Search Results (15)

Ensuring high data quality in traffic crash datasets is critical for effective safety analysis and policymaking. This study presents an AI-assisted framework for auditing crash data integrity by detecting potentially mislabeled records related to construction zone (czone) involvement. A GPT-3.5 model was fine-tuned using a fusion of structured crash attributes and unstructured narrative text (i.e., multimodal input) to predict work zone involvement. The model was applied to 6400 crash reports to flag discrepancies between predicted and recorded labels. Among 80 flagged mismatches, expert review confirmed four records as genuine misclassifications, demonstrating the framework’s capacity to surface high-confidence labeling errors. The model achieved strong overall accuracy (98.75%) and precision (86.67%) for the minority class, but showed low recall (14.29%), reflecting its conservative design that minimizes false positives in an imbalanced dataset. This precision-focused approach supports its use as a semi-automated auditing tool, capable of narrowing the scope for expert review and improving the reliability of large-scale traffic safety datasets. The framework is also adaptable to other misclassified crash attributes or domains where structured and unstructured data can be fused for data quality assurance. Full article

Background/Objectives: This study aims to determine the prevalence of stunting in children under two years old and its association with the maternal profile (including anthropometric measurements), care, feeding practices, and socioeconomic level. It also attempts to assess if stunting may contribute to an underestimation of wasting by performing a preliminary speculative analysis using the expected height for age instead of the real observed height of the children. Methods: The study employed a cross-sectional design, examining mother–child pairs in the rural municipality of Ampefy in the Itasy Region of Madagascar, between 2022 and 2023. A total of 437 mother–child (0–24 months) pairs participated in the study. A questionnaire was administered to collect data on the maternal lifestyle. Maternal medical histories were reviewed, and anthropometric parameters of both the mothers and their child were taken by specialized and trained health professionals with multiple years of experience. Results: The prevalence of stunting in children was 57.4% (95% CI: 52.64–62.10). Stunting was associated with maternal anthropometric measurements (p < 0.001), maternal education (p = 0.004), and breastfeeding (p = 0.047), which appears to have a protective effect. The weight-for-length z-score indicated that only 12.4% of the total children were affected by wasting. In the preliminary speculative analysis using the WHO height-for-age standard, the theoretical prevalence of wasting was estimated to be 42.3%, with a considerable prevalence of severe wasting. The main limitations of this study were the possible selection bias, the limitations inherent to the taking of anthropometric measurements in small children, and therefore, the possible misclassification of the children. The use of a theoretical weight-for-length z-score to estimate a theoretical prevalence of wasting using an untested speculative analysis is also a limitation to the validity of the estimation. Conclusions: Stunting affected over half of the children included in the study (57.4%), but the prevalence of wasting was below what was expected, at 12.4%. In the preliminary speculative analysis using the expected height for age, it was estimated that wasting could possibly affect up to 42.3% of the children. This discrepancy, while it cannot be taken as factual due to the nature of the analysis, could serve as a warning that perhaps the elevated rates of stunting may be masking wasting in some children and other forms of nutritional assessments may be needed in areas where stunting is prevalent. Full article

In the construction industry, ensuring the proper installation, retention, and dismantling of temporary structures, such as jack supports, is critical to maintaining safety and project timelines. However, inconsistencies between on-site data and construction documentation remain a significant challenge. To address this, this study proposes an integrated monitoring framework that combines computer vision-based object detection and document recognition techniques. The system utilizes YOLOv5 for detecting jack supports in both construction drawings and on-site images captured through wearable cameras, while optical character recognition (OCR) and natural language processing (NLP) extract installation and dismantling timelines from work orders. The proposed framework enables continuous monitoring and ensures compliance with retention periods by aligning on-site data with documented requirements. The analysis includes 23 jack supports monitored daily over 28 days under varying environmental conditions, including lighting changes and structural configurations. The results demonstrate that the system achieves an average detection accuracy of 94.1%, effectively identifying discrepancies and reducing misclassifications caused by structural similarities and environmental variations. To further enhance detection reliability, methods such as color differentiation, construction plan overlays, and vertical segmentation were implemented, significantly improving performance. This study validates the effectiveness of integrating visual and textual data sources in dynamic construction environments. The study supports the development of automated monitoring systems by improving accuracy and safety measures while reducing manual intervention, offering practical insights for future construction site management. Full article

Background/Objectives: Cutaneous T-cell lymphoma (CTCL), including Mycosis fungoides (MF) and Sézary syndrome (SS), is a challenging-to-diagnose lymphoproliferative malignancy characterized by T-cell dysfunction and progressive cutaneous and extra cutaneous involvement. Disease severity assessment in CTCL is crucial for guiding treatment. This study aims to evaluate the interrater agreement and interrater reliability of mSWAT among dermatology residents and identify lesion types most prone to scoring variability. Methods: Sixteen dermatology residents with varied experience levels assessed 14 patients with confirmed MF/SS diagnoses. Using mSWAT, residents independently scored lesions severity on a standardized set of patient’s photos. The results were compared with reference mSWAT scores provided by an experienced clinician. Descriptive statistics and the Shapiro–Wilk test were applied to evaluate data distributions, while Student’s t-test assessed score deviations from reference values. Furthemore, we conducted a pilot the high frequency ultrasound (HFUS) study on a single patient, whose mSWAT score and photographs are also presented in the manuscript. Results: Significant discrepancies were observed in 64.29% of cases (9/14), with tumors and infiltrative lesions in erythrodermic SS patients posing particular scoring challenges. Misclassification of tumors as patches or plaques was a frequent issue, leading to underestimations in mSWAT scores. Residents’ assessments of infiltrative lesions were also notably inconsistent. Conclusions: This study highlights significant interobserver variability in mSWAT scoring among less experienced dermatology residents, particularly with tumor and erythrodermic lesions. Findings underscore the need for enhanced training and standardized scoring protocols to improve mSWAT reliability. Similar to other comparable indices, such as PASI, the mSWAT should be employed consistently by the same physician during each assessment to systematically monitor and evaluate the skin condition of a patient under observation. However, broader application requires the acquisition of sufficient experience. The study suggests the use of the HFUS as an objective method of assessment of the skin lesion infiltration in MF/SS patients. Full article

Sea ice is a crucial component of the global climate system. The China–French Ocean Satellite Scatterometer (CFOSAT/SCAT, CSCAT) employs an innovative rotating fan beam system. This study applied principal component analysis (PCA) to extract classification features and developed an ensemble machine learning approach for sea ice detection. PCA identified key features from CSCAT’s backscatter information, representing outer and sweet swath observations. The ensemble model’s performances (OA and Kappa) for the Northern and Southern Hemispheres were 0.930, 0.899, and 0.844, 0.747, respectively. CSCAT achieved an accuracy of over 0.9 for close ice and open water but less than 0.3 for open ice, with misclassification of open ice as closed ice. The sea ice extent discrepancy between CSCAT and the National Snow and Ice Data Center (NSIDC) was −0.06 ± 0.36 million km² in the Northern Hemisphere and −0.03 ± 0.48 million km² in the Southern Hemisphere. CSCAT’s sea ice closely matched synthetic aperture radar (SAR) imagery, indicating effective sea ice and open water differentiation. CSCAT accurately distinguished sea ice from open water but struggled with open ice classification, with misclassifications in the Arctic’s Greenland Sea and Hudson Bay, and the Antarctic’s sea ice–water boundary. Full article

Deep domain adaptation techniques have recently been the subject of much research in machinery fault diagnosis. However, most of the work has been focused on domain alignment, aiming to learn cross-domain features by bridging the gap between source and target domains. Despite the success of these methods in achieving domain alignment, they often overlook the class discrepancy present in cross-domain scenarios. This can result in the misclassification of target domain samples that are located near cluster boundaries or far from their associated class centers. To tackle these challenges, a novel approach called deep domain adaptation with correlation alignment and supervised contrastive learning (DCASCL) is proposed, which synchronously realizes both domain distribution alignment and class distribution alignment. Specifically, the correlation alignment loss is used to enforce the model to generate transferable features, facilitating effective domain distribution alignment. Additionally, classifier discrepancy loss and supervised contrastive learning loss are integrated to carry out feature distribution alignment class-wisely. The supervised contrastive learning loss leverages class-specific information of source and target samples, which efficiently promotes the compactness of samples of the same class and the separation of samples from different classes. Moreover, our approach is extensively validated across three diverse datasets, demonstrating its effectiveness in diagnosing machinery faults across different domains. Full article

Land desertification is a key environmental problem in China, especially in Northwest China, where it seriously affects the sustainable development of natural resources. In this paper, we combine high-resolution satellite remote sensing images and UAV (unmanned aerial vehicle) visible light images to extract desert vegetation data and quickly locate and accurately monitor land desertification in relevant areas according to changes in vegetation coverage. Due to the strong light and dry climate of deserts in Northwest China, which results in deeper vegetation shadow texture and mostly dry shrubs with fewer stems and leaves, the accuracy of the vegetation index commonly used in visible remote sensing image classification is not able to meet the requirements for monitoring and evaluating land desertification. For this reason, in this paper, we took the Hangjin Banner in Bayannur as an example and constructed a new vegetation index, the HSVGVI (hue–saturation–value green enhancement vegetation index), based on the HSV (hue–saturation–value) color space using channel enhancement that can improve the extraction accuracy of desert vegetation and reduce misclassification. In addition, in order to further test the extraction accuracy, samples of densely vegetated and multi-shaded areas were divided in the study area according to the accuracy-influencing factors. At the same time, the HSVGVI was compared with the vegetation indices EXG (excess green index), RGBVI (red–green–blue vegetation index), MGRVI (modified green–red vegetation index), NGBDI (normalized green–red discrepancy index), and VDVI (visible-band discrepancy vegetation index) constructed based on the RGB (red–green–blue) color space. The experimental results show that the extraction accuracy of the EXG and other vegetation indices constructed in RGB color space can only reach 70%, while the extraction accuracy of the HSVGVI can reach more than 95%. In summary, the HSVGVI proposed in this paper can better realize the extraction of desert vegetation data and can provide a reliable technical tool for monitoring and evaluating land desertification. Full article

Background: In Sweden, all patients with urinary bladder cancer (UBC) are recorded in the Swedish National Register for Urinary Bladder Cancer (SNRUBC). The purpose of this study was to validate the registered clinical tumour categories (cT-categories) in the SNRUBC for Norrland University Hospital, Sweden, from 2009 to 2020, inclusive. Methods: The medical records of all 295 patients who underwent radical cystectomy for the treatment of UBC were reviewed retrospectively. Possible factors impacting the cT-categories were identified. To optimise cT-classification, computed tomography urography of all patients with suspected tumour-associated hydronephrosis (TAH) or suspected tumour in bladder diverticulum (TIBD) were retrospectively reviewed by a radiologist. Discrepancy was tested with a logistic regression model. Results: cT-categories differed in 87 cases (29.5%). Adjusted logistic regression analysis found TIBD and TAH as significant predictors for incorrect registration; OR = 7.71 (p < 0.001), and OR = 17.7, (p < 0.001), respectively. In total, 48 patients (68.6%) with TAH and 12 patients (52.2%) with TIBD showed discrepancy regarding the cT-category. Incorrect registration was mostly observed during the years 2009–2012. Conclusion: The study revealed substantial incorrect registration of cT-categories in SNRUBC. A major part of the misclassifications was related to TAH and TIBD. Registration of these variables in the SNRUBC might be considered to improve correct cT-classification. Full article

Question classification based on Bloom’s Taxonomy (BT) has been widely accepted and used as a guideline in designing examination questions in many institutions of higher learning. The misclassification of questions may happen when the classification task is conducted manually due to a discrepancy in the understanding of BT by academics. Hence, several automated examination question classification systems have been proposed by researchers to perform question classification accurately. Most of this research has focused on specific subject areas only or single-sentence type questions. There has been a lack of research on question classification for multi-sentence type and multi-subject questions. This paper proposes a question classification system (QCS) to perform the examination of question classification using a semantic and synthetic approach. The questions were taken from various subjects of an engineering diploma course, and the questions were either single- or multiple-sentence types. The QCS was developed using a natural language toolkit (NLTK), Stanford POS tagger (SPOS), Stanford parser’s universal dependencies (UD), and WordNet similarity approaches. The QCS used the NLTK to process the questions into sentences and then word tokens, such as SPOS, to tag the word tokens and UD to identify the important word tokens, which were the verbs of the examination questions. The identified verbs were then compared with the BT’s verbs list in terms of word sense using the WordNet similarity approach before finally classifying the questions according to BT. The developed QCS achieved an overall 83% accuracy in the classification of a set of 200 examination questions, according to BT. Full article

California’s San Joaquin Valley is both drought-prone and water-scarce but relies on high-intensity agriculture as its primary economy. Climate change adaptation strategies for high-intensity agriculture require reliable and highly resolved land use classification data to accurately account for changes in crop water demand, greenhouse gas (GHG) emissions, and farmgate revenue. Understanding direct and indirect economic impacts from potential changes to high-intensity agriculture to reduce groundwater overdrafts, such as reductions in the cultivated area or switching to less water-intensive crops, is unachievable if land use data are too coarse and inconsistent or misclassified. This study quantified the revenue, crop water requirement, and GHG emission discrepancies resulting from land use misclassification in the United States’ most complex agricultural region, California’s San Joaquin Valley. By comparing three commonly used land use classification datasets—CropScape, Land IQ, and Kern County Agriculture—this study found that CropScape led to considerable revenue and crop water requirement discrepancies compared to other sources. Crop misclassification across all datasets resulted in an underestimation of GHG emissions. The total revenue discrepancies of misclassified crops by area for the 2016 dataset comparisons result in underestimations by CropScape of around USD 3 billion and overestimation by LIQ and Kern Ag of USD 72 million. Reducing crop misclassification discrepancies is essential for crafting climate resilience strategies, especially for California, which generates USD 50 billion in annual agricultural revenue, faces increasing water scarcity, and aims to reach carbon neutrality by 2045. Additional investments are needed to produce spatial land use data that are highly resolved and locally validated, especially in high-intensity agriculture regions dominated by specialty crops with unique characteristics not well suited to national mapping efforts. Full article

Precise prognosis is crucial for selection of adjuvant therapy in breast cancer. Molecular subtyping is increasingly used to complement immunohistochemical and pathological classification and to predict recurrence. This study compares both outcomes in a clinical setting. Molecular subtyping (MammaPrint^®, TargetPrint^®, and BluePrint^®) and pathological classification data were compared in a cohort of 143 breast cancer patients. High risk clinical factors were defined by a value of the proliferation factor Ki67 equal or higher than 14% and/or high histological grade. The results from molecular classification were considered as reference. Core needle biopsies were found to be comparable to surgery samples for molecular classification. Discrepancies were found between molecular and pathological subtyping of the samples, including misclassification of HER2-positive tumors and the identification of a significant percentage of genomic high risk T1N0 tumors. In addition, 20% of clinical low-risk tumors showed genomic high risk, while clinical high-risk samples included 42% of cases with genomic low risk. According to pathological subtyping, a considerable number of breast cancer patients would not receive the appropriate systemic therapy. Our findings support the need to determine the molecular subtype of invasive breast tumors to improve breast cancer management. Full article

Based on a large volume of observational scientific studies and many summary papers, a high consumption of meat and processed meat products has been suggested to have a harmful effect on human health. These results have led guideline panels worldwide to recommend to the general population a reduced consumption of processed meat and meat products, with the overarching aim of lowering disease risk, especially of cancer. We revisited and updated the evidence base, evaluating the methodological quality and the certainty of estimates in the published systematic reviews and meta-analyses that examined the association between processed meat consumption and the risk of cancer at different sites across the body, as well as the overall risk of cancer mortality. We further explored if discrepancies in study designs and risks of bias could explain the heterogeneity observed in meta-analyses. In summary, there are severe methodological limitations to the majority of the previously published systematic reviews and meta-analyses that examined the consumption of processed meat and the risk of cancer. Many lacked the proper assessment of the methodological quality of the primary studies they included, or the literature searches did not fulfill the methodological standards needed in order to be systematic and transparent. The primary studies included in the reviews had a potential risk for the misclassification of exposure, a serious risk of bias due to confounding, a moderate to serious risk of bias due to missing data, and/or a moderate to serious risk of selection of the reported results. All these factors may have potentially led to the overestimation of the risk related to processed meat intake across all cancer outcomes. Thus, with the aim of lowering the risk of cancer, the recommendation to reduce the consumption of processed meat and meat products in the general population seems to be based on evidence that is not methodologically strong. Full article

The automatic classification of various types of cardiomyopathies is desirable but has never been performed using a convolutional neural network (CNN). The purpose of this study was to evaluate currently available CNN models to classify cine magnetic resonance (cine-MR) images of cardiomyopathies. Method: Diastolic and systolic frames of 1200 cine-MR sequences of three categories of subjects (395 normal, 411 hypertrophic cardiomyopathy, and 394 dilated cardiomyopathy) were selected, preprocessed, and labeled. Pretrained, fine-tuned deep learning models (VGG) were used for image classification (sixfold cross-validation and double split testing with hold-out data). The heat activation map algorithm (Grad-CAM) was applied to reveal salient pixel areas leading to the classification. Results: The diastolic–systolic dual-input concatenated VGG model cross-validation accuracy was 0.982 ± 0.009. Summed confusion matrices showed that, for the 1200 inputs, the VGG model led to 22 errors. The classification of a 227-input validation group, carried out by an experienced radiologist and cardiologist, led to a similar number of discrepancies. The image preparation process led to 5% accuracy improvement as compared to nonprepared images. Grad-CAM heat activation maps showed that most misclassifications occurred when extracardiac location caught the attention of the network. Conclusions: CNN networks are very well suited and are 98% accurate for the classification of cardiomyopathies, regardless of the imaging plane, when both diastolic and systolic frames are incorporated. Misclassification is in the same range as inter-observer discrepancies in experienced human readers. Full article

Long-lasting precipitation deficits or heat waves can induce agricultural droughts, which are generally defined as soil moisture deficits that are severe enough to negatively impact vegetation. However, during short soil moisture drought events, the vegetation is not always negatively affected and sometimes even thrives. Due to this duality in agricultural drought impacts, the term “agricultural drought” is ambiguous. Using the ESA’s remotely sensed CCI surface soil moisture estimates and MODIS NDVI vegetation greenness data, we show that, in major European droughts over the past two decades, asynchronies and discrepancies occurred between the surface soil moisture and vegetation droughts. A clear delay is visible between the onset of soil moisture drought and vegetation drought, with correlations generally peaking at the end of the growing season. At lower latitudes, correlations peaked earlier in the season, likely due to an earlier onset of water limited conditions. In certain cases, the vegetation showed a positive anomaly, even during soil moisture drought events. As a result, using the term agricultural drought instead of soil moisture or vegetation drought, could lead to the misclassification of drought events and false drought alarms. We argue that soil moisture and vegetation drought should be considered separately. Full article

Discrepancies in trade statistics can be normal or benign and attributed to a wide variety of unintentional factors, or in some instances within the timber products sector, such discrepancies can be associated with “systemic” factors that distort trade statistics, including (i) measurement and shipment issues, (ii) misreporting of product volumes, (iii) misclassification of timber product types, and (iv) government regulations concerned about trade. This study measured trade discrepancies in logs and lumber trade statistics for China and its trading partner countries from 2002 to 2018 using a time-lagged function, based on the customs data available from Global Trade Information Services (GTIS), with the aim of exploring a more nuanced understanding of trade discrepancies and their “systemic” factors. The results showed that the range of overall discrepancies in logs and lumber trade statistics shrunk over time, from [−0.069, 1.207] in 2002–2007 to [−0.120, 0.408] in 2013–2018. The larger trade flows of logs and lumber from Russia, New Zealand, and the U.S. (each above 10% of total China’s import) showed small trade statistics discrepancy ratios, which were less than ± 0.06. However, trade discrepancies still remained large at the disaggregated level, and significant differences of trade discrepancies between tropical and non-tropical countries. The range of trade discrepancies in hardwood logs increased from 2002 to 2018 and appeared to be attributed to misclassification and misreporting in tropical countries such as Indonesia, the Philippines, Thailand, and Ghana. However, these countries’ trade flows are becoming relatively minor over time. Government policies are suggested to play an important role in influencing both the occurrence and resolution of trade discrepancies. Full article