Accuracy Of Model Research Articles

SEGURANÇA RESIDENCIAL INTELIGENTE: DETECÇÃO DE COMPORTAMENTO SUSPEITO POR MEIO DE ANÁLISE DE VÍDEO

This work presents the development of a computational solution for intelligent home monitoring, designed to detect people, analyze their behaviors and issue alerts for abnormal activities. YOLOv8-Pose technology was used to detect people's key points and a modified MLP (Multilayer Perceptron) neural network with LSTM (Long Short-Term Memory) characteristics was used to classify behavior. The training dataset was created by recording videos using a smartphone strategically positioned to capture the frontal area of a residence. This resulted in 154 video clips where the actor exhibited normal behaviors by walking along the scene and 154 videos where the actor performed actions that could lead to home invasion, such as attempting to climb the fence or break the garage gate. To carry out the experiments, videos with multiple actors and different behaviors were analyzed to evaluate the effectiveness of the developed methodology. The results indicated a high success rate in detecting normal behaviors, although challenges remain in scenarios with partial occlusion. The model's accuracy in classifying behaviors was 91.6%, reflecting its effectiveness in correctly identifying normal and abnormal activities.

Artificial intelligence-based modeling of compressive strength of slurry infiltrated fiber concrete

Purpose The purpose of this paper is to develop a reliable model that would predict the compressive strength of slurry infiltrated fiber concrete (SIFCON) modified with various supplementary cementitious materials (SCMs) using artificial intelligence approach. Design/methodology/approach This study engaged the artificial intelligence to predict the compressive strength of SIFCON through deep neural networks (DNN), artificial neural networks, linear regression, regression trees, support vector machine, ensemble trees, Gaussian process regression and neural networks (NN). A thorough data set of 387 samples was gathered from relevant studies. Eleven variables (cement, silica fume, fly ash, metakaolin, steel slag, fine aggregates, steel fiber fraction, steel fiber aspect ratio, superplasticizer, water to binder ratio and curing ages) were taken as input to predict the output (compressive strength). The accuracy and reliability of the developed models were assessed using a variety of performance metrics. Findings The results showed that the DNN (11-20-20-20-1) predicted the compressive strength of SIFCON better than the other algorithms with R2 and mean square error yielding 95.89% and 8.07. The sensitivity analysis revealed that steel fiber, cement, silica fume, steel fiber aspect ratio and superplasticizer are the most vital variables in estimating the compressive strength of SIFCON. Steel fiber contributed the highest value to the SIFCON’s compressive strength with 16.90% impact. Originality/value This is a novel technique in predicting the compressive strength of SIFCON optimized with different SCMs using supervised learning algorithms, improving its quality and performance.

Sound insulation prediction and optimization of wooden support structure for high-speed train floor based on machine learning

In order to improve the sound insulation performance of high-speed train floors, this study first obtained the necessary data for model training based on the reverberation test method, and then conducted data sorting and feature selection. Next, the maximum mutual information minimum redundancy (mRMR) feature selection algorithm was used to calculate the selected features and screen out a subset of significant features. Subsequently, the decision tree, BP neural network, and support vector machine regression (SVR) methods were applied in sequence, and the standardized feature data were used for the high-speed train floor under the same evaluation criteria of the mean square error (MSE) and coefficient of determination (R2). We conducted training and validation of the sound insulation prediction models for timber-framed support structures. The prediction accuracy of the trained model was compared and evaluated with the finite element statistical energy analysis (FE-SEA) prediction model. Finally, the SVR model was used to optimize the design under constraint conditions. The research results show that based on the research object, sample library, and model training in this article, compared with the FE-SEA model, the prediction error of the SVR model is only 0.3 dB, showing better performance. In engineering practice, the SVR model can effectively optimize the wooden support structure in the floor under certain constraints, and it predicts that the weighted sound insulation of the entire floor is 50.45 dB, which has important engineering application value.

Efficient Screening in Obstructive Sleep Apnea Using Sequential Machine Learning Models, Questionnaires, and Pulse Oximetry Signals: Mixed Methods Study.

Obstructive sleep apnea (OSA) is a prevalent sleep disorder characterized by frequent pauses or shallow breathing during sleep. Polysomnography, the gold standard for OSA assessment, is time consuming and labor intensive, thus limiting diagnostic efficiency. This study aims to develop 2 sequential machine learning models to efficiently screen and differentiate OSA. We used 2 datasets comprising 8444 cases from the Sleep Heart Health Study (SHHS) and 1229 cases from Taipei Veterans General Hospital (TVGH). The Questionnaire Model (Model-Questionnaire) was designed to distinguish OSA from primary insomnia using demographic information and Pittsburgh Sleep Quality Index questionnaires, while the Saturation Model (Model-Saturation) categorized OSA severity based on multiple blood oxygen saturation parameters. The performance of the sequential machine learning models in screening and assessing the severity of OSA was evaluated using an independent test set derived from TVGH. The Model-Questionnaire achieved an F1-score of 0.86, incorporating demographic data and the Pittsburgh Sleep Quality Index. Model-Saturation training by the SHHS dataset displayed an F1-score of 0.82 when using the power spectrum of blood oxygen saturation signals and reached the highest F1-score of 0.85 when considering all saturation-related parameters. Model-saturation training by the TVGH dataset displayed an F1-score of 0.82. The independent test set showed stable results for Model-Questionnaire and Model-Saturation training by the TVGH dataset, but with a slightly decreased F1-score (0.78) in Model-Saturation training by the SHHS dataset. Despite reduced model accuracy across different datasets, precision remained at 0.89 for screening moderate to severe OSA. Although a composite model using multiple saturation parameters exhibits higher accuracy, optimizing this model by identifying key factors is essential. Both models demonstrated adequate at-home screening capabilities for sleep disorders, particularly for patients unsuitable for in-laboratory sleep studies.

Understanding Trust and Reliance Development in AI Advice: Assessing Model Accuracy, Model Explanations, and Experiences from Previous Interactions

People are increasingly interacting with AI systems, but successful interactions depend on people trusting these systems only when appropriate. Since neither gaining trust in AI advice nor restoring lost trust after AI mistakes is warranted, we seek to better understand the development of trust and reliance in sequential human-AI interaction scenarios. In a 2 \({\times}\) 2 between-subject simulated AI experiment, we tested how model accuracy (high vs. low) and explanation type (human-like vs. abstract) affect trust and reliance on AI advice for repeated interactions. In the experiment, participants estimated jail times for 20 criminal law cases, first without and then with AI advice. Our results show that trust and reliance are significantly higher for high model accuracy. In addition, reliance does not decline over the trial sequence, and trust increases significantly with high accuracy. Human-like (vs. abstract) explanations only increased reliance on the high-accuracy condition. We furthermore tested the extent to which trust and reliance in a trial round can be explained by trust and reliance experiences from prior rounds. We find that trust assessments in prior trials correlate with trust in subsequent ones. We also find that the cumulative trust experience of a person in all earlier trial rounds correlates with trust in subsequent ones. Furthermore, we find that the two trust measures, trust and reliance, impact each other: prior trust beliefs not only influence subsequent trust beliefs but likewise influence subsequent reliance behavior, and vice versa. Executing a replication study yielded comparable results to our original study, thereby enhancing the validity of our findings.

Development and validation of a risk score to predict adverse birth outcomes using maternal characteristics in northwest Ethiopia: a retrospective follow-up study

BackgroundAdverse birth outcomes are unfavorable outcomes of pregnancy that are particularly common in low- and middle-income countries. At least one ultrasound is recommended to predict adverse birth outcomes in early pregnancy. However, in low-income countries, imaging equipment and trained manpower are scarce. According to our search of the literature, there is no validated risk prediction model for predicting adverse birth outcomes in Ethiopia. Hence, we developed and validated a model and risk score to predict adverse birth outcomes using maternal characteristics during pregnancy for use in resource-limited settings.MethodsA retrospective follow-up study was conducted from 1 January 2016 to 31 May 2021, and a total of 910 pregnant women were included in this study. Participants were selected using a simple random sampling technique. Stepwise, backward multivariable analysis was conducted. The model's accuracy was assessed using density plots, discrimination, and calibration. The developed model was assessed for internal validity using bootstrapping techniques and evaluated for clinical utility using decision curve analysis across various threshold probabilities.ResultsPremature rupture of Membrane, number of fetuses, residence, pregnancy-induced hypertension, antepartum hemorrhage, hemoglobin level, and labor onset remained in the final multivariable prediction model. The area under the curve of the model was 0.77 (95% confidence interval: 0.73–0.812). The developed risk prediction model had a good performance and was well-calibrated and valid. The decision curve analysis indicated the model provides a higher net benefit across the ranges of threshold probabilities.ConclusionIn general, this study showed the possibility of predicting adverse birth outcomes using maternal characteristics during pregnancy. The risk prediction model using a simplified risk score helps identify high-risk pregnant women for specific interventions. A feasible score would reduce neonatal morbidity and mortality and improve maternal and child health in low-resource settings.

Multi-Objective Optimal Scheduling for Microgrids—Improved Goose Algorithm

Against the background of the dual challenges of global energy demand growth and environmental protection, this paper focuses on the study of microgrid optimization and scheduling technology and constructs a smart microgrid system integrating energy production, storage, conversion, and distribution. By integrating high-precision load forecasting, dynamic power allocation algorithms, and intelligent control technologies, a microgrid scheduling model is proposed. This model simultaneously considers environmental protection and economic efficiency, aiming to achieve the optimal allocation of energy resources and maintain a dynamic balance between supply and demand. The goose optimization algorithm (GO) is innovatively introduced and improved, enhancing the algorithm’s ability to use global search and local fine search in complex optimization problems by simulating the social aggregation of the goose flock, the adaptive monitoring mechanism, and the improved algorithm, which effectively avoids the problem of the local optimal solution. Meanwhile, the combination of super-Latin stereo sampling and the K-means clustering algorithm improves the data processing efficiency and model accuracy. The results demonstrate that the proposed model and algorithm effectively reduce the operating costs of microgrids and mitigate environmental pollution. Using the improved goose algorithm (IGO), the combined operating and environmental costs are reduced by 16.15%, confirming the model’s effectiveness and superiority.

Impact of Petty Tyranny on Employee Turnover Intentions: The Mediating Roles of Toxic Workplace Environment and Emotional Exhaustion in Academia

Based on social exchange theory, social psychology theories, and despotic leadership theory, this study explored the impact of petty tyranny on employee turnover intentions. Specifically, the authors examined the mediating effect of toxic workplace environments through emotional exhaustion on this relationship among academicians. The authors surveyed 421 employees using a five-point Likert scale across six universities in Lahore, Pakistan and employed a time-lag research design. Partial least squares structural equation modeling (PLS-SEM) and artificial neural network (ANN) analyses, including performance comparisons of various algorithms, were used to test the relationships among the variables. The analysis results of the study suggested that petty tyranny does not significantly and directly contribute to employee turnover intentions; however, this relationship is positively and significantly mediated by toxic workplace environments and emotional exhaustion. The results indicated that toxic workplace environments and emotional exhaustion also have a direct effect on employee turnover intentions. A serial full mediation was found between petty tyranny and turnover intentions, mediated through a toxic workplace environment and emotional exhaustion. Similarly, results from the performance comparison of various algorithms reveal trade-offs between precision, recall, and processing time, with ZeroR and Stacking REP Tree emerging as the most effective in terms of overall model accuracy. This study contributes to the literature by examining petty tyranny, workplace environment, and emotional exhaustion, highlighting the need to address tyrannical behavior to improve employee retention in academic organizations. Our study offers valuable practical implications, emphasizing addressing these issues to reduce turnover in academic organizations. Our study also provides recommendations for future research directions.

DGA Domain Detection Based on Transformer and Rapid Selective Kernel Network

Botnets pose a significant challenge in network security by leveraging Domain Generation Algorithms (DGA) to evade traditional security measures. Extracting DGA domain samples is inherently complex, and the current DGA detection models often struggle to capture domain features effectively when facing limited training data. This limitation results in suboptimal detection performance and an imbalance between model accuracy and complexity. To address these challenges, this paper introduces a novel multi-scale feature fusion model that integrates the Transformer architecture with the Rapid Selective Kernel Network (R-SKNet). The proposed model employs the Transformer’s encoder to couple the single-domain character elements with the multiple types of relationships within the global domain block. This paper proposes integrating R-SKNet into DGA detection and developing an efficient channel attention (ECA) module. By enhancing the branch information guidance in the SKNet architecture, the approach achieves adaptive receptive field selection, multi-scale feature capture, and lightweight yet efficient multi-scale convolution. Moreover, the improved Feature Pyramid Network (FPN) architecture, termed EFAM, is utilized to adjust channel weights for outputs at different stages of the backbone network, leading to achieving multi-scale feature fusion. Experimental results demonstrate that, in tasks with limited training samples, the proposed method achieves lower computational complexity and higher detection accuracy compared to mainstream detection models.

Technical Code Analysis of Geomagnetic Flaw Detection of Suppression Rigging Defect Signal Based on Convolutional Neural Network

In this paper, technical code analysis and recognition of the defect signal of the suppression rigging based on a convolutional neural network are carried out given the difficulty and low recognition rate of the defect detection and recognition of the suppression rigging. Firstly, the magnetic induction signal of the suppression rigging defects is collected using CM-801 (Anshan, China), Kalman filtering is used to screen and pre-process the collected data, and the noise reduction data are presented in the form of a cloud image. The pressed rigging defect data set is constructed, and the region of broken wire defect and stress in the image is calibrated. The single-stage object detection algorithm YOLOv5 (You Only Look Once) based on convolutional neural network model calculation is used, the scale detection layer and positioning loss function of the YOLOv5 algorithm are improved and optimized, and the improved YOLOv5 algorithm is used for experiments. The experimental results show that the detection accuracy of the convolution neural network model can reach 97.1%, which can effectively identify the defect signal of the suppressed rigging.

A Deep Learning Network for Accurate Retinal Multidisease Diagnosis Using Multiview Fusion of En Face and B-Scan Images: A Multicenter Study.

Accurate diagnosis of retinal disease based on optical coherence tomography (OCT) requires scrutiny of both B-scan and en face images. The aim of this study was to investigate the effectiveness of fusing en face and B-scan images for better diagnostic performance of deep learning models. A multiview fusion network (MVFN) with a decision fusion module to integrate fast-axis and slow-axis B-scans and en face information was proposed and compared with five state-of-the-art methods: a model using B-scans, a model using en face imaging, a model using three-dimensional volume, and two other relevant methods. They were evaluated using the OCTA-500 public dataset and a private multicenter dataset with 2330 cases; cases from the first center were used for training and cases from the second center were used for external validation. Performance was assessed by averaged area under the curve (AUC), accuracy, sensitivity, specificity, and precision. In the private external test set, our MVFN achieved the highest AUC of 0.994, significantly outperforming the other models (P < 0.01). Similarly, for the OCTA-500 public dataset, our proposed method also outperformed the other methods with the highest AUC of 0.976, further demonstrating its effectiveness. Typical cases were demonstrated using activation heatmaps to illustrate the synergy of combining en face and B-scan images. The fusion of en face and B-scan information is an effective strategy for improving the diagnostic accuracy of deep learning models. Multiview fusion models combining B-scan and en face images demonstrate great potential in improving AI performance for retina disease diagnosis.

A Rice Leaf Area Index Monitoring Method Based on the Fusion of Data from RGB Camera and Multi-Spectral Camera on an Inspection Robot

Automated monitoring of the rice leaf area index (LAI) using near-ground sensing platforms, such as inspection robots, is essential for modern rice precision management. These robots are equipped with various complementary sensors, where specific sensor capabilities partially overlap to provide redundancy and enhanced reliability. Thus, leveraging multi-sensor fusion technology to improve the accuracy of LAI monitoring has become a crucial research focus. This study presents a rice LAI monitoring model based on the fused data from RGB and multi-spectral cameras with an ensemble learning algorithm. The results indicate that the estimation accuracy of the rice LAI monitoring model is effectively improved by fusing the vegetation index and textures from RGB and multi-spectral sensors. The model based on the LightGBM regression algorithm has the most improvement in accuracy, with a coefficient of determination (R2) of 0.892, a root mean square error (RMSE) of 0.270, and a mean absolute error (MAE) of 0.160. Furthermore, the accuracy of LAI estimation in the jointing stage is higher than in the heading stage. At the jointing stage, both LightGBM based on optimal RGB image features and Random Forest based on fused features achieved an R2 of 0.95. This study provides a technical reference for automatically monitoring rice growth parameters in the field using inspection robots.

Autonomous particle-shape-based classification and identification of calcareous soils through machine learning

Recent geotechnical and geological applications have highlighted numerous challenges associated with conventional particle-size based soil classification, particularly for special calcareous soils widely used in marine and offshore constructions. Calcareous soils are always characterized by complex microstructures in terms of loose fabric, irregular particle shapes and abundant intra-particle pores. Therefore, this study aims to establish an autonomous particle-shape categorization criteria for calcareous soils, and further facilitate their intelligent particle identification and classification using image-based machine learning (ML) techniques. First, an extensive collection of images of individual calcareous sand particles were acquired through microscopic photography to establish the datasets for ML training. Four key shape parameters including circularity, roundness, aspect ratio and convexity were measured based on image processing and analysis. Subsequently, an unsupervised ML algorithm, K-means clustering, was applied to these shape parameters to autonomously determine optimal particle-shape categories for calcareous soils. Finally, based on the obtained particle shape classes, two supervised CNN algorithms, AlexNet and YOLO-v3, were trained and applied to automatically identify and classify individual particles or particle assemblies of calcareous soils from various raw image types. The predictive results demonstrate the high accuracy and efficiency of ML models in identifying and classifying the particle shapes of calcareous soils.

Probabilistic Modelling of the Nonlinear Mapping between Voltammogramsand the Grain Size of the Electrode Material in Li-Ion Batteries Using Optimally Tuned Gaussian Process Regression Models

Abstract The electrochemical response, and hence the performance, of lithium-ion batteries (LIBs) is heavily influenced by the microstructure of their electrode material, particularly grain size. Thus, a model that correlates electrode microstructure with electrochemical response can aid in optimizing these characteristics for better battery performance. However, no existing studies address this modeling problem.&amp;#xD; &amp;#xD;This study fills this gap by using data-driven probabilistic modelling to map the electrochemical response (voltammograms) to electrode grain size. The methodology is demonstrated on six datasets containing voltammograms from in-silico cyclic voltammetry experiments on LiMn2O4 electrode particles. Feature vectors created using seven features of the voltammogram were combined with the corresponding known grain size values to construct data-driven probabilistic models via Gaussian process regression (GPR) with five different kernels for each dataset. The hyperparameters of the kernels used in these GPR models were optimised using grid-search. Performance evaluation through bias-variance analysis demonstrated strong predictive accuracy of the developed models. Additionally, the comparisons across datasets and kernels showed the Gaussian and Rational Quadratic kernels' effectiveness in modelling various relationships. The obtained results substantiate the efficacy of the proposed approach, revealing significant insights and encouraging further exploration.

Clinical and molecular prognostic nomograms for patients with papillary renal cell carcinoma

ObjectiveTo summarize the clinicopathological characteristics and prognostic factors of papillary renal cell carcinoma (pRCC) and to construct clinical and molecular prognostic nomograms using existing databases.MethodsClinical prognostic models were developed using the Surveillance, Epidemiology, and End Results (SEER) database, while molecular prognostic models were constructed using The Cancer Genome Atlas (TCGA) database. Cox regression and LASSO regression were employed to identify clinicopathological features and molecular markers related to prognosis. The accuracy of the prognostic models was assessed using ROC curves, C-index, decision curve analysis (DCA) curves, and calibration plots.ResultsIn the 2004–2015 SEER cohort, Cox regression analysis revealed that age, grade, AJCC stage, N stage, M stage, and surgery were independent predictors of overall survival (OS) and cancer-specific survival (CSS) in pRCC patients. ROC curves, C-index, and DCA curves indicated that the prognostic nomogram based on clinical independent predictors had better predictive ability than TNM staging and SEER staging. Additionally, in the TCGA cohort, M stage, clinical stage, and the molecular markers IDO1 and PLK1 were identified as independent risk factors. The prognostic nomogram based on molecular independent risk factors effectively predicted the 3-year and 5-year OS and CSS for pRCC patients.ConclusionsThe clinical and molecular nomograms constructed in this study provide robust predictive tools for individualized prognosis in pRCC patients, offering better accuracy than traditional staging systems.

Real-time surrogate compensation architecture for machine-tool thermal error compensation with high-performance model

Thermal error compensation based on high-performance (high-accuracy prediction of thermal errors under complex machining conditions and environments) models is of great significance to high-precision CNC machining. Currently, the high-performance thermal error models can only run at the edge computer due to the computing power limitation of CNC system under real-time constraints, Therefore, the timeliness of the thermal error compensation instruction cannot be guaranteed; delayed and mutated compensation instructions cause defects in the machining surface, which is the main obstacle that restricts the application of thermal error compensation based on high-performance models. In order to solve this challenge, this research proposes a novel real-time surrogate compensation architecture for machine tool thermal error compensation with high-performance model. Based on the look-ahead machine tool control command, the thermal error of next stage could be predicted according to the high-performance thermal error prediction model running on the edge, and the parameters of thermal error surrogate compensation model running on the CNC system could be fitted for high accuracy real-time compensation of the thermal error during the processing process. The verification experimental results on the CK40S CNC lathe show that the proposed surrogate method can retain more than 95% of the accuracy of the original high performance thermal error prediction model. With the surrogate thermal error compensation, the machining surface quality of the machine tool is flawless, and the batch machining error has been reduced from 18.3 μm to 4.5 μm. Compared with the traditional Multi Liner Regression (MLR) model compensation method, the machining accuracy under complex working conditions is further improved by 50%.

Analyzing Twitter Sentiments on Booster Vaccination with Support Vector Machine (SVM) Method

The Indonesian government has implemented various measures to prevent the spread of the COVID-19 virus, one of which is through a vaccination program with two doses (the first dose and the second dose). However, new variants of the virus have emerged, reducing the effectiveness of the initial vaccinations. To address this, the government introduced a booster vaccination program aimed at enhancing immunity by up to 80%. The government's plan for booster vaccination has received both positive and negative opinions from the public through various media platforms, including Twitter. This study analyzes public opinions on the booster vaccination plan into three classes: positive, negative, and neutral. SVM is a classification method in machine learning categorized as supervised learning, which involves finding an optimal line (hyperplane) as a separator for two different data classes. The stages of this research include data collection, data cleaning, data transformation, and data classification using the Support Vector Machine (SVM) method. The results of this study indicate that the accuracy of the SVM model reaches 80.42%.

Research on structural vertical stiffness evaluation of track-specific suspension bridge under the action of the no-load first train

The track-specific bridge has the characteristics of a large volume of passenger traffic, narrow transverse width, and large excitation. The long-term and high-frequency operating conditions put forward strict requirements for the safety and comfort of the bridge. Aiming at the problem of achieving fast, accurate, and intelligent evaluation of bridges under the coupling action of trains and temperatures, this paper proposes a method for evaluating the structural vertical stiffness of track-specific suspension bridges under the action of the no-load first train. Firstly, the [Formula: see text] (deformation-velocity-time) data of trains on the bridge are analyzed through linkage numerical simulation, track train transponder, and structural safety monitoring system. Secondly, the CART algorithm establishes a regression model for data correction. The mapping relationship between the modified deformation monitoring value and the theoretical calculation value is analyzed to realize the current structural vertical stiffness evaluation. Finally, it is verified by the world’s largest span self-anchored track-specific suspension bridge. The results show that the theoretical deformation value of the most unfavorable deformation section is 268.55 mm, the actual monitoring value is 232.80 mm, the difference is 35.75 mm, and the relative error is 13.3%. The R2 score of the CART regression model is 0.94, and the accuracy of the CART regression model is higher than that of other algorithm regression models. The modified dynamic check coefficient is less than 1, and the ratio of the modified residual deformation value to the maximum deformation value is 7.6%, which meets the specification requirement not exceeding 20%, indicating that the current structural vertical stiffness is in normal operating conditions.

Validation and modification of existing bleeding complications prediction models for percutaneous renal biopsy: a prospective study

Background Bleeding complications following percutaneous renal biopsy (PRB) are a significant clinical concern. This study aimed to validate and refine existing prediction models for post-biopsy bleeding to support more accurate clinical decision-making. Methods Clinical data from 471 PRB patients were examined in this prospective analysis. Ultrasounds were performed immediately and 6 h post-biopsy to identify perinephric hematomas. Patients exhibiting severe pain, a hemoglobin drop of &gt;10 g/L, symptomatic hypotension, hematuria within 7 days post-procedure underwent repeat ultrasound to assess for bleeding complications. Univariate and multivariable logistic regression analyses were conducted to identify factors associated with bleeding risk. The predictive performance of three kidney biopsy risk calculators (KBRC) was evaluated using the area under the receiver operating characteristic (AUROC) curve, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA) to determine clinical utility. Nomograms were developed for each model to facilitate clinical application. Results Univariate analysis identified body mass index (BMI), hemoglobin, and ultrasound findings as significant predictors of bleeding complications. In multivariable analysis, BMI, immediate ultrasound, and 6-h ultrasound data remained significant (p &lt; 0.05). The three models compared included: KBRC-5 (age, body mass index (BMI), platelet count, hemoglobin, kidney size), KBRC-5 with immediate ultrasound data (IKBRC), and KBRC-5 with 6-h hematoma size (SKBRC). The AUROC values for these models were 0.683, 0.786, and 0.867, respectively (p &lt; 0.001). NRI and IDI analyses demonstrated that adding immediate or 6-h ultrasound data significantly improved the risk reclassification ability of the KBRC-5 model (p &lt; 0.05). DCA indicated that IKBRC provided the highest net benefit for risk thresholds between 25% and 77%, while SKBRC was superior for thresholds between 10% and 95%. Nomograms were constructed for each model, allowing clinicians to estimate the probability of bleeding complications by summing scores for each predictor. Calibration curves showed good agreement between predicted and observed probabilities. Conclusion Incorporating real-time ultrasound data post-PRB significantly enhances the predictive accuracy and risk reclassification capability of bleeding risk models. These findings provide critical insights for guiding clinical management decisions in patients undergoing renal biopsy.

Whole slide image based prognosis prediction in rectal cancer using unsupervised artificial intelligence

BackgroundRectal cancer is a common cancer worldwide and lacks effective prognostic markers. The development of prognostic markers by computational pathology methods has attracted increasing attention. This paper aims to construct a prognostic signature from whole slide images for predicting progression-free survival (PFS) of rectal cancer through an unsupervised artificial intelligence algorithm.MethodsA total of 238 patients with rectal cancer from two datasets were collected for the development and validation of the prognostic signature. A tumor detection model was built by transfer learning. Then, on the basis of the tumor patches recognized by the tumor detection model, a convolutional autoencoder model was built for decoding the tumor patches into deep latent features. Next, on the basis of the deep latent features, the tumor patches were divided into different clusters. The cluster number and other hyperparameters were optimized by a nested cross-validation method. The percentage of each cluster from the patient’s tumor patches, which is hereafter called PCF, was calculated for prognostic signature construction. The prognostic signature was constructed by Cox proportional hazard regression with L2 regularization. Finally, bioinformatic analysis was performed to explore the underlying biological mechanisms of the PCFs.ResultsThe accuracy of the tumor detection model in distinguishing tumor patches from non-tumor patches achieved 99.3%. The optimal cluster number was determined to be 9. Therfore, 9 PCFs were calculated to construct the prognostic signature. The prognostic signature achieved a concordance index of 0.701 in the validation cohort. The Kaplan-Meier survival curves showed the prognostic signature had good risk stratification ability. Through the bioinformatic analysis, several PCF-associated genes were identified. These genes were enriched in various gene ontology terms.ConclusionThe developed prognostic signature can effectively predict PFS in patients with rectal cancer and exploration of the underlying biological mechanisms may help to promote its clinical translation.