Performance Of Prediction Models Research Articles

A new model for predicting IRI using TabNet with hunter-prey optimization

ABSTRACT To address the prediction problem of the international roughness index (IRI), which is influenced by multiple factors, we propose a hunter-prey optimization (H-PO) TabNet prediction model. First, we combined multiple data sources, including traffic flow, climate data, pavement basic, and pavement condition data, aligned with annual detection data for highways at a spatial resolution of 100 m. Second, we conducted feature importance analysis on the integrated dataset to identify the most significant factors influencing pavement surface smoothness, which serve as inputs for the prediction model. We employed multiple machine-learning prediction models to predict the IRI using the compiled multisource dataset. After comparing the predictive performance of different models, we selected the TabNet model as the base model and optimised its parameter tuning process using the H-PO algorithm. Finally, ablation experiments were conducted to validate the proposed model. The results demonstrate that the H-PO-TabNet model achieved the best performance, with an R 2 increase of 7.92% compared with a model with default parameter settings. The H-PO-TabNet model also attained the highest overall accuracy, with an R 2 value of 0.8763. This model can improve the accuracy of IRI prediction and provide certain data support for pavement maintenance activities.

Price Prediction Models for Construction Sector – Efficiency and Interpretability

The construction sector, essential for the global economy, is characterized by high price volatility, influenced by economic factors such as material costs, labor market fluctuations, and oil price variations. At the same time, advances in technology and artificial intelligence (AI) have opened new opportunities for improving price predictability through the application of advanced machine learning models. This research aims to compare the performance of several price prediction models, such as Ridge Regression, Random Forest, XGBoost, and Neural Networks, to identify the most efficient model applied in the construction sector The data used in this study were extracted from the Federal Reserve Economic Data (FRED) database and include key variables such as the number of construction employees, job vacancies, the unemployment rate, and oil prices. Based on this data, the four predictive models mentioned were applied and compared, using evaluation metrics such as Mean Squared Error (MSE) and R-squared to determine which model provides the best price prediction. The results provide a detailed assessment of each model and highlight the importance of using AI technologies in managing economic risks in the construction industry, contributing to the optimization of decision-making strategies for companies in this sector. The study provides a valuable set of tools and predictive models that can support managers in the construction industry in making informed and strategic decisions.

Noninvasive assessment of hepatic steatosis grades by ultrasound derived fat fraction in metabolic dysfunction associated steatotic liver disease

To evaluate the diagnostic performance of ultrasound-derived fat fraction (UDFF) and clinical prediction models in assessing hepatic steatosis grades in MASLD patients, with liver biopsy as reference standard.A total of 85 obese patients who were found to have fatty liver by B-mode ultrasound and underwent UDFF measurement, with liver biopsy available, were enrolled. The diagnostic performance of UDFF, clinical prediction models including fatty liver index (FLI), hepatic steatosis index (HSI), ZheJiang University index (ZJU index) and lipid accumulation product (LAP) for hepatic steatosis was assessed. The areas under receiver operating characteristics curves (AUROCs) were utilized to determine the diagnostic efficacy. DeLong test was used to compare the diagnostic performance of different noninvasive methods for hepatic steatosis. The UDFF values of S1, S2 and S3 groups were 17.53 ± 7.49%, 25.00 ± 5.41% and 27.58 ± 6.55%, respectively (P < 0.001). UDFF values were significantly positively correlated with histologic steatosis grades (r = 0.531, P < 0.001). In the ≥ S2 group, the AUROC of UDFF was higher than that of FLI and LAP (P < 0.05), but not significantly different from that of HSI and ZJU index (P > 0.05). In the ≥ S3 group, the AUROC of UDFF was higher than that of FLI, ZJU index and LAP (P < 0.05), but not significantly different from that of HSI (P > 0.05). The UDFF proves effective in assessing hepatic steatosis in patients with MASLD, and its diagnostic efficacy exceeded that of FLI and LAP, but there was no significant difference with HSI, ZJU index in the ≥ S2 group, and with HSI in the ≥ S3 group.

The brain selectively allocates energy to functional brain networks under cognitive control

Abstract Network energy has been conceptualized based on structural balance theory in the physics of complex networks. We utilized this framework to assess the energy of functional brain networks under cognitive control and to understand how energy is allocated across canonical functional networks during various cognitive control tasks. We extracted network energy from functional connectivity patterns of subjects who underwent fMRI scans during cognitive tasks involving working memory, inhibitory control, and cognitive flexibility, in addition to task-free scans. We found that the energy of the whole-brain network increases when exposed to cognitive control tasks compared to the task-free resting state, which serves as a reference point. The brain selectively allocates this elevated energy to canonical functional networks; sensory networks receive more energy to support flexibility for processing sensory stimuli, while cognitive networks relevant to the task, functioning efficiently, require less energy. Furthermore, employing network energy, as a global network measure, improves the performance of predictive modeling, particularly in classifying cognitive control tasks and predicting chronological age. Our results highlight the robustness of this framework and the utility of network energy in understanding brain and cognitive mechanisms, including its promising potential as a biomarker for mental conditions and neurological disorders.

Evaluation of methotrexate Pharmacogenomic variation to predict acute neurotoxicity in children with acute lymphoblastic leukemia.

Methotrexate is an important component of curative therapy in childhood acute lymphoblastic leukemia (ALL), but the role of genetic variation influencing methotrexate clearance and transport in toxicity susceptibility in children with ALL is not well established. Therefore, we evaluated the association between suspected methotrexate pharmacogenomic variants and methotrexate-related neurotoxicity. This study included children (aged 2-20 years) diagnosed with ALL (2005-2019) at six treatment centers in the southwest United States. Clinical information was abstracted from medical records. Suspected neurotoxic events occurring within 21 days of intravenous and/or intrathecal methotrexate delivered between the end of induction and start of maintenance therapy were independently reviewed by at least two pediatric oncologists. Germline DNA was genotyped and 97 methotrexate pharmacogenomic variants of interest with at least grade 3 evidence were identified using the Pharmacogenomics Knowledge Base. Associations between variants and neurotoxicity were assessed by logistic regression. Data were randomly split (80/20) and random forest was constructed to estimate the ability of the variants to correctly classify neurotoxicity. Of the 763 patients included in the study, 8.2% (n = 63) developed methotrexate-associated neurotoxicity. In logistic models, none of the 97 available pharmacogenomic variants reached adjusted statistical significance. However, two variants, rs17222723 (odds ratio [OR] = 2.83 [ref. = T allele], 95% confidence interval [CI]: 1.20-6.15) in ABCC2 and rs1045642 (OR = 0.66 [ref. = minor A allele], 95% CI: 0.44-0.98) in ABCB1, were nominally associated (p-value < 0.05) with neurotoxicity susceptibility. The addition of pharmacogenomic variants did not improve the predictive performance of random forest model (AUC = 0.73) compared to clinical information alone (AUC = 0.74). Overall, our results suggest that associations between neurotoxicity susceptibility and methotrexate pharmacogenomic variants are generally modest and these variants do not significantly improve neurotoxicity risk stratification among children with ALL.

Water Quality by Spectral Proper Orthogonal Decomposition and Deep Learning Algorithms

Water quality plays a pivotal role in human health and environmental sustainability. However, traditional water quality prediction models are limited by high model complexity and long computation time, whereas AI models often struggle with high-dimensional time series and lack physical interpretability. This paper proposes a two-dimensional water quality surrogate model that couples physical numerical models and AI. The model employs physical simulation results as input, applies spectral proper orthogonal decomposition to reduce the dimensionality of the simulation results, utilizes a long short-term memory neural network for matrix forecasting, and reconstructs the two-dimensional concentration field. The simulation and predictive performance of the surrogate model were systematically evaluated through four design scenarios and three sampling dataset lengths, with a particular focus on the convection–diffusion zone and high-concentration zone. The results indicated that the model achieves high prediction accuracy for up to 7 h into the future, with sampling dataset lengths ranging from 20 to 80 h. Specifically, the model achieved an average R2 of 0.92, a MAE of 0.38, and a MAPE of 1.77%, demonstrating its suitability for short-term water quality predictions. The methodology and findings of this study demonstrate the significant potential of integrating spectral proper orthogonal decomposition and deep learning for water quality prediction. By overcoming the limitations of traditional models, the proposed surrogate model provides high-accuracy predictions with enhanced physical interpretability, even in complex, dynamic environments. This work offers a practical tool for rapid responses to water pollution incidents and supports improved watershed water quality management by effectively capturing pollutant diffusion dynamics. Furthermore, the model’s scalability and adaptability make it a valuable resource for addressing intelligent management in environmental science.

Evaluating the impact of modeling choices on the performance of integrated genetic and clinical models.

The value of genetic information for improving the performance of clinical risk prediction models has yielded variable conclusions. Many methodological decisions have the potential to contribute to differential results. We performed multiple modeling experiments integrating clinical and demographic data from electronic health records (EHR) with genetic data to understand which decisions may affect performance. Clinical data in the form of structured diagnostic codes, medications, procedural codes, and demographics were extracted from two large independent health systems and polygenic risk scores (PRS) were generated across all patients of European ancestry with genetic data in the corresponding biobanks. Crohn's disease was studied based on its substantial genetic component, established EHR-based definition, and sufficient prevalence for training and testing. We investigated the impact of choices regarding PRS integration method, training sample, model complexity, and performance metrics. Overall, our results show that including PRS resulted in higher performance but this gain was only robust in situations with limited clinical information. We find consistent performance increases from more compute-intensive models such as random forest, but the impact of other decisions vary by site. This work highlights the importance of considering methodological decision points in interpreting the impact of PRS on prediction performance in clinical models.

Integrating Muscle Depletion with Barcelona Clinic Liver Cancer Staging to Predict Overall Survival in Hepatocellular Carcinoma

Background: Muscle depletion (MD) is a critical factor that influences clinical outcomes in patients with hepatocellular carcinoma (HCC). Although its role in cancer prognosis is recognized, its integration into widely used prognostic systems remains underexplored. This study aimed to evaluate the prognostic impact of MD on overall survival (OS) in HCC patients and to improve existing noninvasive prognostic models by incorporating MD-related metrics. Methods: A retrospective analysis was conducted on 1072 HCC patients treated at Taipei Tzu Chi Hospital between January 2006 and December 2016. All patients had follow-up data and computed tomography (CT) scans at vertebral level L3 for MD evaluation. Independent prognostic factors for OS were identified using Cox proportional hazards models, and the predictive performance of various prognostic models was assessed through the area under the receiver operating characteristic curve (AUROC). Results: The key independent predictors of OS in HCC patients included hepatitis B virus infection, hepatitis C virus infection, liver cirrhosis, tumor size, serum alpha-fetoprotein levels, and MD-related metrics (psoas muscle-to-spine ratio, psoas muscle-to-vertebral ratio, and myosteatosis). Among existing models, the Barcelona Clinic Liver Cancer (BCLC) stage, the Child–Turcotte–Pugh (CTP) class, and the albumin–bilirubin (ALBI) grade demonstrated robust predictive performance for OS. However, incorporating MD significantly improved the predictive accuracy of these models, with the MD–BCLC model showing the highest AUROC (0.804, 95% CI: 0.777–0.832, p &lt; 0.001). Conclusions: MD is an independent and significant prognostic predictor for patients with HCC. Integrating MD metrics into established systems, particularly the BCLC staging system, markedly improves OS prediction, providing a more comprehensive tool for clinical decision-making in the management of HCC.

Performance optimization of binary fluid-filled heat pipes for battery thermal management applications using RSM and ANN schemes: A comparative assessment

In the present scenario, the development of efficient lithium-ion energy storage system-based electric vehicles has been turned into the focus as an effective alternative to conventional transportation systems. However, correspondingly associated potential risks of thermal failure confirm the need for an effective thermal management system to mitigate the adverse effects of excessive heat generated during operation. This work takes an opportunity to present a comparative performance assessment of response surface method (RSM) and artificial neural networks (ANN)-based predictive models to analyze the thermal behavior of oscillating heat pipes (OHP) filled with binary working fluids (acetone-DI water, acetone-methanol, and acetone-ethanol) with mixing ratio (1:1 ≤ MR ≤ 2:1) and volumetric filling ratio (30%≤FR ≤ 70%) as independent variables at different battery discharge rates (1C ≤ DR ≤ 2C). The thermal performance is predicted in terms of maximum cell temperature (MCT), average cell temperature (ACT), and thermal uniformity (TD). The acetone-DI water binary mixture was identified as the optimal working fluid among the fluids evaluated. The ACT, MCT, and TD were determined to be 38.3°C, 40°C, and 3.4°C, respectively, under a filling ratio (FR) of 70%, mixing ratio (MR) of 2:1, and a discharge rate (DR) of 2C. The prediction model performance is measured using correlation coefficient ( R2) value and average absolute prediction error (APE) to acknowledge the best fit. The results confirm the applicability of both the prediction models with the ANN model ( R2-values ≥ 0.99998; average APE ≥ 0.0162%) slightly more accurate relative to the corresponding RSM approach ( R2-values ≥ 0.9814; average APE ≥ 0.2342%).

Predictive model for HBsAg clearance rate in chronic hepatitis B patients treated with pegylated interferon α-2b for 48 weeks.

Chronic hepatitis B (CHB) is a major global health concern. This study aims to investigate the factors influencing hepatitis B surface antigen (HBsAg) clearance in CHB patients treated with pegylated interferon α-2b (Peg-IFNα-2b) for 48weeks and to establish a predictive model. This analysis is based on the "OASIS" project, a prospective real-world multicenter study in China. We included CHB patients who completed 48weeks of Peg-IFNα-2b treatment. Patients were randomly assigned to a training set and a validation set in a ratio of approximately 4:1 by spss 26.0, and were divided into clearance and non-clearance groups based on HBsAg status at 48weeks. Clinical data were analyzed using SPSS 26.0, employing chi-square tests for categorical data and Mann-Whitney U tests for continuous variables. Significant factors (p < 0.05) were incorporated into a binary logistic regression model to identify independent predictors of HBsAg clearance. The predictive model's performance was evaluated using ROC curve analysis. We included 868 subjects, divided into the clearance group (187 cases) and the non-clearance group (681 cases). They were randomly assigned to a training set (702 cases) and a validation set (166 cases). Key predictors included female gender (OR = 1.879), lower baseline HBsAg levels (OR = 0.371), and cirrhosis (OR = 0.438). The final predictive model was: Logit(P) = 0.92 + Gender (Female) * 0.66-HBsAg (log) * 0.96-Cirrhosis * 0.88. ROC analysis showed an AUC of 0.80 for the training set and 0.82 for the validation set, indicating good predictive performance. Gender, baseline HBsAg levels, and cirrhosis are significant predictors of HBsAg clearance in CHB patients after 48weeks of Peg-IFNα-2b therapy. The developed predictive model demonstrates high accuracy and potential clinical utility.

A robust adaptive meta-sample generation method for few-shot time series prediction

The research and exploration of time series prediction (TSP) have attracted much attention recently. Researchers can achieve effective TSP based on the deep learning model and a large amount of data. However, when sufficient high-quality data are not available, the performance of prediction models based on deep learning techniques may degrade. Therefore, this paper focuses on few-shot time series prediction (FTSP) and plans to combine meta-learning and generative models to alleviate the problems caused by insufficient training data. When using meta-learning techniques to process FTSP tasks, researchers set the meta-parameter in model-agnostic meta-learning (MAML) as a meta-sample and construct meta-sample generation methods based on advanced generative modeling theory to achieve better uncertainty coding. The existing meta-sample generation methods in FTSP scenes have an inherent limitation: With the increase of the complexity of prediction tasks, samples based on Gaussian distribution may be sensitive to noise and outliers in the meta-learning environment and lack of uncertainty expression, thus affecting the robustness and accuracy of prediction. Therefore, this paper proposes an adaptive sample generation method called JLSG-Diffusion. Based on the Jensen constraint framework and Laplace modeling theory, this method constructs a sample adapter with reasonable adaptive steps and fast convergence for specific tasks. The advantage is to realize fast adaptive convergence of samples to new tasks at lower cost, effectively control the overall generalization error, and improve the robustness and non-Gaussian generalization of sample posterior reasoning. Moreover, the meta sampler of JLSG-Diffusion embeds meta-learning from the implicit probability measure level of Denoising Diffusion Probabilistic Models (DDPM), which makes the meta-sample distribution directly establish a function mapping with the new task and effectively quantifies the uncertainty of spatiotemporal dimension. Experimental results on three real datasets prove the efficiency and effectiveness of JLSG-Diffusion. Compared with the benchmark methods, the prediction model combined with JLSG-Diffusion shows better accuracy.

Research on rock strength prediction model based on machine learning algorithm

The compressive strength of rocks is one of its mechanical characteristics. It has been a difficult problem to predict rock compressive strength conveniently and efficiently, and to solve the limitations of traditional rock compressive strength tests such as high cost, long time consumption, and reliability assurance. In this study, a data set containing 1774 groups of rock compressive strength test data was constructed through file retrieval, including 9 input parameters: rock type, temperature, confining pressure, dimension of specimen, shape of specimen, and experimental method. Eight supervised learning algorithms were used to learn the rock compressive strength test data, and eight rock compressive strength prediction models considering multiple factors were established to obtain a better method of predicting rock compressive strength. By selecting different features, the optimal feature combination for predicting rock compressive strength was obtained, and the optimal parameters for different models were obtained through the Sparrow Search Algorithm (SSA). Finally, four regression evaluation indicators, including mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R2), were used to evaluate the predictive performance of the established regression models. The results showed that the best-trained model had a MAPE as low as 3.61%, MAE as low as 9.19 MPa, and R2 as high as 0.995. It is noteworthy that AdaBoost was found to be the best model for predicting rock compressive strength. This study presents a significant advancement in the field by demonstrating the effectiveness of machine learning algorithms in this context, which have not been extensively applied to rock compressive strength predictions. The findings suggest that these models can offer substantial improvements over traditional methods, not only in accuracy but also in operational efficiency. This research is important for geotechnical engineering, as accurate rock strength predictions are critical for the design and stability assessments of construction projects, ultimately contributing to safer and more cost-effective engineering solutions.

Research on a Bearing Fault Diagnosis Method Based on a CNN-LSTM-GRU Model

In view of the problem of the insufficient performance of deep learning models in time series prediction and poor comprehensive space–time feature extraction, this paper proposes a diagnostic method (CNN-LSTM-GRU) that integrates convolutional neural network (CNN), long short-term memory (LSTM) network, and gated recurrent unit (GRU) models. In this study, a convolutional neural network (CNN) model is used to process two-dimensional image data in both time and frequency domains, and a convolutional core attention mechanism is introduced to extract spatial features, such as peaks, cliffs, and waveforms, from the samples. A long short-term memory (LSTM) network is embedded in the output processing of the convolutional neural network (CNN) to analyze the long-sequence variation characteristics of rolling bearing vibration signals and enable long-term time series prediction by capturing long-term dependencies in the sequence. In addition, a gated recurrent unit (GRU) is used to refine long-term time series predictions, providing local fine-tuning and improving the accuracy of fault diagnosis. Using a dataset obtained from Case Western Reserve University (CWRU), the average accuracy of CNN-LSTM-GRU fault vibration is greater than 99%, and its superior performance in a noisy environment is demonstrated.

Study protocol: Comparison of different risk prediction modelling approaches for COVID-19 related death using the OpenSAFELY platform

On March 11th 2020, the World Health Organization characterised COVID-19 as a pandemic. Responses to containing the spread of the virus have relied heavily on policies involving restricting contact between people. Evolving policies regarding shielding and individual choices about restricting social contact will rely heavily on perceived risk of poor outcomes from COVID-19. In order to make informed decisions, both individual and collective, good predictive models are required. For outcomes related to an infectious disease, the performance of any risk prediction model will depend heavily on the underlying prevalence of infection in the population of interest. Incorporating measures of how this changes over time may result in important improvements in prediction model performance. This protocol reports details of a planned study to explore the extent to which incorporating time-varying measures of infection burden over time improves the quality of risk prediction models for COVID-19 death in a large population of adult patients in England. To achieve this aim, we will compare the performance of different modelling approaches to risk prediction, including static cohort approaches typically used in chronic disease settings and landmarking approaches incorporating time-varying measures of infection prevalence and policy change, using COVID-19 related deaths data linked to longitudinal primary care electronic health records data within the OpenSAFELY secure analytics platform.

Predictors of outcome following psychological therapy for depression and anxiety in an urban primary care service: a naturalistic Bayesian prediction modeling approach.

England's primary care service for psychological therapy (Improving Access to Psychological Therapies [IAPT]) treats anxiety and depression, with a target recovery rate of 50%. Identifying the characteristics of patients who achieve recovery may assist in optimizing future treatment. This naturalistic cohort study investigated pre-therapy characteristics as predictors of recovery and improvement after IAPT therapy. In a cohort of patients attending an IAPT service in South London, we recruited 263 participants and conducted a baseline interview to gather extensive pre-therapy characteristics. Bayesian prediction models and variable selection were used to identify baseline variables prognostic of good clinical outcomes. Recovery (primary outcome) was defined using (IAPT) service-defined score thresholds for both depression (Patient Health Questionnaire [PHQ-9]) and anxiety (Generalized Anxiety Disorder [GAD-7]). Depression and anxiety outcomes were also evaluated as standalone (PHQ-9/GAD-7) scores after therapy. Prediction model performance metrics were estimated using cross-validation. Predictor variables explained 26% (recovery), 37% (depression), and 31% (anxiety) of the variance in outcomes, respectively. Variables prognostic of recovery were lower pre-treatment depression severity and not meeting criteria for obsessive compulsive disorder. Post-therapy depression and anxiety severity scores were predicted by lower symptom severity and higher ratings of health-related quality of life (EuroQol questionnaire [EQ5D]) at baseline. Almost a third of the variance in clinical outcomes was explained by pre-treatment symptom severity scores. These constructs benefit from being rapidly accessible in healthcare services. If replicated in external samples, the early identification of patients who are less likely to recover may facilitate earlier triage to alternative interventions.

Investment Market Trend Forecasting and Strategy Optimization based on Big Data Analytics

The financial market is an important part of the global economy, and there exists a certain special internal regularity in its changes. With the rapid development of big data technology, the huge amount of data in the financial market provides new opportunities for the prediction of investment trends and strategy optimization. Based on big data analytics technology, combined with machine learning models, the study explores how to construct investment market trend prediction models through effective data preprocessing and feature engineering, and evaluates and compares the performance of different prediction models. How to accurately predict the direction of the financial market will be an important reference value for relevant practitioners as well as researchers. Traditional financial time series research mainly focuses on linear trend fitting based on statistical knowledge, however, the trend of the financial market is not limited to pure mathematical and rational function curves, the financial market is affected by a combination of various factors from all walks of life. Big data analytics has significant application potential in the investment market, providing investors with more scientific and accurate decision support.

Data-space Validation of High-dimensional Models by Comparing Sample Quantiles

We present a simple method for assessing the predictive performance of high-dimensional models directly in data space when only samples are available. Our approach is to compare the quantiles of observables predicted by a model to those of the observables themselves. In cases where the dimensionality of the observables is large (e.g., multiband galaxy photometry), we advocate that the comparison is made after projection onto a set of principal axes to reduce the dimensionality. We demonstrate our method on a series of two-dimensional examples. We then apply it to results from a state-of-the-art generative model for galaxy photometry (pop-cosmos) that generates predictions of colors and magnitudes by forward simulating from a 16-dimensional distribution of physical parameters represented by a score-based diffusion model. We validate the predictive performance of this model directly in a space of nine broadband colors. Although motivated by this specific example, we expect that the techniques we present will be broadly useful for evaluating the performance of flexible, nonparametric population models of this kind, and other settings where two sets of samples are to be compared.

Predicting the risk of cesarean hysterectomy in the management of placenta accreta spectrum disorders: a new model based on clinical findings and ultrasonography.

To develop a nomogram to predict the probability of cesarean hysterectomy (CH) in placenta accreta spectrum disorders (PASD) patients. Data from 520 patients who underwent surgery with a preliminary diagnosis of PASD at a tertiary center in southeast Turkey between 2013 and 2023 were collected, and 302 patients were included in the study. A predictive model based on clinical and ultrasonographic variables was developed using penalized maximum likelihood estimation (PMLE) regression analysis. Maternal age (aOR = 1.22, 95% CI 1.08-1.44, p = 0.001) and prior uterine surgeries (aOR = 3.18, 95% CI 1.57-8.29, p = 0.001) were identified as demographic factors with an increased likelihood of CH in the nomogram, and advanced gestational weeks demonstrated a negative correlation (aOR: 0.78, 95% CI 0.56-1.02, p = 0.07). Regarding the ultrasonographic findings, the presence of the "multiple lacunae within the placenta" (aOR = 48.53, 95% CI 18.42-257.40, p < 0.001) and the "anterior placental location" (aOR = 9.60, 95% CI 2.96-50.76, p < 0.001) significantly increased the probability of CH. In addition, "hypervascularization on Doppler flow with irregularity in the line between the bladder and uterine serosa" (aOR = 7.90, 95% CI 2.66-35.12, p < 0.001) and the "retroplacental myometrial thickness of < 1mm" (aOR = 2.49, 95% CI 0.89-8.27, p = 0.08) were related to the probability of CH. Harrell's C-index was 0.974, and the kappa value was 0.819 for the prediction model's performance evaluation. We developed a nomogram to predict the probability of cesarean hysterectomy in patients with PASD, incorporating maternal age, gestational weeks, prior uterine surgeries, ultrasound findings, and placental location. The most closely associated findings with CH in patients with PASD were the presence of multiple placental lacunae and the anterior location of the placenta.

Numerical Study on Heat Leakage, Thermal Stratification, and Self-Pressurization Characteristics in Liquid Helium Storage Tanks

During the operation of liquid-phase He-4 (LHe-4) storage tanks, heat leakage changes the thermophysical parameters and phase properties of the LHe-4 in the tanks, resulting in the thermal layering phenomenon. This phenomenon is characterized by the LHe-4 temperature gradient and pressure increase (self-pressurization) phenomena in the tanks. Based on the Layer-by-Layer model, a heat transfer model of a composite adiabatic structure with multilayer insulation and liquid nitrogen screen (LNCS) insulation was established, and the Neumann boundary heat flux of the thermal response model was determined. A numerical simulation model of the thermal response of a liquid helium storage tank was established. The spatial and temporal evolutions of the pressure distribution, natural convection characteristics, thermal stratification characteristics, and self-pressurization characteristics of the LHe-4 tank were investigated. Finally, the self-pressurization thermodynamic model of the LHe-4 storage tank was built based on the isothermal saturation and homogeneous model. It is shown that the predictive performance of the mLee model for the self-boosting characteristics (relative deviation of 14.32%) was significantly improved compared with that of the Lee model (relative deviation of 39.64%). The thermal stratification degree (TSD) of the tank increased with the operation time, with TSDs of 1.023, 1.028, and 1.036 at 1 h, 2 h, and 3 h, which exacerbated the self-pressurization of the tank. The wall surface in contact with the phase interface is a strong evaporation point, so the interfacial mass transfer rate maps show a pattern of high at both ends and low in the middle.

Associations of plasma acylcarnitine and bile acid levels with incidence of coronary heart disease in Chinese adults

Objective: To explore the associations of plasma acylcarnitine and bile acid levels with the risk of incident coronary heart disease (CHD) in Chinese adults. Methods: The baseline survey of China Kadoorie Biobank (CKB) took place in 10 areas across China during 2004-2008, and the first resurvey took place from July to October 2008, with collection of data via questionnaire, physical examination and blood samples. The current study was based on 2 159 individuals with targeted mass spectrometry metabolomic measurements from the first resurvey of CKB. The associations of acylcarnitines and bile acids with incident CHD were assessed using Cox proportional hazards regression models. Unweighted metabolites scores were constructed to assess the overall effect of acylcarnitines and bile acids on incident CHD. The impact of metabolites on the performance of CHD prediction model was evaluated with the receiver operating characteristic (ROC) area under the curve (AUC). Follow-up for CHD incidence was censored on December 31, 2018. Results: The mean age of the participants was (53.1±9.8) years and 754 were males (34.9%). During (10.5±0.1) years of follow-up, 140 cases of CHD were recorded. Four metabolites including acylcarnitines C3-OH, C5:1, C5:1-DC, and deoxycholic acid (DCA) showed associations with CHD incidence and the HR (95%CI) were 1.474 (1.230-1.767), 0.761 (0.637-0.909), 0.773 (0.650-0.918), and 1.309 (1.113-1.539), respectively [false discovery rate (FDR)0.05]. All metabolite scores, including short-chain, medium-chain, long-chain acylcarnitines, primary and secondary bile acids scores were associated with the risk of CHD (FDR0.05). Compared to the traditional models, the addition of DCA or 4 key metabolites increased the AUC of the predictive model from 0.803 (0.761-0.845) to 0.812 (0.772-0.852) and 0.817 (0.778-0.857), respectively (all P0.05). Conclusions: Acylcarnitine and bile acid levels are associated with the risk of CHD, and DCA or 4 key metabolites can improve the predictive ability for CHD incidence.