Validation Dataset Research Articles (Page 1)

Atrial fibrillation (AF) is a prevalent arrhythmia with significant health risks, often underdiagnosed due to limitations in traditional screening methods. This study investigates the effectiveness of an AI-based electronic stethoscope for AF screening, comparing it to other portable devices. A retrospective study was conducted using 496 cardiac sound recordings from patients with and without AF. The recordings were divided into derivation and validation datasets. An AI model, combining ResNet34 and a 12-layer Vision Transformer (ViT), was developed and trained on the derivation dataset. The model's performance was evaluated using sensitivity, specificity, accuracy, positive and negative predictive values, and the area under the receiver operating characteristic (ROC) curve (AUC). Additionally, a non-consecutive day twice cardiac sound collection was performed on 74 samples to assess the model's consistency. The AI model achieved high performance metrics in both derivation and validation datasets. In the derivation dataset, sensitivity was 0.95 (95% CI, 0.90-0.97), specificity was 0.90 (95% CI, 0.83-0.94), accuracy was 0.92 (95% CI, 0.90-0.96), positive predictive value was 0.92 (95% CI, 0.87-0.96), and negative predictive value was 0.93 (95% CI, 0.86-0.96). In the validation dataset, sensitivity was 0.94 (95% CI, 0.88-0.98), specificity was 0.91 (95% CI, 0.83-0.96), accuracy was 0.93 (95% CI, 0.89-0.96), positive predictive value was 0.93 (95% CI, 0.86-0.97), and negative predictive value was 0.93 (95% CI, 0.85-0.97). The AUC for the derivation dataset was 0.92 (95% CI, 0.89-0.96), and for the validation dataset, it was 0.93 (95% CI, 0.88-0.97). The non-consecutive day cardiac sound collection resulted in a Cohen's Kappa value of 0.74, indicating good consistency in the model's judgments. The AI-based electronic stethoscope shows promise as a reliable and accessible tool for AF screening, with potential applications in primary healthcare and general population screening.

Early and severe (ES) midline shift (MLS ≥ 10mm) simultaneously occurring within 24h after endovascular thrombectomy (EVT) is a life-threatening emergency that requires immediate intervention. This study aims to describe ES-MLS and develop a predictive model in patients with anterior circulation occlusion who have undergone EVT. This retrospective cohort study utilized data from a prospective registry. Functional outcome was defined as a modified Rankin Scale score of 0-2. Radiomic features extracted from pre-EVT diffusion-weighted imaging were subjected to LASSO regression with fourfold cross-validation. Clinical features were selected via multivariable regression and integrated into a nomogram, with performance evaluated through receiver operating characteristic curve analysis in both training and validation datasets. A total of 481 patients (median age 68 [IQR 58-76], 39.7% female) were included in this study, which consisted of a training dataset (n = 361) and a validation dataset (n = 120). In the ES-MLS group, 85.7% had died and none had a functional outcome at the 90-day follow-up. Recanalization, NIHSS score, and two radiomic features were identified as factors associated with ES-MLS in the nomogram. The predictive model exhibited an area under the curve (AUC) of 0.844 (95% confidence interval [CI], 0.803-0.880) in the training dataset and 0.823 (95% CI, 0.743-0.887) in the validation dataset. This is the initial structured overview of ES-MLS after EVT, featuring a model designed for personalized prediction of ES-MLS. The tool may enhance patient selection before EVT and refine the aggressive monitoring strategy after EVT.

Invasive candidiasis (IC) remains an infection with high incidence and mortality rates in the ICU setting, particularly among patients treated with broad-spectrum antibiotics. This study aims to investigate the association between detailed antibiotic usage profiles and the occurrence of IC, and to develop an IC-predictive model specialized for patients who received broad-spectrum antibiotics. We retrospectively collected detailed information on antibiotic categories, treatment duration, combination therapies and other clinical data of enrolled patients. Univariate and multivariate logistic regression analyses were performed to identify risk factors for IC and to construct a nomogram model. We analyzed 1,260 patients treated with broad-spectrum antibiotics and 877 without. After adjusting for IC-related risk factors using propensity score matching (PSM) and inverse probability of treatment weighting (IPTW), broad-spectrum antibiotics remained an independent risk factor for IC. Among patients receiving antibiotic monotherapy, lipopeptides, glycopeptides and oxazolidinones were the top three antibiotic classes associated with an increased risk of IC. The duration of antibiotic therapy showed a positive correlation with IC risk. Combination therapy significantly increased the risk of IC (odds ratio [OR] = 2.341, 95% confidence interval [CI]: 1.316-4.162), with the combination of beta-lactams/beta-lactamase inhibitors and glycopeptides showing the highest IC risk. Based on univariate and multivariate regression analyses, we developed an IC risk nomogram specific to patients receiving broad-spectrum antibiotics, including smoking history, sepsis, continuous renal replacement therapy (CRRT), prognostic nutritional index (PNI), use of beta-lactams/beta-lactamase inhibitors and plasma (1,3)-β-D-glucan (BDG) positivity. The model demonstrated good predictive performance with an area under the curve (AUC) of 0.863 (95% CI: 0.806-0.920) in the training dataset and 0.784 (95% CI: 0.685-0.883) in the validation dataset. Decision curve analysis (DCA) and clinical impact curve (CIC) analysis demonstrated favorable clinical benefits of the model. Our findings suggest that specific antibiotic profiles-type, duration, and combination-were significantly associated with IC. Furthermore, we developed a nomogram to predict IC risk among patients treated with broad-spectrum antibiotics, which showed good predictive performance and potential clinical utility.

Delirium is an acute syndrome characterized by fluctuating attention, cognitive impairment, and severe disorganization of behavior, which has been shown to affect up to 31% of patients in the intensive care unit (ICU). Early detection can enable timely interventions and improved health outcomes. While artificial intelligence (AI) models have shown great potential for ICU delirium prediction using structured electronic health records (EHR), most studies have either not leveraged state-of-the-art AI models, been limited to single-center cohorts, or relied on small datasets for development and validation. In this study, we introduce DeLLiriuM, a novel LLM-based delirium prediction model that utilizes EHR data from the first 24 hours of ICU admission to estimate a patient's risk of developing delirium for the remainder of their ICU stay. We developed and validated DeLLiriuM using ICU admissions from 104,303 patients across 195 hospitals in three large databases: the eICU Collaborative Research Database, the Medical Information Mart for Intensive Care (MIMIC)-IV, and the University of Florida's Integrated Data Repository. Our DeLLiriuM model achieved superior performance compared to all baseline models on the external validation set, measured by the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC) metric. DeLLiriuM attained AUROC 82.4 (95% confidence interval 81.8-83.0) and AUPRC 11.8 (95% confidence interval 11.3-12.4) across 77,543 patients spanning 194 hospitals. Our approach of transforming structured EHR data into an unstructured text format, the primary data modality for LLMs, enables our DeLLiriuM model to capture clinical contextual information, resulting in improved predictive performance. To the best of our knowledge, DeLLiriuM is the first LLM-based delirium prediction tool for the ICU that utilizes structured EHR data with LLMs rather than clinical notes with LLMs or traditional structured feature representations used in AI models.

Introduction Falls occur in all age groups and represent a significant public health concern. Previous studies have implemented artificial intelligence (AI), including machine learning (ML) and deep learning (DL) algorithms for fall risk prediction, but the comparative performance between models and the applicability for younger populations remains unclear. This study aims to develop and compare different ML/DL models and identify key predictive features across age groups. Methods We enrolled 1441 community-dwelling adults aged over 20 years in southern Taiwan and collected demographic, clinical, and physical performance data. Participants were categorized based on fall history. Five ML models (KNN, RF, GBDT, XGBoost, and CatBoost) and two DL (GRU, AGRU) models were trained and evaluated using accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUROC). Feature importance was interpreted using SHapley Additive exPlanations (SHAP) values in the best-performing model. Age-stratified subgroup analyses were conducted for groups aged 20-45, 46-65, and >65 years. Results The AGRU model achieved the highest accuracy (91.39%) and AUROC (0.934) in the overall group and outperformed other models across all subgroups. Feature importance analysis revealed pulse rate, living alone, systolic blood pressure, 5-times Sit-to-Stand test, and sex as major predictors of falls in the overall group. The top five predicting factors varied across age groups. Conclusion We developed a robust and interpretable DL model for identifying fall risk across different age groups. Age-specific risk factors highlight the need for tailored preventive strategies. External validation using an independent dataset demonstrated moderate generalizability. Larger and more diverse datasets for validation and integration of sequential or sensor-based data are essential for practical applications.

Underactive bladder (UAB) is a common disorder that significantly affects patients' quality of life, necessitating the exploration of underlying molecular mechanisms for more effective management. This study aims to elucidate the gene expression profiles associated with UAB by employing a combination of bioinformatics analyses and experimental validation to identify pivotal hub genes and potential therapeutic targets. We accessed the GSE122060 and GSE100219 datasets from the Gene Expression Omnibus (GEO) database to identify differentially expressed genes (DEGs), followed by functional enrichment analysis, construction of a protein-protein interaction (PPI) network, screening for hub genes and assess the accuracy and diagnostic value of the hub genes with the validation dataset GSE28242. Eighty-five DEGs were identified from the GEO dataset, with functional enrichment analysis focusing primarily on biological processes like neutrophil migration, cell chemotaxis, and bacterial defense responses. Twelve key genes were identified in the PPI network using CytoHubba and MCODE plugins. Of these, C3, CLEC4E, CSF3R, CXCR2, FPR2, and IDO1 showed significant upregulation in the validation set compared to the control group. Receiver operating characteristic (ROC) curve analysis demonstrated that these six hub genes possess high diagnostic potential, with area under the curve (AUC) values greater than 0.76. Additionally, a hub gene-transcription factor (TF) interaction network, a hub gene-TF-miRNA co-regulatory network and a hub gene-drug interaction network were constructed, revealing that five TFs and five miRNAs regulate three or more hub genes. Quantitative real-time polymerase chain reaction (qRT-PCR) validation confirmed the differential expression patterns of the 12 key genes in the PPI network in TGF-β1 treated SV-HUC-1 cells. In conclusion, our findings suggest that CLEC4E, CSF3R, CXCR2, FPR2, and IDO1 can serve as promising diagnostic biomarkers for UAB, while the identified TFs and miRNAs could unveil new avenues for drug discovery and therapeutic interventions targeting UAB progression.

The stringent regulatory requirements on nitrogen oxides (NOx) emissions from diesel compression ignition engines require accurate and reliable models for real time monitoring and diagnostics. Although traditional methods such as physical sensors and virtual engine control module (ECM) sensors provide essential data, they are only used for estimation. Ubiquitous literature primarily focuses on deterministic models with little emphasis on capturing the various uncertainties. The lack of probabilistic frameworks restricts the applicability of these models for robust diagnostics. The objective of this paper is to develop and validate a probabilistic model to predict engine-out NOx emissions using Gaussian process regression. Our approach is as follows. We employ three variants of Gaussian process models: the first with a standard radial basis function kernel with input window, the second incorporating a deep kernel using convolutional neural networks to capture temporal dependencies, and the third enriching the deep kernel with a causal graph derived via graph convolutional networks. The causal graph embeds physics knowledge into the learning process. All models are compared against a virtual ECM sensor using both quantitative and qualitative metrics. We conclude that our model provides an improvement in predictive performance when using an input window and a deep kernel structure. Even more compelling is the further enhancement achieved by the incorporation of a causal graph into the deep kernel. These findings are corroborated across different verification and validation datasets.

Objective To establish a secondary prevention screening model for predicting metabolic syndrome (MetS) based on community obstructive sleep apnea (OSA) screening, using simple and easily accessible indicators, to help early identification of high-risk individuals and improve prognosis and reduce mortality. Methods This study enrolled adults newly diagnosed with OSA from community settings in China, collecting comprehensive demographic and lifestyle data. To identify key predictive variables, least absolute shrinkage and selection operator (LASSO) regression was employed for feature selection. Nine machine learning algorithms, such as logistic regression, random forest, and support vector machine (SVM), were then used to build predictive models, with each undergoing rigorous training, hyperparameter tuning, and evaluation on stratified training, validation, and test datasets. Model performance was evaluated using multiple metrics, including the area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, specificity, F1 score, calibration curves, and clinical decision curve analysis (DCA). To improve interpretability, Shapley additive explanations (SHAP) analysis was applied to quantify each predictor's contribution to the model's output. Results Among the nine machine learning algorithms evaluated, the logistic regression model exhibited superior performance. The finalized model achieved an AUC of 0.814 on the test dataset, demonstrating strong discriminative ability. Key performance metrics included a sensitivity of 0.794, specificity of 0.647, accuracy of 0.693, and an F1 score of 0.617. Feature importance analysis highlighted body mass index (BMI), age, and gender as the most significant predictors of MetS. Calibration curves and clinical DCA further confirmed the model's reliability, showing close alignment between predicted probabilities and observed outcomes, thus affirming its clinical utility. External validation reinforced the model's robustness, yielding an AUC of 0.818, with consistent discrimination and well-calibrated predictions. Conclusion This study successfully developed a MetS prediction model based on community environment. The model relies solely on simple, easily obtainable self-reported indicators and demonstrates good predictive performance. This model, as a primary screening tool, enables residents to assess their MetS risk status independently, without relying on complex biochemical tests or the assistance of specialized medical personnel.

Identifying the mechanisms of action (MOA) driving a drug's anti-cancer efficacy is critical for its clinical success, guiding the search for its best biomarkers, indications and combinations. Yet, systematically identifying MOAs remains challenging due to drugs often engaging multiple targets with varying affinities across different cellular contexts. Addressing this challenge, we present DeepTarget, a computational tool that integrates large-scale drug and genetic knockdown viability screens with omics data to predict a drug's MOAs driving its cancer cell killing. To test its performance, we curated eight datasets of high-confidence drug-target pairs focused on cancer drugs and benchmarked DeepTarget. We show that DeepTarget outperforms recent tools in predicting drug targets and their mutation-specificity, achieving strong predictive performance across diverse validation datasets. We experimentally validate DeepTarget's predictions in two case studies: (a) Demonstrating that pyrimethamine, an anti-parasitic drug, affects cellular viability through modulation of mitochondrial function, specifically the oxidative phosphorylation pathway, and (b) Confirming that T790-mutated EGFR mediates ibrutinib response in BTK-negative solid tumors. Additionally, we demonstrate that kinase inhibitors predicted by DeepTarget to have higher target specificity show increased progression in clinical trials. We provide DeepTarget as an open-source tool ( https://github.com/CBIIT-CGBB/DeepTarget ) along with predicted target profiles for 1,500 cancer-related drugs and 33,000 unpublished natural product extracts. DeepTarget represents a significant computational advancement among target discovery methods that complements the leading structure-based methods by considering cellular context and can potentially accelerate drug development and repurposing efforts in oncology.

Background The transformation of smooth muscle cells (SMCs) into alternative phenotypes is a key process in atherosclerosis pathogenesis. Recent studies have revealed oncological parallels between atherosclerosis and cancer, such as DNA damage and oncogenic pathway activation in SMCs, but the precise molecular mechanisms remain poorly understood. This study integrates cancer gene sets using bioinformatics to identify key hub genes associated with atherosclerosis and explores their immune molecular mechanisms. Methods Datasets from the Gene Expression Omnibus (GEO) were analyzed to identify differentially expressed genes (DEGs) and module genes using Limma and WGCNA. Machine learning algorithms (SVM-RFE, LASSO regression, and random forest) were employed to identify cancer-related hub genes for early atherosclerosis diagnosis. A diagnostic model was constructed and validated. UMAP plots from single-cell RNA sequencing data were used to analyze the expression patterns of hub genes, particularly focusing on macrophage-like SMCs in SMC lineage-traced mouse models. Biomarker expression was validated in both human and mouse experiments. Results Four cancer-related hub genes (CRGs) were identified: Interferon Regulatory Factor 7 (IRF7), Formin Homology 2 Domain Containing 1 (FHOD1), Tumor Necrosis Factor (TNF), and Zinc Finger SWIM Domain Containing 3 (ZSWIM3). A diagnostic nomogram using IRF7, FHOD1, and TNF demonstrated high accuracy and reliability in both training and validation datasets. Immune microenvironment analysis revealed significant differences between atherosclerosis and control groups. Spearman correlation analysis highlighted associations between hub genes and immune cell infiltration. Single-cell RNA sequencing identified distinct SMC-derived cell clusters and phenotypic transitions, with increased expression of IRF7 and FHOD1 in macrophages potentially derived from SMCs in both human carotid plaques and mouse models. Conclusion This study integrates cancer gene sets to identify key hub genes in atherosclerosis, emphasizing its parallels with cancer. The diagnostic nomogram based on IRF7, FHOD1, and TNF provides a reliable tool for early diagnosis, while insights into SMC phenotypic switching and immune microenvironment modulation offer potential therapeutic targets.

Background: Pulmonary capillary wedge pressure (PCWP) provides an objective assessment of congestion status in heart failure (HF) patients, but its use is limited by the need for an invasive procedure, trained personnel, and specialized equipment to obtain a measurement. Cardiosense (Chicago, IL) has developed a machine learning (ML) algorithm that detects elevated PCWP non-invasively from data acquired by a chest-worn wearable device (CardioTag). We present data for a potential in-clinic point-of-care tool that improves the identification of hemodynamic congestion, with a focus on outpatient and low-acuity settings. Methods: The ePCWP System is a ML model developed to identify elevated PCWP (>18 mmHg) using non-invasive physiological biosignals from the CardioTag device, which simultaneously collects electrocardiogram, seismocardiogram, and photoplethysmogram data. Concurrent CardioTag and right-heart catheterization (RHC) data were collected prospectively in an observational study across 15 US sites from 1,116 patients undergoing standard-of-care RHC. Patients were either diagnosed with HFrEF, HFpEF, HFmrEF, or were suspected of HF before the RHC procedure. Standard of care physical examination, used to evaluate congestion status, was captured and used for comparative analysis. The training dataset contained 726 subjects and the validation dataset contained 153 subjects. Results: Five-fold cross-validation of the training dataset showed an overall accuracy of 0.79, sensitivity of 0.75 (CI: [0.69, 0.80]), and a specificity of 0.81 (CI: [0.77, 0.78]). The validation dataset showed an overall accuracy of 0.81, sensitivity of 0.76 (CI: [0.63, 0.86]), and a specificity of 0.82 (CI: [0.66, 0.89]). Figure 1 shows the overall classification performance of the ePCWP System (left) and a comparison to standard-of-care physical exam (right). Conclusion: We developed a non-invasive point-of-care tool that is capable of providing rapid, accurate assessments of congestion for patients with HF. This tool might be used to support convenient, frequent inpatient monitoring to augment discharge decisions and guide post-discharge follow-up care towards timely interventions and improvements in patient outcomes.

BackgroundSpondylitis, particularly infectious forms caused by Mycobacterium tuberculosis and Brucella species, presents significant clinical challenges due to overlapping symptoms and diagnostic difficulties. Accurate differentiation is crucial for effective treatment, necessitating advanced imaging techniques and radiomics to enhance diagnostic precision and improve patient outcomes in cases of tuberculosis spondylitis (TS) and brucella spondylitis (BS).MethodsThis retrospective cohort study included 195 patients diagnosed with TS or BS from January 2020 to December 2024 in center1, with an external validation cohort of 57 patients in center 2. Inclusion criteria consisted of relevant clinical symptoms for at least six months, positive serological tests, Magnetic Resonance Imaging (MRI) abnormalities, and complete medical records. Imaging was performed at two centers, employing standardized protocols for T1-weighted(T1WI) and T2-weighted(T2WI) and fat-suppression T2WI (FS T2WI) MRI. Region of Interest (ROI) segmentation and radiomics feature extraction were conducted using the Deepwise platform, yielding a total of 1049 Computed Tomography (CT) and 5829 MRI features. Nine predictive models were developed and validated through nested five-fold cross-validation, assessing performance metrics such as Area Under the Curve (AUC), sensitivity, and specificity. Statistical significance was set at p < 0.05.ResultsA total of 195 patients with 207 lesions were analyzed, comprising 116 patients of TS and 79 patients of BS. An external validation cohort included 57 patients with 60 lesions. Nine predictive models were developed using selected features: the CT model utilized 80 features, while T1WI, T2WI, and FS T2WI models employed 14, 33, and 32 features, respectively. Multi-modality models (CT, T1WI, T2WI and FS T2WI model) combined features from various sequences, achieving optimal performance with an AUC of 0.8136 in validation dataset. Model efficacy was validated through receiver operating characteristic (ROC) curve analysis, decision curve analyses, and calibration plots. Additionally, SHapley Additive exPlanations(SHAP) analysis was used to interpret model predictions, identifying key influential features, which are detailed in the supplementary materials.ConclusionOur research indicates that multimodal imaging-based radiomics hold significant potential for the differential diagnosis of BS and TS.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-025-11863-w.

Background: End-diastolic (ED) and end-systolic (ES) frames are critical for left ventricular (LV) volume measurements in echocardiography but show high inter- and intra-observer variability. Deep learning (DL) methods have emerged for ED/ES detection; however, these typically rely on manually annotated reference frames and often fail to generalize across different image types, such as contrast and non-contrast echocardiographic views. Methods: A fully automated novel framework was developed for localizing ED/ES frames in both contrast and non-contrast cine loops without the use of manual annotations. The process begins with the YOLO (v12) object detection DL model to identify the LV as a region of interest (ROI); alternatively, a fixed bounding box or no localization step may be used. The largest ROI is selected to crop the cine loop. Robust principal component analysis is then applied to decompose the cine into a low-rank matrix, followed by singular value decomposition to extract the top three left singular vectors ( U ). Pseudo-periodic cardiac cycles are identified in each U using the Spectral Dominance Ratio. Zero-crossings and their variances are computed, and the U with the lowest variance (with at least two cycles) is chosen to represent the cardiac cycle. A peak detection algorithm is used to identify local extrema corresponding to the ED/ES frames. Results: The method was validated using a UAB dataset (N=984; 912 contrast, 72 non-contrast) and the publicly available EchoNet-Dynamic dataset (N=10,030, non-contrast) for external validation. The YOLO model was trained exclusively on the UAB dataset (1394 images for training, 298 images for validation, and 300 images for testing). On the UAB test set, the model achieved a mean Average Precision (mAP50) of 0.994 and mAP50-95 of 0.717. Mean absolute errors (MAE) in the UAB dataset were 2.65 ± 2.95 frames (median 2) for ED and 1.58 ± 1.49 frames (median 1) for ES. In the EchoNet dataset, the MAE was 3.75 ± 4.02 frames (median 2) for ED and 2.72 ± 2.81 frames (median 2) for ES. The framework excluded 5 UAB and 115 EchoNet cases due to only one cardiac cycle in the U . Conclusion: A robust and generalizable framework has been presented for localizing ED/ES frames without reliance on manually labeled training data. This approach supports both contrast and non-contrast images and can function with or without DL-based ROI detection, offering a scalable fully automated solution for echocardiographic analysis.

Effective management of introduced species requires a clear understanding of their habitat requirements. Species distribution models (SDMs) offer a powerful tool for addressing this challenge. We applied seven modeling techniques to predict a suitable habitat for the introduced Chukar Partridge ( Alectoris chukar ), including artificial neural networks, generalized additive models, k-nearest neighbor, random forests, support vector machines, extreme gradient boosting, and a weighted ensemble approach. Using site-level data on physiography, climate, land cover, and habitat range, we modeled Chukar distributions by simulating historical introduction efforts and extrapolating predictions into surrounding areas to assess cross-regional transferability. Model performance was evaluated using independent, geographically distinct validation datasets. Our results demonstrate that machine learning-based SDMs provide accurate and transferable predictions of Chukar habitat suitability. This study highlights the value of machine learning for predicting establishment success while emphasizing the importance of incorporating species movement behavior and site fidelity into SDM frameworks. Overall, our findings contribute to advancing conservation planning, species reintroductions, and adaptive management strategies.

ABSTRACT Phytoplankton underpin marine food webs and carbon cycling, converting dissolved carbon dioxide into organic matter and exporting it to deeper layers. However, these organisms are sensitive to environmental changes that affect their growth and community structure differently, which may be represented by their taxonomic structure or cell size categories. Consequently, there is increasing interest in developing and improving satellite-based models for estimating the abundance of phytoplankton size classes (PSCs) and different taxonomic groups. Satellites can reliably estimate two key properties related to phytoplankton biomass and ocean dynamics: chlorophyll-a concentration (Chla), the primary pigment of phytoplankton, and sea surface temperature (SST), which is associated with water masses and often related to nutrient availability. In this study, we tested different approaches and developed regional models to retrieve PSCs from satellite data. The regional models were fitted to the South Brazil Bight (SBB) in the Southwestern Atlantic Ocean. The in situ training and validation datasets were obtained from oceanographic cruises conducted in the SBB during 2019–2022. We applied different model parameterisation schemes to compare SST-independent and SST-dependent models with both global and regional fits. The models were applied to both in situ data and satellite observations from Ocean and Land Colour Instrument (OLCI) sensors on board Sentinel 3A and 3B satellites, alongside the Multi-scale Ultra-high Resolution (MUR) SST product. The regional SST-dependent approach consistently outperformed alternatives across all size classes, achieving correlation coefficients (ρ) greater than 0.7, bias less than 0.14, and mean absolute error (MAE) of less than 0.36. By comparison, the regional SST-independent approach (ρ > 0.54, bias < 0.17, MAE < 0.38) and the global SST-dependent approach (ρ > 0.59, bias < 0.11, and MAE < 0.40) showed weaker performance. These results highlight the importance of regional SST-dependent models for improving PSC estimation accuracy in the SBB and similar regions where SST variability affects nutrient availability, phytoplankton biomass, and community structure.

This paper presents a TensorFlow-native implementation for automated crack detection in concrete structures, addressing the critical need for efficient and objective infrastructure monitoring. Leveraging a Convolutional Neural Network architecture with 24.8 million parameters, the model was trained on a large-scale dataset of 40,000 images, each with a 227x227 RGB resolution. The methodology, incorporating specific framework optimizations and a rigorous training configuration, achieved a remarkable overall classification accuracy of 99.375% on the validation dataset. The model demonstrated balanced performance with precision values of 0.993 and 0.994, recall values of 0.994 and 0.993, and F1-scores of 0.994 and 0.994 for both "No Crack" and "Crack" classes. This high accuracy, coupled with balanced metrics, underscores the model's effectiveness and reliability for practical applications. The proposed solution significantly enhances real-time structural health monitoring systems, mitigating the limitations of traditional manual inspections and facilitating proactive maintenance strategies for concrete infrastructure.

The accurate classification of skin cancer types is a critical task in medical diagnostics, requiring robust and reliable models to distinguish between various skin lesions. Despite advancements in deep learning, developing models that generalize well to unseen data remains a challenge. Current methodologies primarily utilize convolutional neural networks (CNNs) for image classification tasks, leveraging architectures such as ResNet, VGG, and Inception. These models have shown promise in improving classification accuracy for skin cancer detection. However, existing models often face limitations, including overfitting to the training data and difficulty in handling imbalanced datasets. This results in decreased performance on validation and test datasets, reducing their practical applicability in clinical settings. Additionally, these models may lack the fine-grained discrimination required to accurately classify a diverse range of skin lesion types. To address the limitations of traditional CNN-based approaches, we propose a novel model based on the EfficientNetV2L architecture, optimized for skin lesion classification. Our approach introduces adaptive early stopping and learning rate callbacks to enhance generalization and prevent overfitting. Trained on the ISIC dataset, the model achieved a high classification accuracy of 99.22%, demonstrating robustness across various lesion types. This work contributes a powerful, efficient, and clinically relevant solution to the field of automated skin cancer diagnosis.

Thanks to the popularity of smartphones with high-quality cameras and social media platforms, an exceptional amount of image data is generated and shared daily. This visual data can provide unprecedented insights into daily life and can be used to help answer research questions in psychology. However, the traditional methods used to analyze visual data are burdensome and are either time-intensive (e.g., content analysis) or require technical training (e.g., developing and training deep learning models). Zero-shot learning, where a pretrained model is used without any additional training, requires less technical expertise and may be a particularly attractive method for psychology researchers aiming to analyze image data. In this tutorial, we aim to provide an overview and step-by-step guide on how to analyze visual data with zero-shot learning. Specifically, we demonstrate how to use two popular models (Contrastive Language-Image Pretraining and Large Language and Vision Assistant) to identify a beverage in an image from a data set where we manipulated the type of beverage present, the setting, and the prominence of the beverage in the image (foreground, midground, background). To guide researchers through this process, we provide open code and data on GitHub and as a Google Colab notebook. Finally, we discuss how to interpret and report accuracy, how to create a validation data set, what steps need to be taken to implement the models with new data, and discuss future challenges and limitations of the method. To conclude, zero-shot learning requires less technical expertise and may be a particularly attractive method for psychology researchers aiming to analyze image data. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Background Although some prediction models have been developed to evaluate postpartum haemorrhage in caesarean delivery with complications, limited attention has focused on unplanned return to the operating room (UPROR), especially in twin pregnancies. On this note, this study seeks to investigate the risk factors for UPROR and Postpartum Haemorrhage (PPH) in twin pregnancies after caesarean section (CS) and develop a nomogram for predicting PPH. Objective This study aimed to investigate the risk factors for UPROR and PPH in twin pregnancies after CS in China and develop a nomogram for PPH prediction. Methods A multicentre retrospective cohort study was conducted. There were a total of 1198 twin pregnant women who underwent a CS at the Women’s Hospital, School of Medicine, Zhejiang University in Hangzhou, Ninghai Maternal and Child Health Hospital, Fuyang Women and Children’s Hospital in China from January 2017 to December 2021. All 1198 pregnant women were randomly divided into two groups (D for development and V for validation), one for training and one for validation by ratio 7:3. A nomogram was developed to predict PPH (blood loss ≥1000 ml) and UPROR based on the model generated by logistic regression analysis. The training cohort and the validation cohort were evaluated in PPH, and a decision curve analysis was developed. Results 16.77% (201/1198) women experienced PPH, 142 of which (142/840, 16.90%) in the training cohort and 59 (59/358, 16.48%) in the validation cohort. Seven optimal variates were obtained as predictors of PPH in twin pregnancies, including assisted reproductive technology (ART), advanced gestational weeks, placenta previa, emergency operation, total birth weight, and the use of uterotonic and anticoagulants. The AUC for the nomogram was 0.75 (95% CI, 0.71-0.79) for the training cohort, while that was 0.83 (95% CI, 0.79–0.88) for the validation dataset. 3.67% (44/1198) of women experienced UPROR for tamponade after the CS; PPH was the cause in all cases, none of whom had a hysterectomy. Six optimal variates were obtained as predictors of UPROR in twin pregnancies, including advanced maternal age, ART, parity ≥ 1, placenta previa, total amount of amniotic fluid (ml) ≥ 1500, and twin growth discordance. The AUC for the nomogram was 0.74 (95% CI, 0.66–0.82). Conclusion The novel nomogram prediction model for UPROR in twin pregnancies via cesarean section has clinical potentials, including the prevention of PPH in twin pregnancies.

Population pharmacokinetics of mycophenolic acid and Bayesian estimator in lung transplant adults recipients in the early post-transplant period.