Development Of Predictive Models Research Articles

A Security Situation Prediction Model for Industrial Control Network Based on Explainable Belief Rule Base

Industrial Control Systems (ICSs) are vital components of industrial production, and their security posture significantly impacts operational safety. Given that ICSs frequently interact with external networks, cyberattacks can disrupt system symmetry, thereby affecting industrial processes. This paper aims to predict the network security posture of ICSs to ensure system symmetry. A prediction model for the network security posture of ICSs was established utilizing Evidence Reasoning (ER) and Explainable Belief Rule Base (BRB-e) technologies. Initially, an evaluation framework for the ICS architecture was constructed, integrating data from various layers using ER. The development of the BRB prediction model requires input from domain experts to set initial parameters; however, the subjective nature of these settings may reduce prediction accuracy. To address this issue, an ICS network security posture prediction model based on the Explainable Belief Rule Base (BRB-e) was proposed. The modeling criteria for explainability were defined based on the characteristics of the ICS network, followed by the design of the inference process for the BRB-e prediction model to enhance accuracy and precision. Additionally, a parameter optimization method for the explainable BRB-e prediction model is presented using a constrained Projection Equilibrium Optimization (P-EO) algorithm. Experiments utilizing industrial datasets were conducted to validate the reliability and effectiveness of the prediction model. Comparative analyses indicated that the BRB-e model demonstrates distinct advantages in both prediction accuracy and explainability when compared to other algorithms.

SPECT/CT in Early Response Assessment of Patients with Metastatic Castration-Resistant Prostate Cancer Receiving 177Lu-PSMA-617.

177Lu-PSMA-617 (LuPSMA) is a newly established treatment for patients with metastatic castration-resistant prostate cancer (mCRPC), but survival outcomes vary widely, and predictors of treatment responses are needed. This study investigated the role of total tumor volumes (TTVs) and new lesions (NLs) determined by LuPSMA SPECT/CT in early cycles to predict subsequent outcomes in a real-world practice setting. Methods: Between June and December 2022, consecutive patients with mCRPC who received at least 2 administrations of LuPSMA with SPECT/CT 24 h after treatment were retrospectively reviewed. We evaluated associations between TTVs and the appearance of NLs at cycles 2 and 3 with subsequent prostate-specific antigen (PSA) progression-free survival and overall survival (OS) using multivariate Cox regression. All analyses were adjusted for changes in PSA level relative to baseline. Results: Sixty-six mCRPC patients (median age, 74 y) received a median of 4 (interquartile range, 3-5) cycles of LuPSMA. Median follow-up starting at cycle 2 was 42 wk (interquartile range, 33-48 wk), with 24 of 66 patients deceased at the time of the analysis. Changes in TTV measured at the start of cycles 2 and 3 relative to baseline correlated significantly with corresponding changes in PSA level (r = 0.55 and 0.56), but absolute TTVs did not correlate significantly (r = 0.00 and 0.18). Patients with a higher absolute TTV at the start of cycle 2 had worse PSA progression-free survival and OS (hazard ratio [HR], 1.4 [95% CI, 1.1-1.8] and 2.1 [95% CI, 1.5-2.9]), with consistent results at the start of cycle 3 (HR, 2 [95% CI, 1.4-2.9] and 2 [95% CI, 1.2-3.2]). NLs were identified in 13 of 66 and 11 of 51 patients at the start of cycles 2 and 3. NLs at the start of cycle 2 were associated with worse OS (HR, 5.8 [95% CI, 1.9-17.5]), with consistent results at the start of cycle 3 (HR, 4.9 [95% CI, 1.3-18.6]). In multivariate analysis, a higher TTV and the appearance of NLs at the start of cycles 2 and 3 were independently associated with poorer OS. Conclusion: Higher TTVs and NLs on LuPSMA SPECT/CT at the start of cycles 2 and 3 were independently associated with higher risk of death. These measures provided prognostic information independent of changes in PSA. Development of prognostic and predictive models including TTV, NLs, and PSA changes is warranted.

Experimental and modelling analysis of waste material-based geopolymer concrete incorporated with crumb rubber particles

Rubberized geopolymer concrete (RuGPC) emerges as an eco-friendly alternative to conventional concrete, significantly reducing greenhouse gas emissions. This study investigates the use of steel slag (SS) in varying proportions (30 %, 35 %, and 40 %) as a replacement for ground granulated blast furnace slag (GGBS) in geopolymer concrete, with crumb rubber (CR) replacing crusher dust (CD) as fine aggregate due to its increasing demand. The research focuses on understanding the impact of aluminosilicate materials on the mechanical, thermal, and microstructural properties of geopolymer concrete cured at 60°C. Advanced characterization techniques, including Scanning Electron Microscopy (SEM), Energy-Dispersive X-ray Spectroscopy (EDX), X-ray Diffraction (XRD), and Fourier Transform Infrared Spectroscopy (FTIR), were employed. SEM and EDX analyses revealed that the microstructural properties of GGBS and SS materials, mainly Na/Si, Si/Al, H₂O/Na₂O, and Na/Al ratios, significantly influence RuGPC performance through gel formation. FTIR analysis indicates a shift in the stretching vibrations of GGBS and SS to lower wavenumbers due to geopolymerization changes. XRD results show the formation of C-S-H gel at around 27–30° 2theta, attributed to increased GGBS and SS content. Despite efforts to incorporate CR into geopolymer matrices, challenges in mitigating strength degradation persist. To address this, a predictive model was developed to understand the key factors affecting RuGPC performance. Six machine learning techniques—M5P (pruned and unpruned), random forest, random tree, linear regression, and support vector machine with various kernels (PUK, RBF, PK, and NPK), and artificial neural network (ANN)—were employed to predict the physical and thermal behavior of RuGPC. The analysis identified ANN-based models as the most effective. Sensitivity analysis highlighted the grade of rubber and the CR replacement percentage by volume of CD as the most influential parameters determining RuGPC compressive strength, density, and thermal conductivity. Moreover, The economic analysis revealed that RuGPC mixtures were 1.2–11.61 % more cost-effective than OPC concrete. These findings underscore the importance of predictive model development in optimizing RuGPC properties for practical applications, offering valuable insights for decision-making processes.

Development of a Prediction Model for Specific Fuel Consumption in Rotary Tillage Based on Actual Operation

Tractor fuel consumption has typically been predicted using indoor test results under specific conditions. This study analyzes the factors affecting fuel consumption during rotary tillage in actual fields and develops a prediction model. The test field was divided into sections using a grid method, and rotary tillage operations were performed to measure various parameters, including soil strength, tractor’s transmission and PTO gear stages, tillage pitch, travel speed, engine and PTO shaft torque and speed, and fuel consumption. Pearson correlation identified variables affecting specific fuel consumption, and regression analysis was used to develop a prediction model. The model’s accuracy was analyzed using the coefficient of determination (R2) and root mean square error (RMSE), and it was compared with the ASABE’s fuel consumption prediction model. The test results showed that higher transmission and PTO gear stages, and tillage pitch decreased specific fuel consumption, while soil strength had no significant effect. Thus, operating at higher gear and PTO stages within suitable conditions enhances energy efficiency in rotary tillage. Statistical analysis showed that specific fuel consumption significantly correlated with tractor travel speed, PTO shaft power, and PTO shaft speed. The prediction model, including these variables, had the highest accuracy with R2 of 0.91 and RMSE of 0.011 L/kW·h. The developed prediction model showed significantly improved accuracy compared to the ASABE model, indicating that it can predict specific fuel consumption based on key operational variables in actual rotary tillage operations.

Development and validation of a risk prediction model for postpartum urinary incontinence in primiparas with singleton pregnancies: a multicenter clinical investigation

BackgroundPostpartum urinary incontinence (UI) is a serious condition that significantly affects the quality of life. Several studies have demonstrated that it is associated with pelvic floor dysfunction. This study aimed to develop and validate a UI risk prediction model to identify primiparas with singleton pregnancies at high risk.MethodsA multistage stratified random sampling process was used. UI was measured using the International Standard Consultation on Incontinence Questionnaire Form (a modified Bristol questionnaire, ICIQ-FLUTS). Records of 1,340 primiparas with singleton pregnancies were reviewed, and data were collected from January 2014 to December 2014 in multiple centers. A univariate logistic regression analysis was performed, followed by a multivariable logistic regression analysis of the data. Using bootstrap resampling, we constructed a nomogram to assess postpartum UI risk.ResultsA total of 1,340 patients were enrolled, including 345 with postpartum UI and 995 with non-postpartum UI. The occurrence of postpartum UI was significantly related to the mode of delivery, family history of UI, coffee or tea consumption, antenatal UI, and frequent cough. The nomogram exhibited good discriminatory ability with a C-index of 0.718 (95% confidence interval: 0.684–0.752) and a bootstrap-corrected C-index of 0.716. Additionally, the calibration curve demonstrated that the predicted outcomes aligned well with the actual observations. Ultimately, the decision curve analysis indicated that the nomogram exhibited favorable clinical applicability.ConclusionThe decision curve analysis suggests that the nomogram could provide clinical value. The clinician will then feel more confident about making clinical recommendations regarding postpartum UI screening for primiparous women with singleton pregnancies.

Development and validation of a clinical nomogram prediction model for surgical site infection following lumbar disc herniation surgery

Surgical site infection (SSI) following lumbar disc herniation (LDH) surgery leads to prolonged hospital stays, increased costs and reoperations. Therefore, we aim to develop and validate a nomogram to predict the risk of SSI following LDH surgery, thereby helping spine surgeons design personalized prevention strategies and promote early recovery. Data from 647 patients with SSI who underwent LDH surgery at the First Affiliated Hospital of Air Force Medical University (AFMU) from 2020 to 2023 were collected. Ultimately, 241 patients with SSI were selected based on inclusion and exclusion criteria. Patients were randomly divided into training and validation sets with a ratio of 7:3. LASSO regression, univariate, and multivariate logistic regression were utilized to identify target variables and establish the prediction model, which was subsequently validated. Six factors—Age, Body Mass Index (BMI), Postoperative Suction Drainage (PSD), Gelatin Sponge (GS), None-Preoperative Antibiotic (NPTA), and Thrombin Time (TT)—were selected to construct the nomogram model. In the training set, the area under the curve (AUC) for the nomogram was 0.818 (95% CI 0.779–0.857). In the validation set, the AUC was 0.782 (95% CI 0.717–0.846). Calibration curves for both sets showed satisfactory agreement between predicted and actual SSI probabilities. Decision curve analysis indicated that the nomogram is clinically useful with a threshold range of 1–90%. The Clinical Impact Curve (CIC) demonstrated an acceptable cost-benefit ratio. The developed nomogram model effectively predicts the risk of SSI following LDH surgery, enabling spine surgeons to formulate more professional and rational clinical prevention strategies.

Early Warning of Energy Storage Battery Fault Based on Improved Autoformer and Adaptive Threshold

To enhance voltage prediction accuracy in energy storage batteries and address the limitations of fixed threshold warning methods, a fault warning approach based on an improved Autoformer model and adaptive thresholds is proposed. First, a spatiotemporal filtering layer is introduced into the autocorrelation mechanism to analyze the trend features of voltage sequences across different frequency domains. Additionally, an adaptive gating residual connection is used to link the sublayer and current layer output features, which helps to improve the model's adaptive feature selection capability. This innovation enables the development of a robust voltage prediction model based on the enhanced Autoformer. Then, a similarity‐based adaptive threshold, using interval estimation, is employed to rapidly track variations in battery voltage, enabling dynamic adjustment of voltage thresholds. Finally, the proposed method is validated with real voltage data from an operational energy storage station. The experimental results shows that the proposed model has higher accuracy and robustness compared to similar methods. The adaptive threshold can reduce the false alarm rate by ≈18% and issue alarms at three sampling points ahead of the battery management system alarm, improving fault warning accuracy and illustrating that early fault warning is effectively and practically carried out using the method.

Development of prediction models for carbapenem-resistant Klebsiella pneumoniae acquisition and prognosis in adult patients

ObjectivesTo explore the risk factors and clinical outcomes of carbapenem-resistant Klebsiella pneumoniae (CRKP) infection and establish nomograms to predict the probability of CRKP infection and mortality in adult patients.MethodsPatients infected with KP from August 2019 to April 2021 in a tertiary hospital in Shanghai were enrolled. Risk factors associated with CRKP and 30-day mortality were identified using multivariate logistic regression analysis and Cox regression analysis.ResultsOverall, 467 patients with KP infection were enrolled, wherein 210 (45.0%) patients were infected with CRKP and 257 (55.0%) patients with carbapenem-susceptible K. pneumoniae (CSKP). Five factors, namely Charlson’s Comorbidity Index (CCI) ≥ 3, the use of central venous catheterization, prior hospitalization during the 3 months before infection, and previous exposure to carbapenems and broad-spectrum β-lactams, were found to be independently associated with CRKP infection. Based on these parameters, the nomogram showed a better performance as indicated by C-index of 0.94 (95% confidence interval [CI]: 0.92–0.96) and well-fitted calibration curves. CRKP was independently associated with 30-day mortality. Multivariate Cox regression analysis revealed that age ≥65 years, higher CCI scores, higher Sequential Organ Failure Assessment scores, the presence of respiratory failure, albumin levels ≤30 g/L, and non-appropriate treatments in 3 days, were associated with 30-day mortality.ConclusionThe predictive nomogram established in this study can facilitate the clinicians to make better clinical decisions when treating patients with KP infection.

Comparative assessment of the energy performance index of modern urban and traditional rural housing in lower Himachal Pradesh, India to develop prediction model

Comparative assessment of the energy performance index of modern urban and traditional rural housing in lower Himachal Pradesh, India to develop prediction model

Enhancing accuracy in X-ray radiation-based multiphase flow meters: Integration of grey wolf optimization and MLP neural networks

This research investigates the development of an advanced predictive model aimed at accurately determining the volumetric percentages of water, oil, and gas within oil pipeline systems. Utilizing an innovative approach that incorporates an X-ray source alongside two sodium iodide detectors, the study leverages the Monte Carlo N-Particle (MCNP) simulation code to model the behavior of three-phase fluids under varied conditions. The model meticulously simulates various volumetric configurations of water, oil, and gas, resulting in a comprehensive dataset that provides key spectral information. The initial phase involved the extraction of ten temporal and frequency-related features from each detector, culminating in a pool of twenty features. The analytical process then applied the Grey Wolf Optimization (GWO) algorithm to select the most indicative features for predictive modeling. Out of the initial set, seven features—short-time energy, frequency deviation, relative spectral density, spectral margin, main peak position, spectral coefficient, and frequency intensity—were identified as critical for enhancing model accuracy. These features were subsequently fed into a meticulously structured multilayer perceptron (MLP) neural network. This network, designed with two hidden layers containing 20 and 10 neurons, respectively, demonstrated exceptional capability, achieving a root mean square error (RMSE) of less than 0.06 in the prediction of oil and gas volumetric percentages. The study emphasizes the significant impact of integrating refined feature selection techniques and robust neural network architectures on the precision and reliability of volumetric predictions in multiphase flow systems within oil pipelines. This approach not only enhances predictive accuracy but also contributes to more efficient resource management and operational planning in the oil and gas industry.

PR-FCNN: a data-driven hybrid approach for predicting PM2.5 concentration

The atmosphere’s fine articulate Matter (PM2.5) poses various health-related risks. Even though multiple efforts have been made to lower the emissions of these substances, the mortality rate is continuously increasing, requiring immediate inclination of the scientific community towards the design and development of advanced predictive models. Conventional statistical approaches have become dormant due to their limitations in capturing the innate relationships between the pollutants, particularly for predicting PM2.5 concentrations. In contrast, machine and deep learning techniques have shown great potential for forecasting air quality, providing more accuracy than its predecessor techniques. The present study investigates the utilization of hybrid approaches by integrating machine learning models with deep learning models to improve the prediction capabilities of PM2.5 concentration. It uses datasets from the World Air Quality Index (WAQI) and the State of Global Air (SOGA) to analyze the performance of the models on both the daily and annual data, respectively. This ensures the model’s effectiveness on a diversified dataset. The present study implements Random Forest (RF), Polynomial Regression (PR), XGBoost, and Extra Tree Regressor (ETR) coupled with Fully Connected Neural Network (FCNN), Long Short-Term Memory (LSTM), and Bi-directional LSTM (Bi-LSTM) for obtaining optimized results. Finally, after a thorough investigation, the hybrid PR model coupled with FCNN (PR-FCNN) is found to be the best model with improved R-squared (R2) values, portraying its potential for predicting PM2.5 concentration accurately. Based on the experimentation, the preset study recommends implementing hybrid approaches, offering better predictive accuracy in forecasting air pollutants, especially PM2.5.

Immune checkpoint inhibitors-related thyroid dysfunction: influencing factor analysis, prediction model development, and management strategy proposal

BackgroundWith the extensive utilization of immune checkpoint inhibitors (ICIs) across various cancers, ICIs-related thyroid dysfunction (ICI-TD) has become a growing concern in clinical practice. This study aimed to devise an individualized management strategy for ICI-TD to enhance the early identification and proactive management in cancer patients.MethodsWe designed and conducted a three-phase study. Initially, we analyzed the influencing factors through a systematic review and meta-analysis, which adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. Moreover, the study protocol was registered with PROSPERO (CRD42019131133). Subsequently, prediction models for ICI-TD were developed utilizing 11 algorithms based on the real-world cohort data from July 20, 2018 (the approval date of the first ICIs, Pembrolizumab in China), to October 31, 2022. Considering discrimination, calibration, and clinical utility, we selected the model with the best performance for web calculator development. Finally, individualized management strategies for ICI-TD were proposed by combining evidence-based analysis with practical considerations.ResultsThe systematic review encompassed 21 observational studies involving 4,145 patients, revealing associations between ICI-TD and factors such as female gender, age, receipt of Pembrolizumab (versus other ICIs), and baseline levels of thyroid-stimulating hormone, free thyroxine, and antithyroid antibodies. In the prediction model development phase, 621 participants were enrolled, with 36 patients developing ICI-TD. The model based on the LightGBM algorithm demonstrated superior performance, leading to the development of a web calculator. Based on these findings and existing guidelines, individualized monitoring and treatment pathways for pharmacists were devised.ConclusionThis study offers comprehensive insights into managing ICI-TD, potentially enhancing tailored cancer immunotherapy management.

Critical heat flux enhancement by porous coatings and prediction model development under External Reactor Vessel Cooling conditions

Critical heat flux (CHF) is the important limitation to estimate the effectiveness of the In-vessel retention (IVR) strategy for the severe accident mitigation. However, as the reactor power increases in the new-generation nuclear power plant, it is necessary to ensure the sufficient safety margin by taking some measures for the CHF enhancement. Porous coatings are recognized as an effective technique to improve the CHF. In this paper, the applicability of the porous coatings has been evaluated. The porous coatings consist of a dense layer and a porous layer. CHF experiments were constructed at the REPEC-III facility with 1:1 height ratio to the prototypic External Reactor Vessel Cooling (ERVC) environment. CHF on the porous surface is enhanced 32.4%–48.5 % under the ERVC conditions. Based on the experimental data, a mechanistic model considering the porous surface characteristics to predict CHF of the flow boiling under the ERVC natural circulation is developed and has a good predictive ability. The predicted CHF is compared with the experimental data of large-scale ERVC experiments. The mechanistic model can predict CHF within ±15 % error. Through the full-height experimental and theoretical analyses, porous coatings are an effective method to improve CHF and has a good application perspective for the IVR strategy in the nuclear power plant.

Development and validation of prediction models for prehospital triage of military trauma patients.

The introduction of wireless sensors will enable military care providers to continuously and remotely assess/monitor vital signs. Prediction models are needed to use such data adequately and aid military care providers in their on-scene decision-making to optimise prehospital triage and improve patient outcomes. A prospective cohort comprising data from eight Emergency Medical Services and seven inclusive trauma regions was used to develop and validate prediction models that could aid military care providers in their prehospital triage decisions. Healthy (American Society of Anesthesiologists physical status classification 1 or 2) admitted adult trauma patients (aged ≥16 and ≤50 years), who suffered from a trauma mechanism that could occur to military personnel and were transported by ambulance from the scene of injury to a hospital, were included. A full model strategy was used, including prehospital predictors that are expected to be automaticly collectible by wireless sensors or to be incorporated in a personalised device that could run the models. Models were developed to predict early critical-resource use (ECRU), severe head injury (Abbreviated Injury Scale (AIS) ≥4), serious thoracic injury (AIS ≥3) and severe internal bleeding (>20% blood loss). Model performance was evaluated in terms of discrimination and calibration. Prediction models were developed with data from 4625 patients (80.0%) and validated with data from 1157 patients (20.0%). The models had good to excellent discriminative performance for the predicted outcomes in the validation cohort, with an area under the curve of 0.80 (95% CI 0.76 to 0.84) for ECRU, 0.83 (0.76 to 0.91) for severe head injury, 0.75 (0.70 to 0.80) for serious thoracic injury and 0.85 (0.78 to 0.93) for severe internal bleeding. All models showed satisfactory calibration in the validation cohort. The developed models could reliably predict outcomes in a simulated military trauma population and potentially support prehospital care providers in their triage decisions.

Development of prediction models for strength behavior of MSWIA mixed with fiber and cement

Development of prediction models for strength behavior of MSWIA mixed with fiber and cement

Time Prediction in Ship Block Manufacturing Based on Transfer Learning

Accurate time prediction is critical to the success of ship block manufacturing. However, the emergence of new ship types with limited historical data poses challenges to existing prediction methods. In response, this paper proposes a novel framework for ship block manufacturing time prediction, integrating clustering and the transfer learning algorithm. Firstly, the concept of distributed centroids was innovatively adopted to achieve the clustering of categorical attribute features. Secondly, abundant historical data from other types of blocks (source domain) were incorporated into the neural network model to explore the effects of block features on manufacturing time, and the model was further transferred to blocks with limited data (target domain). Leveraging the similarities and differences between source and target domain blocks, actions involving freezing and fine-tuning parameters were adopted for the predictive model development. Despite a small sample size of only 80, our proposed block time prediction method achieves an impressive mean absolute percentage error (MAPE) of 8.62%. In contrast, the MAPE for the predictive model without a transfer learning algorithm is notably higher at 14.97%. Experimental validation demonstrates the superior performance of our approach compared to alternative methods in scenarios with small sample datasets. This research addresses a critical gap in ship block manufacturing time prediction.

Decision support tools for effective bus fleet electrification: Replacement factors and fleet size prediction

The electrification of public transit systems represents a crucial strategy for advancing sustainable urban mobility. Thus, the development of efficient charging infrastructure and the optimization of fleet size emerge as major challenges for transit agencies. Switching from diesel buses to electric buses (Ebuses) will require increasing the fleet size to accommodate the limited range of Ebuses and the significant idle time required for charging. This study develops prediction models to estimate the required Ebus fleet size to maintain same transit route services for the case of overnight depot charging, using data from Ebuses operating in the City of Toronto. The analysis reveals that Ebuses equipped with diesel auxiliary heaters are less sensitive to temperature fluctuations compared to battery-heated buses. Thus, the required replacement factor, indicating the additional fleet needed to switch from diesel to Ebuses, varies depending on the heating system. Specifically, diesel-heated buses require a lower replacement factor (1.3) compared to battery-heated buses (1.4), with winter conditions exacerbating this disparity. Furthermore, the study employs vehicular, operational, route, and external variables to develop the prediction models. Additionally, SHAP analysis is utilized to interpret the machine learning models and evaluate the influence of the inputs on the required fleet size. The results show that the total distance traveled, and the average temperature are the primary factors affecting the fleet size for Ebuses using their batteries for heating, whereas the total distance traveled, and the average bus speed are the primary factors affecting the fleet size for Ebuses with diesel auxiliary heaters.

12383 Development of an AI-Derived Endometriosis Prediction Model

12383 Development of an AI-Derived Endometriosis Prediction Model

Development of machine learning prediction model for AKI after craniotomy and evacuation of hematoma in craniocerebral trauma.

The aim of this study was to develop a machine-learning prediction model for AKI after craniotomy and evacuation of hematoma in craniocerebral trauma. We included patients who underwent craniotomy and evacuation of hematoma due to traumatic brain injury in our hospital from January 2015 to December 2020. Ten machine learning methods were selected to model prediction, including XGBoost, Logistic Regression, Light GBM, Random Forest, AdaBoost, GaussianNB, ComplementNB, Support Vector Machines, and KNeighbors. We totally included 710 patients. 497 patients were used for the training of the machine learning models and the remaining patients were used to test the performance of the models. In the validation cohort, the AdaBoost model got the highest area under the receiver operating characteristic curve (AUC) (0.909; 95% CI, 0.849-0.970) compared with other models. The AdaBoost model showed an AUC of 0.909 (95% CI, 0.849-0.970) in the validation cohort. Although there was an underestimated acute kidney injury risk for the model in the calibration curve, there was a net benefit for the AdaBoost model in the decision curve. Our machine learning model was evaluated to have a good performance in the validation cohorts and could be a useful tool in the clinical practice.

Development and Validation of a Comprehensive Predictive Model for Surgical Planning in Patients with Renal Cell Carcinoma and Inferior Vena Cava Tumor Thrombus

ObjectiveTo investigate preoperative predictors for selecting different surgical approaches in patients with renal cell carcinoma with inferior vena cava (IVC) tumor thrombus (RCC-IVCTT), and to establish and validate corresponding predictive models. MethodsClinical data of 583 RCC-IVCTT patients were retrospectively analyzed. Of these, 465 cases were used to construct predictive models, and 118 cases were used for validation. Univariate and multivariate analyses identified independent predictors for surgical strategies. Two nomogram prediction models were established based on relevant independent factors to predict surgical approach. ResultsIn the development cohort, 342 patients underwent IVC thrombectomy (IVCT), 91 underwent IVC cavectomy (IVCC), and 32 underwent IVC reconstruction (IVCR). Multivariate logistic regression analysis identified the following predictors for inability to perform IVCT: higher Mayo classification of tumor thrombus (TT), IVC wall invasion, presence of bland thrombus, pan-immune-inflammation value (PIV) > 358×109, and maximum anteroposterior (AP) diameter of IVC at renal vein ostium (RVo) > 24mm. Platelet > 170×109/L and inadequate collateral circulation were predictors for IVCR. The developed model predicted capacity of the nomogram was evaluated in terms of its calibration, discrimination, and clinical utility. The validation set confirmed these findings. ConclusionThe comprehensive preoperative predictive model for RCC-IVCTT patients aids in preoperative determination of the required surgical approach and necessity for IVC angiography, facilitating perioperative preparation and reducing unnecessary invasive examinations.