Precision And Recall Research Articles

Development and Validation of a Machine Learning Model for Early Prediction of Sepsis Onset in Hospital Inpatients from All Departments

Background: With 11 million sepsis-related deaths worldwide, the development of tools for early prediction of sepsis onset in hospitalized patients is a global health priority. We developed a machine learning algorithm, capable of detecting the early onset of sepsis in all hospital departments. Methods: Predictors of sepsis from 45,127 patients from all departments of Valenciennes Hospital (France) were retrospectively collected for training. The binary classifier SEPSI Score for sepsis prediction was constructed using a gradient boosted trees approach, and assessed on the study dataset of 5270 patient stays, including 121 sepsis cases (2.3%). Finally, the performance of the model and its ability to detect early sepsis onset were evaluated and compared with existing sepsis scoring systems. Results: The mean positive predictive value of the SEPSI Score was 0.610 compared to 0.174 for the SOFA (Sepsis-related Organ Failure Assessment) score. The mean area under the precision–recall curve was 0.738 for SEPSI Score versus 0.174 for the most efficient score (SOFA). High sensitivity (0.845) and specificity (0.987) were also reported for SEPSI Score. The model was more accurate than all tested scores, up to 3 hours before sepsis onset. Half of sepsis cases were detected by the model at least 48 hours before their medically confirmed onset. Conclusions: The SEPSI Score model accurately predicted the early onset of sepsis, with performance exceeding existing scoring systems. It could be a valuable predictive tool in all hospital departments, allowing early management of sepsis patients. Its impact on associated morbidity-mortality needs to be further assessed.

Prognosis modelling of adverse events for post-PCI treated AMI patients based on inflammation and nutrition indexes

ObjectiveThis study aimed to evaluate the predictive performance of inflammatory and nutritional indices for adverse cardiovascular events (ACE) in patients with acute myocardial infarction (AMI) after percutaneous coronary intervention (PCI) using a machine learning (ML) algorithm.MethodsAMI patients who underwent PCI were recruited and randomly divided into non/ACE groups. Inflammatory and nutritional indices were graded according to the laboratory examination reports. Logistic Regression was used to screen for factors that were significant for ML model establishment. The performances of the algorithms were evaluated in terms of accuracy, kappa, F1, receiver operating characteristic, precision recall curve, etc.ResultsAge, LVEF%, Killip Grade, heart rate, creatinine, albumin, neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), and prognostic nutritional index (PNI) were significantly correlated with ACE by Logistic regression analysis (P < 0.05). These nine factors were employed to establish stepwise regression (SR), random forest (RF), naïve Bayes (NB), decision trees (DT), and artificial neutron network (ANN), whose performances were evaluated in terms of accuracy, kappa, F1, receiver operating characteristic, precision recall curve, etc. The accuracy of the decision tree was greater than that of other trees. The area under the curves was the highest in the ANN model compared with the other models.ConclusionANN predictive performance had an advantage over other ML algorithms based on age, LVEF%, Killip Grade, heart rate, creatinine, albumin, NLR, PLR, and PNI.

Machine learning models for preventative mobile health asthma control

Introduction Asthma attacks are set off by triggers such as pollutants from the environment, respiratory viruses, physical activity and allergens. The aim of this research is to create a machine learning model using data from mobile health technology to predict and appropriately warn a patient to avoid such triggers. Methods Lightweight machine learning models, XGBoost, Random Forest, and LightGBM were trained and tested on cleaned asthma data with a 70-30 train-test split. The models were measured on Precision Score, Accuracy Score, Recall Score, F1 Score and model speed. Results The best model, XGBoost, obtained an Accuracy score of 0.902, Recall score of 0.904, Precision score of 0.835, and F1 score of 0.860 and a model training speed of 14 s. Conclusion As proved by the XGBoost model, predicting asthma triggers can be a viable option for asthma control using machine learning. In addition, the actionable information of triggers, allows patients to make behavior changes. However there will still need to be work testing the system in a mobile health system.

Deep Learning and Automatic Detection of Pleomorphic Esophageal Lesions—A Necessary Step for Minimally Invasive Panendoscopy

Background: Capsule endoscopy (CE) improved the digestive tract assessment; yet, its reading burden is substantial. Deep-learning (DL) algorithms were developed for the detection of enteric and gastric lesions. Nonetheless, their application in the esophagus lacks evidence. The study aim was to develop a DL model for esophageal pleomorphic lesion (PL) detection. Methods: A bicentric retrospective study was conducted using 598 CE exams. Three different CE devices provided 7982 esophageal frames, including 2942 PL lesions. The data were divided into the training/validation and test groups, in a patient-split design. Three runs were conducted, each with unique patient sets. The sensitivity, specificity, accuracy, positive and negative predictive value (PPV and NPV), area under the conventional receiver operating characteristic curve (AUC-ROC), and precision–recall curve (AUC-PR) were calculated per run. The model’s diagnostic performance was assessed using the median and range values. Results: The median sensitivity, specificity, PPV, and NPV were 75.8% (63.6–82.1%), 95.8% (93.7–97.9%), 71.9% (50.0–90.1%), and 96.4% (94.2–97.6%), respectively. The median accuracy was 93.5% (91.8–93.8%). The median AUC-ROC and AUC-PR were 0.82 and 0.93. Conclusions: This study focused on the automatic detection of pleomorphic esophageal lesions, potentially enhancing the diagnostic yield of this type of lesion, compared to conventional methods. Specific esophageal DL algorithms may provide a significant contribution and bridge the gap for the implementation of minimally invasive CE-enhanced panendoscopy.

A machine learning model accurately identifies glycogen storage disease Ia patients based on plasma acylcarnitine profiles

BackgroundGlycogen storage disease (GSD) Ia is an ultra-rare inherited disorder of carbohydrate metabolism. Patients often present in the first months of life with fasting hypoketotic hypoglycemia and hepatomegaly. The diagnosis of GSD Ia relies on a combination of different biomarkers, mostly routine clinical chemical markers and subsequent genetic confirmation. However, a specific and reliable biomarker is lacking. As GSD Ia patients demonstrate altered lipid metabolism and mitochondrial fatty acid oxidation, we built a machine learning model to identify GSD Ia patients based on plasma acylcarnitine profiles.MethodsWe collected plasma acylcarnitine profiles from 3958 patients, of whom 31 have GSD Ia. Synthetic samples were generated to address the problem of class imbalance in the dataset. We built several machine learning models based on gradient-boosted trees. Our approach included hyperparameter tuning and feature selection and generalization was checked using both nested cross-validation and a held-out test set.ResultsThe binary classifier was able to correctly identify 5/6 GSD Ia patients in a held-out test set without generating significant amounts of false positive results. The best model showed excellent performance with a mean received operator curve (ROC) AUC of 0.955 and precision-recall (PR) curve AUC of 0.674 in nested CV.ConclusionsThis study demonstrates an innovative approach to applying machine learning to ultra-rare diseases by accurately identifying GSD Ia patients based on plasma free carnitine and acylcarnitine concentrations, leveraging subtle acylcarnitine abnormalities. Acylcarnitine features that were strong predictors for GSD Ia include C16-carnitine, C14OH-carnitine, total carnitine and acetylcarnitine. The model demonstrated high sensitivity and specificity, with selected parameters that were not only robust but also highly interpretable. Our approach offers potential prospect for the inclusion of GSD Ia in newborn screening. Rare diseases are underrepresented in machine learning studies and this work highlights the potential for these techniques, even in ultra-rare diseases such as GSD Ia.

Integrating machine learning and nano-QSAR models to predict the oxidative stress potential caused by single and mixed carbon nanomaterials in algal cells.

In silico methods are increasingly important in predicting the ecotoxicity of engineered nanomaterials (ENMs), encompassing both individual and mixture toxicity predictions. It is widely recognized that ENMs trigger oxidative stress effects by generating intracellular reactive oxygen species (ROS), serving as a key mechanism in their cytotoxicity studies. However, existing in silico methods still face significant challenges in predicting the oxidative stress effects induced by ENMs. Herein, we utilized laboratory-derived toxicity data and machine learning methods to develop quantitative nanostructure-activity relationship (nano-QSAR) classification and regression models, aiming to predict the oxidative stress effects of five carbon nanomaterials (fullerene, graphene, graphene oxide, single-walled carbon nanotubes, and multi-walled carbon nanotubes) and their binary mixtures on Scenedesmus obliquus cells. We constructed five nano-QSAR classification models by combining zeta potential (ζP) with the C4.5 decision tree, support vector machine, artificial neural network, naive Bayes, and K-nearest neighbor algorithms. Moreover, we constructed three classification models by integrating the features including ζP, hydrodynamic diameter (DH), and specific surface area (SSA) with the logistic regression, random forest, and Adaboost algorithms. The Accuracy, Recall, Precision and harmonic mean of Precision and Recall (F1-score) values of these models were all higher than 0.600, indicating an excellent performance in distinguishing whether CNMs have the potential to generate ROS. In addition, using the ζP, DH, and SSA descriptors, we combined decision tree regression, random forest regression, gradient boosting, and the Adaboost algorithm, and successfully constructed four nano-QSAR regression models with applicable application domains (all training and testing data points lie within 95% confidence intervals), goodness-of-fit (Rtrain2 ≥ 0.850), and robustness (cross-validation R2 ≥ 0.650) as well as predictive power (Rtest2 ≥ 0.610). The method developed would establish a fundamental basis for more precise evaluations of ecological risks posed by these materials from a mechanistic standpoint.

Modality-, feature-, and strategy-dependent organization of low-level working memory.

Previous research has shown that, when multiple similar items are maintained in working memory, recall precision declines. Less is known about how heterogeneous sets of items across different features within and between modalities impact recall precision. In two experiments, we investigated modality (Experiment 1, n = 79) and feature-specific (Experiment 2, n = 154) load effects on working memory performance. First, we found a cross-modal advantage in continuous recall: Orientations that are memorized together with a pitch are recalled more precisely than orientations that are memorized together with another orientation. The results of our second experiment, however, suggest that this is not a pure effect of sensory modality but rather a feature-dependent effect. We combined orientations, pitches, and colors in pairs. We found that memorizing orientations together with a color benefits orientation recall to a similar extent as the cross-modal benefit. To investigate this absence of interference between orientations and colors held in working memory, we analyzed subjective reports of strategies used for the different features. We found that, although orientations and pitches rely almost exclusively on sensory strategies, colors are memorized not only visually but also with abstract and verbal strategies. Thus, although color stimuli are also visually presented, they might be represented by independent neural circuits. Our results suggest that working memory storage is organized in a modality-, feature-, and strategy-dependent way.

Predicting the decline of physical function among the older adults in China: A Cohort study based on China Longitudinal Health and Longevity Survey (CLHLS)

BackgroundAs the arrival of healthy aging, maintaining physical function (PF) in older adults is crucial for their health, so it is necessary to detect the decline of PF among them and take intervention measures. MethodsWe construct eight machine learning models to predict declines of PF in this study. The performance of the models was tested by Area Under Curve (AUC), sensitivity, specificity, accuracy, precision-recall (PR) curve and calibration degree. Decision Curve Analysis (DCA) curve was used to evaluate their discrimination ability and clinical practicability. ResultsThere were 2,017 participants in this study. We found that logistic regression models performed the best, with AUC, sensitivity, specificity and accuracy of 0.803, 0.698, 0.761 and 0.744 respectively, and its DCA curve, calibration degree and PR curve also performed well. ConclusionLogistic regression can be used as optimal model to identify the risk of PF decline among older adults in China.

Clinical Application of a Big Data Machine Learning Analysis Model for Osteoporotic Fracture Risk Assessment Built on Multicenter Clinical Data in Qingdao City.

Osteoporotic fractures (OPF) pose a public health issue, imposing significant burdens on families and societies worldwide. Currently, there is a lack of comprehensive and validated risk assessment models for OPF. This study aims to develop a model to assess and predict the risk of OPF in Qingdao City, China. From January 2021 to January 2023, we recruited 84 osteoporotic patients diagnosed with fractures from Qingdao University Affiliated Hospital, Qingdao Municipal Hospital, Qingdao Hiser Hospital Affiliated of Qingdao University, and Qingdao Central Hospital as the experimental group. In addition, 112 osteoporotic patients without fractures were recruited as the control group. In this study, we employed seven machine learning models, namely Adaboost, random forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Logistic Regression (LR), Naive Bayes (NB), and Gradient Boosting Decision Trees (GBDT), to analyze the risk factors influencing the occurrence of OPF. Next, we plotted receiver operating characteristic (ROC), Precision-Recall (PR), and calibration curves to evaluate the predictive values of the different risk assessment models for OPF. Among the seven models built based on the training set data, the Adaboost model showed area under the curve (AUC), sensitivity, and specificity values close to 1, indicating the best classification performance. In the test set, the AUC values for the RF, SVM, LR, KNN, NB, AdaBoost, and GBDT models were 0.936, 0.905, 0.88, 0.93, 0.862, 0.939, and 0.859, respectively (p < 0.001). All sensitivity and specificity values for these models were higher than 0.8, with sensitivity and specificity values of the Adaboost model closest to 1. Additionally, six models had an area under the Precision-Recall curve (prAUC) values higher than 0.9, except KNN at 0.284 (p < 0.001). The calibration curves of the seven models did not significantly deviate from the ideal curve, indicating acceptable discriminative ability and predictive performance of the predictive model. All results showed that trabecular bone score (TBS) was the most important variable affecting the model, followed by osteocalcin (OST) and hunchback. Given the various clinical data from patients with OPF, we assessed and demonstrated the good predictive performance of our risk predictive models. This model will enable us to take timely intervention measures to reduce the incidence of OPF and improve patient prognosis.

Deformable multi-level feature network applied to nucleus segmentation

IntroductionThe nucleus plays a crucial role in medical diagnosis, and accurate nucleus segmentation is essential for disease assessment. However, existing methods have limitations in handling the diversity of nuclei and differences in staining conditions, restricting their practical application.MethodsA novel deformable multi-level feature network (DMFNet) is proposed for nucleus segmentation. This network is based on convolutional neural network and divides feature processing and mask generation into two levels. At the feature level, deformable convolution is used to enhance feature extraction ability, and multi-scale features are integrated through a balanced feature pyramid. At the mask level, a one-stage framework is adopted to directly perform instance segmentation based on location.ResultsExperimental results on the MoNuSeg 2018 dataset show that the mean average precision (mAP) and mean average recall (mAR) of DMFNet reach 37.8% and 47.4% respectively, outperforming many current advanced methods. Ablation experiments verified the effectiveness of each module of the network.DiscussionDMFNet provides an effective solution for nucleus segmentation and has important application value in medical image analysis.

The use of cloud based machine learning to predict outcome in intracerebral haemorrhage without explicit programming expertise

Machine Learning (ML) techniques require novel computer programming skills along with clinical domain knowledge to produce a useful model. We demonstrate the use of a cloud-based ML tool that does not require any programming expertise to develop, validate and deploy a prognostic model for Intracerebral Haemorrhage (ICH). The data of patients admitted with Spontaneous Intracerebral haemorrhage from January 2015 to December 2019 was accessed from our prospectively maintained hospital stroke registry. 80% of the dataset was used for training, 10% for validation, and 10% for testing. Seventeen input variables were used to predict the dichotomized outcomes (Good outcome mRS 0–3/ Bad outcome mRS 4–6), using machine learning (ML) and logistic regression (LR) models. The two different approaches were evaluated using Area Under the Curve (AUC) for Receiver Operating Characteristic (ROC), Precision recall and accuracy. Our data set comprised of a cohort of 1000 patients. The data was split 8:1 for training & testing respectively. The AUC ROC of the ML model was 0.86 with an accuracy of 75.7%. With LR AUC ROC was 0.74 with an accuracy of 73.8%. Feature importance chart showed that Glasgow coma score (GCS) at presentation had the highest relative importance, followed by hematoma volume and age in both approaches. Machine learning models perform better when compared to logistic regression. Models can be developed by clinicians possessing domain expertise and no programming experience using cloud based tools. The models so developed lend themselves to be incorporated into clinical workflow.

Improving the Vehicle Small Object Detection Algorithm of Yolov5

To address the problems of low accuracy and poor robustness in vehicle small object detection for autonomous driving tasks, this study aims to propose an improved vehicle small object detection algorithm model based on YOLOv5. Firstly, some convolutions in the backbone network are replaced with receptive field attention convolutions, and the weights of the convolution kernels are dynamically assigned based on the importance of image features to ensure the extraction of important features. Secondly, adding a channel attention mechanism to the backbone network enhances the attention to small target features. Finally, the Focal-EIoU loss function is introduced to increase the attention on high-quality samples in the regression stage of object detection boxes. When the model is applied to the small object test set of the KITTI dataset, the precision rate, recall rate and mean average precision are 88.5%, 82.8%, and 84.9%, respectively, and the frame processing rate reaches 87.83FPS.

Electrocardiogram-Based Deep Learningto Predict Mortality in RepairedTetralogy of Fallot.

Artificial intelligence-enhanced electrocardiogram (AI-ECG) analysis shows promise to predict mortality in adults with acquired cardiovascular diseases. However, its application to the growing repaired tetralogy of Fallot (rTOF) population remains unexplored. This study aimed to develop and externally validate an AI-ECG model to predict 5-year mortality in rTOF. A convolutional neural network was trained on electrocardiograms (ECGs) obtained at Boston Children's Hospital and tested on Boston (internal testing) and Toronto (external validation) INDICATOR (International Multicenter TOF Registry) cohorts to predict 5-year mortality. Model performance was evaluated on single ECGs per patient using area under the receiver operating (AUROC) and precision recall (AUPRC) curves. The internal testing and external validation cohorts comprised of 1,054 patients (13,077 ECGs at median age 17.8 [Q1-Q3: 7.9-30.5] years; 54% male; 6.1% mortality) and 335 patients (5,014 ECGs at median age 38.3 [Q1-Q3: 29.1-48.7] years; 57% male; 8.4% mortality), respectively. Model performance was similar during internal testing (AUROC 0.83, AUPRC 0.18) and external validation (AUROC 0.81, AUPRC 0.21). AI-ECG performed similarly to the biventricular global function index (an imaging biomarker) and outperformed QRS duration. AI-ECG 5-year mortality prediction, but not QRS duration, was a significant independent predictor when added into a Cox regression model with biventricular global function index to predict shorter time-to-death on internal and external cohorts. Saliency mapping identified QRS fragmentation, wide and low amplitude QRS complexes, and flattened T waves as high-risk features. This externally validated AI-ECG model may complement imaging biomarkers to improve risk stratification in patients with rTOF.

Short communication: The importance of random effects in detecting purging of inbreeding depression: A model comparison in Pannon White rabbits.

Inbreeding depression (ID) is a well-documented phenomenon associated with reduced fitness and possible extinction. However, ID can be mitigated or even eliminated through the interplay of inbreeding and selection, a process known as purging. The aim of this study was to compare the predictive power of two commonly used approaches in models with and without random dam effects to detect purging (full and reduced models). Specifically, we compared the full and reduced models based on the Kalinowski ancestral inbreeding coefficient (KAL) with the full and reduced inbreeding-purging (IP) models based on the purged inbreeding coefficient. Our analysis utilised kindling records from 1379 Pannon rabbits born between 1992 and 1997. We modelled the effects of dam and litter inbreeding on kit survival at birth (zero/one), an important fitness trait, along with the effects of parity and the effects of kindling season. Random dam effects were only considered in the full KAL and IP models. We assessed the classification abilities (predictive power) of these models using Receiver Operating Characteristic (ROC) and Precision-Recall (PR) curves, where larger areas under the curve (AUC) indicate better classification performance. The full KAL model (AUC-ROC = 0.8156, AUC-PR = 0.9861) and the full IP model (AUC-ROC = 0.8152, AUC-PR = 0.9860), both of which include random dam effects, demonstrated high predictive power based on both methods. In contrast, the reduced KAL model (AUC-ROC = 0.5730, AUC-PR = 0.9553) and the reduced IP model (AUC-ROC = 0.5686, AUC-PR = 0.9546), which did not include random dam effects, had significantly lower AUC values. Based on our empirical results using the receiver ROC and PR curves, it could be concluded that the inclusion of random dam effects significantly increased the predictive power of the KAL and IP approaches. This finding has high importance as the inclusion of 'polygenic' random effects is not standard - and possibly never applied - in the IP models, unlike in KAL models where their use is more common.

WE-DeepLabV3+: A lightweight segmentation model for Panax notoginseng leaf diseases

Panax notoginseng plays an important role in traditional Chinese medicine. However, diseases pose a significant threat to the quality and yield of P. notoginseng. The main challenge related to the identification of P. notoginseng leaf diseases is how to achieve good performance in the case of small diseased spots on P. notoginseng leaf, overlapping edges of diseased leaf and the difficulty in mobile deployment. A lightweight semantic segmentation model Window Efficient-DeepLabv3+ is proposed for segmentation and quantification of P. notoginseng leaf diseases. We propose the Window Attention-ASPP module and present a hierarchical stacking of features, which improves model accuracy for minor target lesions while reducing parameters. In addition, a lightweight backbone network MobileNetV2 is utilized as a feature extraction module. The decoding stage introduces the Efficient Channel Attention module, which effectively improves the accuracy of the segmentation of the blade contour. Experimental results yielded the Mean Intersection Over Union, Mean Precision, and Mean Recall metrics of WE-DeepLabV3+ network to be 82.0 %, 87.6 %, and 92.4 % respectively, outperforming other segmentation models such as UNet, PSPNet, CaraNet, SegNet, and BiSeNetV2. Moreover, the number of parameters has been reduced by 90.6 % with only 5.1 M parameters. Finally, the method is used to quantify the disease of P. notoginseng leaf, the error is only 1.15 % and 0.82 %, which proves that it can quantify the disease severity accurately. Thus, the proposed method holds great significance for raising the yield and quality of P. notoginseng, also providing reliable guidance for precise fertilization and drug control.

Development and validation of a deep learning algorithm for the classification of the level of surgical difficulty in impacted mandibular third molar surgery

The aim of this study was to develop and validate a convolutional neural network (CNN) algorithm for the detection of impacted mandibular third molars in panoramic radiographs and the classification of the surgical extraction difficulty level. A dataset of 1730 panoramic radiographs was collected; 1300 images were allocated to training and 430 to testing. The performance of the model was evaluated using the confusion matrix for multiclass classification, and the actual scores were compared to those of two human experts. The area under the precision–recall curve of the YOLOv5 model ranged from 72% to 89% for the variables in the surgical difficulty index (level of occlusion, available retromolar space, impaction angle, number of roots, root curvature, and relationship to the second molar). The area under the receiver operating characteristic curve showed promising results of the YOLOv5 model for classifying third molars into three surgical difficulty levels (micro-average AUC 87%). Furthermore, the algorithm scores demonstrated good agreement with the human experts. In conclusion, the YOLOv5 model has the potential to accurately detect and classify the position of mandibular third molars, with high performance for every criterion in radiographic images. The proposed model could serve as an aid in improving clinician performance and could be integrated into a screening system.

The effect of resampling techniques on the performances of machine learning clinical risk prediction models in the setting of severe class imbalance: development and internal validation in a retrospective cohort

PurposeThe availability of population datasets and machine learning techniques heralded a new era of sophisticated prediction models involving a large number of routinely collected variables. However, severe class imbalance in clinical datasets is a major challenge. The aim of this study is to investigate the impact of commonly-used resampling techniques in combination with commonly-used machine learning algorithms in a clinical dataset, to determine whether combination(s) of these approaches improve upon the original multivariable logistic regression with no resampling.MethodsWe previously developed and internally validated a multivariable logistic regression 30-day mortality prediction model in 30,619 patients using preoperative and intraoperative features.Using the same dataset, we systematically evaluated and compared model performances after application of resampling techniques [random under-sampling, near miss under-sampling, random oversampling, and synthetic minority oversampling (SMOTE)] in combination with machine learning algorithms (logistic regression, elastic net, decision trees, random forest, and extreme gradient boosting).ResultsWe found that in the setting of severe class imbalance, the impact of resampling techniques on model performance varied by the machine learning algorithm and the evaluation metric. Existing resampling techniques did not meaningfully improve area under receiving operating curve (AUROC). The area under the precision recall curve (AUPRC) was only increased by random under-sampling and SMOTE for decision trees, and oversampling and SMOTE for extreme gradient boosting. Importantly, some combinations of algorithm and resampling technique decreased AUROC and AUPRC compared to no resampling.ConclusionExisting resampling techniques had a variable impact on models, depending on the algorithms and the evaluation metrics. Future research is needed to improve predictive performances in the setting of severe class imbalance.

Risk prediction model for psoriatic arthritis: NHANES data and multi-algorithm approach.

To develop a simplified predictive model for identifying psoriatic arthritis (PsA) in psoriasis patients. Data from the National Health and Nutrition Examination Survey (NHANES) database were analyzed, including patients with psoriasis without arthritis (PsC) or PsA. The least absolute shrinkage and selection operator, Boruta algorithm, random forest, and stepwise regression were employed to select key variables from 38 potential predictors. Logistic regression models were constructed for each combination of selected variables and evaluated using receiver operating characteristic (ROC) curves, precision-recall (PR) curves, calibration plots, Brier scores, and decision curve analysis (DCA). The study included 587 patients with psoriasis, 238 of whom had PsA. The variable combinations proposed by the Boruta algorithm exhibited the best overall performance. Key predictors in the Borutamodel included age, fasting glucose, education level, thyroid disease, hypertension, and chronic bronchitis. This model achieved area under the curve (AUC) of 0.781 (95% CI, 0.737-0.826) for the training set and 0.780 (95% CI, 0.712-0.848) for the testing set in the ROC curve analyses. The AUC values in the PR curves were 0.687 (95% CI, 0.611-0.757) and 0.653 (95% CI, 0.535-0.770), respectively. The Brier scores of 0.186 and 0.191 for the testing and training sets indicated a good fit, further supported by the calibration curves. DCA showed a net clinical benefit for decision thresholds ranging from 0.2 to 0.8 in both datasets. The Borutamodel represents a promising tool for early risk assessment of PsA. Key Points • National Database Utilization: This study leverages the NHANES database to predict psoriatic arthritis risk, addressing previous limitations tied to regional or ethnic constraints. • Comprehensive Variable Analyses: The research examines 38 variables, including demographics, health conditions, laboratory results, and lifestyle factors, using four distinct screening methods and thorough evaluations of model performance. • Innovative Risk Model: The study introduces a novel risk assessment model that integrates age, fasting glucose, education, and comorbidities including hypertension, thyroid disease, and chronic bronchitis, thus moving beyond traditional focus on skin lesions and joint symptoms.

Real-time Mask Detection with YOLOv3 and Ensemble Residual Networks

The proposed system introduced an advanced face mask detections framework leveraging Adaptively YOLOv3 with Hybrid Atrous Convolution-depend Face Detector, optimized through an Ensembles residual Learning Network. This approach aims to enhance the accuracy level and efficacy of face mask detection in real-time situation. By integrating of the adaptive capability of YOLOv3 on the spatial benefits of atrous convolution defined, the system effectively captured varying of facial features and masks placements. The ensemble residual learning further refined detection by amalgamating of multiple learning models, by reducing false positives values and negatives. This hybrid model is meticulous optimized to maintaining of higher detection precision rate and recall, for ensuring robust performance even over diverse and complex conditions. The system’s efficacy will underscored by reduces the inference time and enhanced processing throughput rate, by making it highly suitable for large-scale deployment on public safety applications. By rigorously testing phase, the proposed system demonstrated significant improvements in detection accuracy rate, inference speed, and overall reliability, positioning it will be a pivotal tool in enforcing healthy safety protocols on various public and private settings.

Decision model of public opinion risk in campus social network based on hybrid dynamic deletion and shortest path algorithm

Aiming at the problem that traditional network public opinion monitoring and searching are inefficient and can easily cause resource waste, the study firstly, through the dynamic deletion-shortest path algorithm to classify network text, and on this basis, innovatively constructs a text sentiment classification model based on the variant of convolutional neural network and recurrent neural network, and secondly, uses attention mechanism to classify the model. improvement of the classification model by using the attention mechanism. The research results show that the average precision rate, recall rate, and F-value of the dynamic deletion-shortest path algorithm are 97.30%, 79.55%, and 87.53%, and the classification speed is 397 KB/s, which is better than the traditional shortest path algorithm. In the classification effect measurement of long text, the accuracy and F-value of the recurrent neural network variant model are above 84%, and the accuracy of the text sentiment classification model with the introduction of the attention mechanism is improved by 3.89% compared to the pre-improvement period. In summary, the dynamic deletion-shortest path algorithm proposed in the study and the sentiment classification model with the introduction of the attention mechanism have superior performance and can provide certain application value for campus social network opinion risk decision-making.