consistent image acquisition conditions. The resulting model was deployed in a mobile application designed to support field-level assessment by non-specialist users. These findings demonstrate the practical feasibility of applying vision-based models to support decision-making in dairy production systems, particularly in settings where traditional data collection methods are unavailable or impractical. Full article

Background and Objectives: Body composition changes with aging and menopause, often leading to increased adiposity and a shift in fat distribution. While BMI is commonly used in clinical practice, it does not accurately reflect fat mass or distribution. This study aims to evaluate age-related changes in both total and regional adiposity using DXA-derived indices in Korean women aged ≥ 40 years and to assess the limitations of BMI-based obesity classification. Materials and Methods: This retrospective multicenter study analyzed the DXA scans and clinical records of 914 Korean women aged 40–80 years who attended menopause clinics across multiple institutions between 2018 and 2021. We analyzed five adiposity indices: body mass index (BMI), total body fat percentage (TB%F), fat mass index (FMI), visceral adipose tissue (VAT) area, and android-to-gynoid (A/G) fat ratio. Excess adiposity was defined as BMI ≥ 23 kg/m², TB%F ≥ 40%, FMI ≥ 9 kg/m², VAT > 100 cm², or A/G ratio > 1.0. Age group comparisons were made using ANOVA, and misclassification was assessed by comparing BMI with other indices. Results: Mean BMI increased with age, peaking in the 60s before declining in the 70s. TB%F and FMI peaked in the 50s, while VAT and A/G ratio increased continuously with age. Excess adiposity was found in 41.9% of women by TB%F, 40.5% by FMI, and 59.4% by VAT in the 70s. Notably, 22% of women with normal BMI (<23 kg/m²) had VAT > 100 cm², and 35.7% had A/G > 1.0, indicating central obesity. Conclusions: DXA-based indices provide a more accurate assessment of adiposity and associated cardiometabolic risks in aging women than BMI alone. Clinical screening strategies should consider incorporating regional fat distribution markers, particularly in midlife and postmenopausal populations, to better identify individuals at risk. Full article

Accurate body condition score (BCS) monitoring in dairy cows is essential for optimizing health, productivity, and welfare. Traditional manual scoring methods are labor-intensive and subjective, driving interest in automated imaging-based systems. This study evaluated the effectiveness of 2D imaging and deep learning for BCS classification using three camera perspectives—front, back, and top-down—to identify the most reliable viewpoint. The research involved 56 Norwegian Red milking cows at the Center for Livestock Experiments (SHF) of Norges Miljo-og Biovitenskaplige Universitet (NMBU) in Norway. Images were classified into BCS categories of 2.5, 3.0, and 3.5 using a YOLOv8 model. The back view achieved the highest classification precision (mAP@0.5 = 0.439), confirming that key morphological features for BCS assessment are best captured from this angle. Challenges included misclassification due to overlapping features, especially in Class 2.5 and background data. The study recommends improvements in algorithmic feature extraction, dataset expansion, and multi-view integration to enhance accuracy. Integration with precision farming tools enables continuous monitoring and early detection of health issues. This research highlights the potential of 2D imaging as a cost-effective alternative to 3D systems, particularly for small and medium-sized farms, supporting more effective herd management and improved animal welfare. Full article

Purpose: This retrospective study aimed to compare the efficacy of cycloplegic (CR) versus non-cycloplegic refraction (NCR) methods in detecting refractive errors among children and adolescents. Methods: Electronic data from pediatric ophthalmology clinics at the University Hospital “Sveti Duh”; Zagreb, Croatia, from January 2008 to July 2023, were analyzed. Comprehensive eye examinations, including Logarithmic Visual Acuity tests, subjective refraction, cycloplegic retinoscopy, slit lamp, and fundus examinations, were conducted. Results: The dataset included 1075 individuals, with 180 undergoing NCR and 895 undergoing CR. In premyopes, the NCR group had a longer follow-up (5.04 vs. 3.45 years; p < 0.001) with similar SE progression. In low myopia, NCR showed more negative first visit SE (−1.86 D vs. −1.35 D; p < 0.001) and faster progression (p = 0.01). In high myopia, follow-up was longer in NCR (5.08 vs. 2.08 years; p = 0.03) with no other significant differences. SE progression was highest in 4–6-year-olds and significantly faster in NCR (−0.61 vs. −0.40 D/year; p = 0.05). Conclusions: Cycloplegic refraction is essential for accurately assessing refractive status, especially in cases of low myopia, as it prevents misclassification and ensures precise evaluation in children and adolescents, thereby facilitating the appropriate diagnosis and treatment of refractive errors. Full article

Objective: The classification of skin cancer is very helpful in its early diagnosis and treatment, considering the complexity involved in differentiating AK from BCC and SK. These conditions are generally not easily detectable due to their comparable clinical presentations. Method: This paper presents a new approach to hyperspectral imaging for enhancing the visualization of skin lesions called the Spectrum-Aided Vision Enhancer (SAVE), which has the ability to convert any RGB image into a narrow-band image (NBI) by combining hyperspectral imaging (HSI) to increase the contrast of the area of the cancerous lesions when compared with the normal tissue, thereby increasing the accuracy of classification. The current study investigates the use of ten different machine learning algorithms for the purpose of classification of AK, BCC, and SK, including convolutional neural network (CNN), random forest (RF), you only look once (YOLO) version 8, support vector machine (SVM), ResNet50, MobileNetV2, Logistic Regression, SVM with stochastic gradient descent (SGD) Classifier, SVM with logarithmic (LOG) Classifier and SVM- Polynomial Classifier, in assessing the capability of the system to differentiate AK from BCC and SK with heightened accuracy. Results: The results demonstrated that SAVE enhanced classification performance and increased its accuracy, sensitivity, and specificity compared to a traditional RGB imaging approach. Conclusions: This advanced method offers dermatologists a tool for early and accurate diagnosis, reducing the likelihood of misclassification and improving patient outcomes. Full article

Gender recognition from pedestrian imagery is acknowledged by many as a quasi-solved problem, yet most existing approaches evaluate performance in a within-domain setting, i.e., when the test and training data, though disjoint, closely resemble each other. This work provides the first exhaustive cross-domain assessment of six architectures considered to represent the state of the art: ALM, VAC, Rethinking, LML, YinYang-Net, and MAMBA, across three widely known benchmarks: PA-100K, PETA, and RAP. All train/test combinations between datasets were evaluated, yielding 54 comparable experiments. The results revealed a performance split: median in-domain F1 approached 90% in most models, while the average drop under domain shift was up to 16.4 percentage points, with the most recent approaches degrading the most. The adaptive-masking ALM achieved an F1 above 80% in most transfer scenarios, particularly those involving high-resolution or pose-stable domains, highlighting the importance of strong inductive biases over architectural novelty alone. Further, to characterize robustness quantitatively, we introduced the Unified Robustness Metric (URM), which integrates the average cross-domain degradation performance into a single score. A qualitative saliency analysis also corroborated the numerical findings by exposing over-confidence and contextual bias in misclassifications. Overall, this study suggests that challenges in gender recognition are much more evident in cross-domain settings than under the commonly reported within-domain context. Finally, we formalize an open evaluation protocol that can serve as a baseline for future works of this kind. Full article

Background: In the emergency setting, many diagnostic pathways incorporate change in high-sensitivity cardiac troponin (hs-cTn) concentrations (i.e., the delta) to classify patients as low-risk (rule-out) or high-risk (rule-in) for possible myocardial infarction (MI). However, the impact of analytical variation on the delta for correct classification is unknown, especially at concentrations below and around the 99th percentile. Our objective was to assess the impact of delta variation for correct risk classification across the European Society of Cardiology (ESC 0/1 h and 0/2 h), the High-STEACS, and the common change criteria (3C) pathways. Methods: A yearlong accuracy study for hs-cTnT was performed where laboratories across Canada tested three patient-based samples (level 1 target value = 6 ng/L, level 2 target value = 9 ng/L, level 3 target value = 12 ng/L) monthly across 41 different analyzers. The assigned low-delta between levels 1 and 2 was 3 ng/L (i.e., 9 − 6 = 3 ng/L) and the assigned high-delta between levels 1 and 3 was 6 ng/L (i.e., 12 − 6 = 6 ng/L). The low- and high-deltas for each analyzer were determined monthly from the measured values, with the difference calculated from the assigned deltas. The obtained deltas were then assessed via the different pathways on correct classification (i.e., percent correct with 95% confidence intervals, CI) and using non-parametric analyses. Results: The median (interquartile range) difference between the measured versus assigned low-delta (n = 436) and high-delta (n = 439) was −1 ng/L (−1 to 0). The correct classification differed among the pathways. The ESC 0/1 h pathway yielded the lowest percentage of correct classification at 35.3% (95% CI: 30.8 to 40.0) for the low-delta and 90.0% (95% CI: 86.8 to 92.6) for the high-delta. The 3C and ESC 0/2 h pathways yielded higher and equivalent estimates on correct classification: 95.2% (95% CI: 92.7 to 97.0) for the low-delta and 98.2% (95% CI: 96.4 to 99.2) for the high-delta. The High-STEACS pathway yielded 99.5% (95% CI: 98.4 to 99.9) of correct classifications for the high-delta but only 36.2% (95% CI: 31.7 to 40.9) for the low-delta. Conclusions: Analytical variation will impact risk classification for MI when using hs-cTn deltas alone per the pathways. The 3C and ESC 0/2 h pathways have <5% misclassification when using deltas for hs-cTnT in this dataset. Additional studies with different hs-cTnI assays at concentrations below and near the 99th percentile are warranted to confirm these findings. Full article

It is well-known that accurate classification of histopathological images is essential for effective diagnosis of colorectal cancer. Our study presents three attention-based decision fusion models that combine pre-trained CNNs (Inception V3, Xception, and MobileNet) with a spatial attention mechanism to enhance feature extraction and focus on critical image regions. A key innovation is the attention-driven fusion strategy at the decision level, where model predictions are weighted by relevance and confidence to improve classification performance. The proposed models were tested on diverse datasets, including 17,531 colorectal cancer histopathological images collected from the Royal Hospital in the Sultanate of Oman and a publicly accessible repository, to assess their generalizability. The performance results achieved high accuracy (98–100%), strong MCC and Kappa scores, and low misclassification rates, highlighting the robustness of the proposed models. These models outperformed individual transfer learning approaches (p = 0.009), with performance differences attributed to the characteristics of the datasets. Gradient-weighted class activation highlighted key predictive regions, enhancing interpretability. Our findings suggest that the proposed models demonstrate the potential for accurately classifying CRC images, highlighting their value for research and future exploration in diagnostic support. Full article

Predictive maintenance is essential for reducing industrial downtime and costs, yet real-world datasets frequently encounter class imbalance and require cost-sensitive evaluation due to costly misclassification errors. This study utilises the SCANIA Component X dataset to advance predictive maintenance through machine learning, employing seven supervised algorithms, Support Vector Machine, Random Forest, Decision Tree, K-Nearest Neighbours, Multi-Layer Perceptron, XGBoost, and LightGBM, trained on time-series features extracted via a sliding window approach. A bespoke cost-sensitive metric, aligned with SCANIA’s misclassification cost matrix, assesses model performance. Three imbalance mitigation strategies, downsampling, downsampling with SMOTETomek, and manual class weighting, were explored, with downsampling proving most effective. Random Forest and Support Vector Machine models achieved high accuracy and low misclassification costs, whilst a voting ensemble further enhanced cost efficiency. This research emphasises the critical role of cost-aware evaluation and imbalance handling, proposing an ensemble-based framework to improve predictive maintenance in industrial applications Full article

Background and Objectives: The use of antiseizure medications (ASMs) during pregnancy is critical to seizure control in women with epilepsy but raises concerns regarding the use of these drugs and their possible effect on the maternal serum biochemical markers used for second-trimester prenatal screening. The aim of this study was to assess the effect of ASMs on the levels of maternal serum alpha-fetoprotein (AFP), unconjugated estriol (uE3), and human chorionic gonadotropin (hCG) assessed in the serum biomarker analyses part of second-trimester prenatal screening. Materials and Methods: This retrospective cohort study included 43 pregnant women in the ASM-exposed group (levetiracetam, lamotrigine, carbamazepine, or combined therapy) and 43 matched controls without medication use. Groups were matched based on maternal age, gravidity, parity, abortion history, gestational age at testing, body mass index, and smoking status with propensity score matching. Serum AFP, uE3, and hCG levels measured at 15–20 weeks of gestation were compared between groups. The incidence of fetal congenital anomalies or aneuploidies was also compared between groups. Results: Pregnant women in the ASM-exposed group had significantly higher maternal serum AFP (1.34 ± 0.42 vs. 1.01 ± 0.31 MoM; p < 0.001) and uE3 (1.28 ± 0.39 vs. 1.05 ± 0.34 MoM; p = 0.004) than the controls. However, hCG did not differ significantly between the groups (1.07 ± 0.46 vs. 1.01 ± 0.42 MoM; p = 0.523). Regarding the ASM subgroups (levetiracetam, lamotrigine, and carbamazepine), there were no significant differences in the serum biomarkers (p > 0.05). There was no significant difference between the ASM-exposed and control groups in terms of the incidence of congenital anomalies or aneuploidies (2.3% in the ASM-exposed group vs. 2.3% in the control group; p = 1.000). Conclusions: The use of ASMs during pregnancy significantly alters second-trimester maternal serum biochemical markers, including our primary concerns, AFP and uE3, which could cause inaccurate interpretations of second-trimester prenatal screening. Clinicians should carefully consider maternal medication exposure when interpreting these biochemical markers in pregnant women with epilepsy to prevent the misclassification of fetal risks and avoid unnecessary invasive procedures. Full article