Calibration Plots Research Articles

Risk factor analysis and establishment of a predictive model for complications of elderly advanced gastric cancer with Clavien-Dindo classification ≥ II grade

BackgroundThe occurrence of complications following radical gastrectomy for gastric cancer significantly impacts patients’ quality of life. Elderly patients are susceptible to postoperative complications. This study seeks to investigate the risk factors associated with Clavien-Dindo ≥IIgrade complications following radical gastrectomy for advanced gastric cancer in elderly patients, develop a nomogram risk prediction model, and validate its accuracy.MethodsRetrospective collection of clinical and pathological data was conducted on 442 elderly patients with advanced gastric cancer who underwent radical gastrectomy at Shaanxi Provincial People’s Hospital from January 2015 to December 2020. They were randomly divided into a training set (n = 310) and a validation set (n = 132) in a 7:3 ratio. The severity of postoperative complications was graded using the Clavien-Dindo classification system, resulting in two complication groups: Clavien-Dindo <II group (n = 229) and Clavien-Dindo ≥II group (n = 81). Univariate and multivariate logistic regression analyses were performed to identify independent risk factors affecting the occurrence of Clavien-Dindo ≥II grade complications, and a predictive model was established based on the results. The model was then validated using the validation set.ResultsAmong the 442 patients included in the study, 121 cases (27.38%) experienced postoperative complications, with 111 cases (25.11%) classified as Clavien-Dindo ≥II grade complications. Multivariable logistic analysis revealed that the Prognostic Nutritional Index (PNI), surgical duration, age, and history of Diabetes mellitus were independent risk factors for the occurrence of Clavien-Dindo ≥IIgrade complications in elderly patients with advanced gastric cancer after surgery (P < 0.05). The nomogram model constructed based on these factors demonstrated good discriminative ability, as indicated by the area under the Receiver Operating Characteristic (ROC) curve. Calibration plots showed that the predicted probability of gastric cancer lymph node metastasis using the nomogram model was well aligned with actual outcomes. Decision curve analysis indicated the clinical utility of the nomogram model across a wide range of thresholds, demonstrating its practicality and potential for clinical benefit.ConclusionThis study identified Prognostic Nutritional Index (PNI), surgical duration, age, and history of Diabetes mellitus as risk factors for the occurrence of Clavien-Dindo ≥ II grade complications in elderly patients with advanced gastric cancer after surgery. Based on these four risk factors, a nomogram risk prediction model was constructed. This model can be used to personalize the prediction of the risk of Clavien-Dindo ≥ II grade complications occurring after surgery in elderly patients with advanced gastric cancer.

Attribute Value Weighted Averaged One-Dependence Estimators with Kullback–Leibler Divergence

The Averaged One-Dependence Estimators (AODE) algorithm is an improvement of the naive Bayes algorithm, which allows all the attributes dependent on one common attribute, called parent attribute, thus forming One-Dependence Estimators (ODE). The classification probability is estimated by averaging the conditional probability of the ODE. When there is a dependency relationship between attributes, the AODE algorithm can better capture these relationships, thus improving classification performance. The AODE algorithm treats the parent and child attribute values equally in different ODEs. However, the parent attribute value and the child attribute value in different ODEs have different importance for classification. In this paper, two attribute value weighted AODE based on Kullback–Leibler divergence are proposed, one is parent attribute value weighted AODE, and the other is child attribute value weighted AODE. Comparative experiments were carried out with 30 datasets in the UCI database, and the experiments indicate that the performance of the parent attribute value weighted AODE algorithm with Kullback–Leibler divergence is significantly better than the original AODE algorithm, and its performance is also better than the mutual information weighted AODE algorithm.

Development and validation of a clinical prediction model for osteonecrosis of the jaw in patients receiving zoledronic acid using FAERS and canadian databases.

Osteonecrosis of the jaw (ONJ) stands as a severe complication linked to the use of bisphosphonates, particularly zoledronic acid, which is widely prescribed for managing conditions like osteoporosis and bone metastasis. This study is geared towards the development and validation of a clinical prediction model for ONJ in patients undergoing zoledronic acid treatment. We harnessed data from the FDA Adverse Event Reporting System (FAERS) as our training dataset, while the Canada Vigilance Adverse Reaction (CVAR) database served as the testing dataset. The study encompassed patients treated with zoledronic acid and subsequently diagnosed with ONJ. We analysed a range of predictive factors, including breast cancer, bone metastasis, osteoporosis, vitamin D and calcium levels, comorbidities, the number of concomitant medications, dosage, age, weight, and gender. Logistic regression and nomogram analysis were the chosen methodologies for constructing the predictive model. To evaluate the model's performance, we utilized receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). The study encompassed a total of 2,126 patients in the training cohort, 911 patients in the internal test cohort from the FAERS database, and 121 patients in the external test cohort from the CVAR database. Notable predictors for ONJ included bone metastasis (OR: 1.65, 95% CI: 1.22-2.24), osteoporosis (OR: 0.33, 95% CI: 0.21-0.52), the number of concomitant medications (OR: 1.07, 95% CI: 1.05-1.09), and the dosage of zoledronic acid (OR: 1.24, 95% CI: 1.10-1.39). The nomogram exhibited robust discriminatory power, evidenced by an area under the curve (AUC) of 0.77 in the training cohort, 0.76 in the internal test cohort, and 0.90 in the external test cohort. Calibration plots demonstrated a strong alignment between observed and predicted probabilities. Furthermore, DCA highlighted the prediction model's significant net benefit across various threshold probabilities. By leveraging data from both the FAERS and Canadian databases, this study has successfully developed and validated a clinical prediction model for ONJ in patients receiving zoledronic acid. This model stands as a valuable tool for clinicians, enabling them to pinpoint high-risk patients and make evidence-based treatment decisions to minimize the risk of ONJ.

Identifying Key Genes as Progression Indicators of Prostate Cancer with Castration Resistance Based on Dynamic Network Biomarker Algorithm and Weighted Gene Correlation Network Analysis.

Background: Androgen deprivation therapy (ADT) is the mainstay of treatment for prostate cancer, yet dynamic molecular changes from hormone-sensitive to castration-resistant states in patients treated with ADT remain unclear. Methods: In this study, we combined the dynamic network biomarker (DNB) method and the weighted gene co-expression network analysis (WGCNA) to identify key genes associated with the progression to a castration-resistant state in prostate cancer via the integration of single-cell and bulk RNA sequencing data. Based on the gene expression profiles of CRPC in the GEO dataset, the DNB method was used to clarify the condition of epithelial cells and find out the most significant transition signal DNB modules and genes included. Then, we calculated gene modules associated with the clinical phenotype stage based on the WGCNA. IHC was conducted to validate the expression of the key genes in CRPC and primary PCa patients Results:Nomograms, calibration plots, and ROC curves were applied to evaluate the good prognostic accuracy of the risk prediction model. Results: By combining single-cell RNA sequence data and bulk RNA sequence data, we identified a set of DNBs, whose roles involved in androgen-associated activities indicated the signals of a prostate cancer cell transition from an androgen-dependent state to a castration-resistant state. In addition, a risk prediction model including the risk score of four key genes (SCD, NARS2, ALDH1A1, and NFXL1) and other clinical-pathological characteristics was constructed and verified to be able to reasonably predict the prognosis of patients receiving ADT. Conclusions: In summary, four key genes from DNBs were identified as potential diagnostic markers for patients treated with ADT and a risk score-based nomogram will facilitate precise prognosis prediction and individualized therapeutic interventions of CRPC.

Machine learning models to predict 30-day mortality for critical patients with myocardial infarction: a retrospective analysis from MIMIC-IV database.

The identification of efficient predictors for short-term mortality among patients with myocardial infarction (MI) in coronary care units (CCU) remains a challenge. This study seeks to investigate the potential of machine learning (ML) to improve risk prediction and develop a predictive model specifically tailored for 30-day mortality in critical MI patients. This study focused on MI patients extracted from the Medical Information Mart for Intensive Care-IV database. The patient cohort was randomly stratified into derivation (n = 1,389, 70%) and validation (n = 595, 30%) groups. Independent risk factors were identified through eXtreme Gradient Boosting (XGBoost) and random decision forest (RDF) methodologies. Subsequently, multivariate logistic regression analysis was employed to construct predictive models. The discrimination, calibration and clinical utility were assessed utilizing metrics such as receiver operating characteristic (ROC) curve, calibration plot and decision curve analysis (DCA). A total of 1,984 patients were identified (mean [SD] age, 69.4 [13.0] years; 659 [33.2%] female). The predictive performance of the XGBoost and RDF-based models demonstrated similar efficacy. Subsequently, a 30-day mortality prediction algorithm was developed using the same selected variables, and a regression model was visually represented through a nomogram. In the validation group, the nomogram (Area Under the Curve [AUC]: 0.835, 95% Confidence Interval [CI]: [0.774-0.897]) exhibited superior discriminative capability for 30-day mortality compared to the Sequential Organ Failure Assessment (SOFA) score [AUC: 0.735, 95% CI: (0.662-0.809)]. The nomogram (Accuracy: 0.914) and the SOFA score (Accuracy: 0.913) demonstrated satisfactory calibration. DCA indicated that the nomogram outperformed the SOFA score, providing a net benefit in predicting mortality. The ML-based predictive model demonstrated significant efficacy in forecasting 30-day mortality among MI patients admitted to the CCU. The prognostic factors identified were age, blood urea nitrogen, heart rate, pulse oximetry-derived oxygen saturation, bicarbonate, and metoprolol use. This model serves as a valuable decision-making tool for clinicians.

Construction and validation of a nomogram for predicting cancer-specific survival in middle-aged patients with advanced hepatocellular carcinoma: A SEER-based study.

Hepatocellular carcinoma is the predominant form of primary liver cancer and is the leading cause of cancer-related death. The aim of this study was to construct a nomogram to predict cancer-specific survival (CSS) in middle-aged patients with advanced hepatocellular carcinoma. Clinical data were downloaded from the Surveillance, Epidemiology and End Results (SEER) database for middle-aged patients diagnosed with advanced hepatocellular carcinoma (AJCC stage III and IV) from 2000 to 2019. The patients were randomized in a 7:3 ratio into training cohort and validation cohort. Univariate and multivariate Cox regression analyses were performed in the training cohort to screen for independent risk factors associated with cancer-specific survival for the construction of nomogram. The nomogram was examined and evaluated using the consistency index (C-index), area under the curve (AUC), and calibration plots. The clinical application value of the model was evaluated using decision curve analysis (DCA). A total of 3026 patients were selected, including 2244 in the training cohort and 962 in the validation cohort. Multivariate analysis revealed gender, marital status, American Joint Committee on Cancer (AJCC) stage, tumor size, bone metastasis, lung metastasis, alpha-fetoprotein (AFP) level, surgery, radiotherapy, chemotherapy as independent risk factors, which were all included in the construction of the nomogram. In the training cohort, the AUC values were 0.74 (95% CI: 0.76-0.72), 0.78 (95% CI: 0.82-0.75), and 0.82 (95% CI: 0.86-0.78) at 1-, 3-, and 5-year CSS, respectively. The calibration plots showed good consistency between the actual and predicted values. The DCA curves indicated that the nomogram model could more accurately predict CSS at 1-, 3-, and 5-year in middle-aged patients with advanced hepatocellular carcinoma compared with the AJCC staging system. Highly similar results to the training cohort were also observed in the validation cohort. In the risk stratification system, good differentiation was shown between the 2 groups, and Kaplan-Meier survival analysis indicated that surgery could prolong patient survival. In this study, we developed a nomogram and risk stratification system for predicting CSS in middle-aged patients with advanced hepatocellular carcinoma. The prediction model has good predictive performance and can help clinicians in judging prognosis and clinical decision making.

Bicenter validation of a risk model for the preoperative prediction of extraprostatic extension of localized prostate cancer combining clinical and multiparametric MRI parameters

BackgroundThis study aimed to validate a previously published risk model (RM) which combines clinical and multiparametric MRI (mpMRI) parameters to predict extraprostatic extension (EPE) of prostate cancer (PC) prior to radical prostatectomy (RP).Materials and methodsA previously published RM combining clinical with mpMRI parameters including European Society of Urogenital Radiology (ESUR) classification for EPE was retrospectively evaluated in a cohort of two urological university hospitals in Germany. Consecutive patients (n = 205, January 2015 –June 2021) with available preoperative MRI images, clinical information including PSA, prostate volume, ESUR classification for EPE, histopathological results of MRI-fusion biopsy and RP specimen were included. Validation was performed by receiver operating characteristic analysis and calibration plots. The RM‘s performance was compared to ESUR criteria.ResultsHistopathological T3 stage was detected in 43% of the patients (n = 89); 45% at Essen and 42% at Düsseldorf. Discrimination performance between pT2 and pT3 of the RM in the entire cohort was AUC = 0.86 (AUC = 0.88 at site 1 and AUC = 0.85 at site 2). Calibration was good over the entire probability range. The discrimination performance of ESUR classification alone was comparable (AUC = 0.87).ConclusionsThe RM showed good discriminative performance to predict EPE for decision-making for RP as a patient-tailored risk stratification. However, when experienced MRI reading is available, standardized MRI reading with ESUR scoring is comparable regarding information outcome. A main limitation is the potentially limited transferability to other populations because of the high prevalence of EPE in our subgroups.

Golf Club Selection with AI-Based Game Planning.

In the dynamic realm of golf, where every swing can make the difference between victory and defeat, the strategic selection of golf clubs has become a crucial factor in determining the outcome of a game. Advancements in artificial intelligence have opened new avenues for enhancing the decision-making process, empowering golfers to achieve optimal performance on the course. In this paper, we introduce an AI-based game planning system that assists players in selecting the best club for a given scenario. The system considers factors such as distance, terrain, wind strength and direction, and quality of lie. A rule-based model provides the four best club options based on the player's maximum shot data for each club. The player picks a club, shot, and target and a probabilistic classification model identifies whether the shot represents a birdie opportunity, par zone, bogey zone, or worse. The results of our model show that taking into account factors such as terrain and atmospheric features increases the likelihood of a better shot outcome.

External validation of the parsimonious EuroLung risk models: analysis of the Brazilian Lung Cancer Registry.

The purpose of this study was to assess performance in the Brazilian Lung Cancer Registry Database by using the parsimonious EuroLung risk models for morbidity and mortality. The EuroLung1 and EuroLung2 models were tested and evaluated through calibration (calibration plot, Brier score, and the Hosmer-Lemeshow test) and discrimination (ROC AUCs), in a national multicenter registry of 1,031 patients undergoing anatomic lung resection. The evaluation of performance in Brazilian health care facilities utilizing risk-adjustment models, specifically EuroLung1 and EuroLung2, revealed substantial miscalibration, as evidenced by calibration plots and Hosmer-Lemeshow tests in both models. In terms of calibration, EuroLung1 exhibited a calibration plot with overlapping points, characterized by a slope of 1.11 and a Brier score of 0.15; the Hosmer-Lemeshow test yielded a statistically significant p-value of 0.015; and the corresponding ROC AUC was 0.678 (95% CI: 0.636-0.721). The EuroLung2 model displayed better calibration, featuring fewer overlapping points in the calibration plot, with a slope of 1.22, with acceptable discrimination, as indicated by a ROC AUC of 0.756 (95% CI: 0.670-0.842). Both models failed to accurately predict morbidity and mortality outcomes in this specific health care context. Discrepancies between the EuroLung model predictions and outcomes in Brazil underscore the need for model refinement and for a probe into inefficiencies in the Brazilian health care system.

Construction and validation of a dynamic nomogram using Lasso-logistic regression for predicting the severity of severe fever with thrombocytopenia syndrome patients at admission

BackgroundSevere fever with thrombocytopenia syndrome (SFTS) is a highly fatal infectious disease caused by the SFTS virus (SFTSV), posing a significant public health threat. This study aimed to construct a dynamic model for the early identification of SFTS patients at high risk of disease progression.MethodsAll eligible patients enrolled between April 2014 and July 2023 were divided into training and validation sets. Thirty-four clinical variables in the training set underwent analysis using least absolute shrinkage and selection operator (LASSO) logistic regression. Selected variables were then input into the multivariate logistic regression model to construct a dynamic nomogram. The model’s performance was assessed using the area under the receiver operating characteristic curve (AUC-ROC), concordance index (C-index), calibration curve, and decision curve analysis (DCA) in both training and validation sets. Kaplan-Meier survival analysis was utilized to evaluate prognostic performance.Results299 SFTS patients entered the final investigation, with 208 patients in the training set and 90 patients in the validation set. LASSO and the multivariate logistic regression identified six significant prediction factors: age (OR, 1.060; 95% CI, 1.017–1.109; P = 0.007), CREA (OR, 1.017; 95% CI, 1.003–1.031; P = 0.019), PT (OR, 1.765; 95% CI, 1.175–2.752; P = 0.008), D-dimer (OR, 1.039; 95% CI, 1.005–1.078; P = 0.032), nervous system symptoms (OR, 8.244; 95% CI, 3.035–26.858; P < 0.001) and hemorrhage symptoms (OR, 3.414; 95% CI, 1.096–10.974; P = 0.035). The AUC-ROC, C-index, calibration plots, and DCA demonstrated the robust performance of the nomogram in predicting severity at admission, and Kaplan-Meier survival analysis indicated its utility in predicting 28-day mortality among SFTS patients. The dynamic nomogram is accessible at https://sfts.shinyapps.io/SFTS_severity_nomogram/.ConclusionThis study provided a practical and readily applicable tool for the early identification of high-risk SFTS patients, enabling the timely initiation of intensified treatments and protocol adjustments to mitigate disease progression.

Development and validation of a nomogram for predicting the interstitial fibrosis and tubular atrophy in patients with lupus nephritis.

Interstitial fibrosis and tubular atrophy (IFTA) were frequent histologic features of lupus nephritis (LN), and LN patients with IFTA have poor renal outcomes. In this study, we aimed to construct prediction models for the IFTA in LN patients. This retrospective study included 303 patients with biopsy proven LN at the Affiliated Hospital of Qingdao University and Union Hospital of Fujian Medical University. The participants were randomly divided into development and validation cohorts. They were further divided into IFTA and non-IFTA groups. The least absolute shrinkage and selection operator (LASSO) regression model with laboratory test results collected at the time of kidney biopsy was used to optimize feature selection for the risk model. Multivariable logistic regression analysis was applied to build a predicting model incorporating the feature selected in the LASSO regression model. Discrimination, calibration, and clinical usefulness of the predicting model were assessed using the C-index, calibration plot, and ROC curve analysis. Internal validation was assessed using the bootstrapping validation. A nomogram for individual assessment was constructed based on the preferable model. Predictors contained in the prediction nomogram included age, body mass index (BMI), mean arterial pressure (MAP), logANA, C3, eGFR and serum uric acid. The model displayed good discrimination with a C-index of 0.794 (95% CI 0.734-0.854) and good calibration. High C-index value of 0.857 (95% CI 0.776-0.938) could still be reached in the interval validation. A nomogram model based on the LASSO model was created for producing a probability score of IFTA in LN patients. With excellent predictive abilities, the nomogram may provide a simple and reliable tool to distinguish LN patients with IFTA and helps physicians make clinical decisions in their comprehensive assessment.

Impact of preoperative inflammatory and nutritional markers on the prognosis of patients with peritoneal metastasis of colorectal cancer.

Identifying patients with peritoneal metastasis (PMs) of colorectal cancer (CRC) who will benefit from cytoreductive surgery and hyperthermic intraperitoneal chemotherapy is crucial before surgery. Inflammatory and nutritional indicators play essential roles in cancer development and metastasis. To investigate the association of preoperative inflammatory and nutritional markers with prognosis in patients with CRC-PM. We included 133 patients diagnosed with CRC-PM between July 2012 and July 2018. Patients' demographics, overall survival (OS), and preoperative inflammatory and nutritional markers were evaluated. The Kaplan-Meier method and log-rank test were used to estimate differences. Of the 133 patients, 94 (70.6%) had normal hemoglobin (Hb) and 54 (40.6%) had a high neutrophil-to-lymphocyte ratio (NLR). The median OS (mOS) was significantly lower for patients with high NLR (7.9 months) than for those with low NLR (25.4 months; P = 0.002). Similarly, patients with normal Hb had a longer mOS (18.5 months) than those with low Hb (6.3 months; P < 0.001). Multivariate analysis identified age, carbohydrate antigen 199 levels, NLR, Hb, and peritoneal cancer index as independent predictors of OS. Based on these findings, a nomogram was constructed, which demonstrated a good capacity for prediction, with a C-index of 0.715 (95% confidence interval: 0.684-0.740). Furthermore, the 1- and 2-year survival calibration plots showed good agreement between predicted and actual OS rates. The areas under the curve for the 1- and 2-year survival predictions of the nomogram were 0.6238 and 0.6234, respectively. High NLR and low Hb were identified as independent predictive risk factors for poor prognosis in patients with CRC-PM. The established nomogram demonstrated high accuracy in predicting OS for patients with CRC-PM, indicating its potential as a valuable prognostic tool for this patient population.

Novel nomogram for the prediction of sepsis-induced coagulopathy in the PICU: A multicentre retrospective study

IntroductionSepsis-induced coagulopathy (SIC) is a severe complication of sepsis, characterized by poor prognosis and high mortality. However, the predictors of SIC in pediatric patients have yet to be identified. Our aim was to develop a user-friendly and efficient nomogram for predicting SIC in sepsis patients admitted to the pediatric intensive care unit (PICU). Materials and methodsWe screened 948 sepsis patients admitted to the PICU in three hospitals located in Shandong, China. Least absolute shrinkage and selector operation (LASSO) regression was used in the training cohort for variable selection and regularization. The selected variables were utilized to construct a nomogram for predicting the risk of SIC among sepsis patients admitted to the PICU. ResultsOverall, SIC was observed in 324 (40.3 %) patients. The morbidity of SIC in sepsis patients is associated with age, fibrinogen, prothrombin time, C-reactive protein, lactate and the pediatric sequential organ failure assessment score. We developed a nomogram for the early identification of SIC in the training cohort (area under the curve [AUC] 0.869, 95 % confidence interval [CI] 0.830–0.907, sensitivity 75.7 %, specificity 84.8 %) and validation cohorts (validation cohort 1: AUC 0.854, 95 % CI 0.805–0.903, sensitivity 72.0 %, specificity 86.9 %; validation cohort 2: AUC 0.853, 95 % CI 0.796–0.910, sensitivity 70.1 %, specificity 87.8 %). The calibration plots of the nomogram demonstrated a high level of concordance in the SIC probabilities between the observed and predicted values. ConclusionsThe novel nomogram showed excellent predictive performance for the morbidity of SIC among sepsis patients admitted to the PICU, potentially assisting healthcare professionals in early identification and intervention for SIC.

Conventional machine learning-based prediction models did not outperform the International IgA Nephropathy Prediction Tool.

Immunoglobulin A nephropathy (IgAN) is a major cause of end-stage kidney disease (ESKD). The International IgA Nephropathy Prediction Tool (IIgAN-PT) predicts IgAN prognosis, but improvement in the prediction performance using machine learning (ML)-based methods is needed. We analyzed 4,425 biopsy-confirmed patients with IgAN and ≥6 months of follow-up from nine tertiary university hospitals in Korea. The study population was divided into development and validation cohorts. Using the collected 87 clinicodemographic and pathological variables, ML-based prediction models for ESKD or estimated glomerular filtration rate were constructed: 1) the conventional CatBoost model, 2) the optimized CatBoost model with Cox proportional hazards, 3) the deep Cox proportional hazards model, and 4) the deep Cox mixture model. The area under the curve (AUC) and calibration plots were used to investigate the discriminative and calibration performance of the models, which were then compared with those of the IIgAN-PT full model. The full model showed excellent performance (AUC [95% confidence interval] for 5-year outcome, 0.896 [0.8530.940]), with acceptable calibration results. The ML-based models showed good performance in predicting adverse kidney outcomes and revealed acceptable discrimination performance in the external validation (AUC [95% confidence interval] for the 5-year outcome: 1) 0.829 [0.791-0.866]; 2) 0.847 [0.804-0.890]; 3) 0.823 [0.784-0.862]; and 4) 0.832 [0.794-0.870]), although they underestimated the external validation cohort risks. With the validation data, the overall performance of the IIgAN-PT was non-inferior to that of the ML-based model. Conclusions: Our ML-based models showed good performance in predicting adverse kidney outcomes in patients with IgAN but they did not outperform the IIgAN-PT.

Implementing machine learning to predict survival outcomes in patients with resected pulmonary large cell neuroendocrine carcinoma

ABSTRACT Background The post-surgical prognosis for Pulmonary Large Cell Neuroendocrine Carcinoma (PLCNEC) patients remains largely unexplored. Developing a precise prognostic model is vital to assist clinicians in patient counseling and creating effective treatment strategies. Research design and methods This retrospective study utilized the Surveillance, Epidemiology, and End Results database from 2000 to 2018 to identify key prognostic features for Overall Survival (OS) in PLCNEC using Boruta analysis. Predictive models employing XGBoost, Random Forest, Decision Trees, Elastic Net, and Support Vector Machine were constructed and evaluated based on Area Under the Receiver Operating Characteristic Curve (AUC), calibration plots, Brier scores, and Decision Curve Analysis (DCA). Results Analysis of 604 patients revealed eight significant predictors of OS. The Random Forest model outperformed others, with AUC values of 0.765 and 0.756 for 3 and 5-year survival predictions in the training set, and 0.739 and 0.706 in the validation set, respectively. Its superior validation cohort performance was confirmed by its AUC, calibration, and DCA metrics. Conclusions This study introduces a novel machine learning-based prognostic model with a supportive web-based platform, offering valuable tools for healthcare professionals. These advancements facilitate more personalized clinical decision-making for PLCNEC patients following primary tumor resection.

Predictive etiological classification of acute ischemic stroke through interpretable machine learning algorithms: a multicenter, prospective cohort study

BackgroundThe prognosis, recurrence rates, and secondary prevention strategies varied significantly among different subtypes of acute ischemic stroke (AIS). Machine learning (ML) techniques can uncover intricate, non-linear relationships within medical data, enabling the identification of factors associated with etiological classification. However, there is currently a lack of research utilizing ML algorithms for predicting AIS etiology.ObjectiveWe aimed to use interpretable ML algorithms to develop AIS etiology prediction models, identify critical factors in etiology classification, and enhance existing clinical categorization.MethodsThis study involved patients with the Third China National Stroke Registry (CNSR-III). Nine models, which included Natural Gradient Boosting (NGBoost), Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Light Gradient Boosting Machine (LGBM), Gradient Boosting Decision Tree (GBDT), Adaptive Boosting (AdaBoost), Support Vector Machine (SVM), and logistic regression (LR), were employed to predict large artery atherosclerosis (LAA), small vessel occlusion (SVO), and cardioembolism (CE) using an 80:20 randomly split training and test set. We designed an SFS-XGB with 10-fold cross-validation for feature selection. The primary evaluation metrics for the models included the area under the receiver operating characteristic curve (AUC) for discrimination and the Brier score (or calibration plots) for calibration.ResultsA total of 5,213 patients were included, comprising 2,471 (47.4%) with LAA, 2,153 (41.3%) with SVO, and 589 (11.3%) with CE. In both LAA and SVO models, the AUC values of the ML models were significantly higher than that of the LR model (P < 0.001). The optimal model for predicting SVO (AUC [RF model] = 0.932) outperformed the optimal LAA model (AUC [NGB model] = 0.917) and the optimal CE model (AUC [LGBM model] = 0.846). Each model displayed relatively satisfactory calibration. Further analysis showed that the optimal CE model could identify potential CE patients in the undetermined etiology (SUE) group, accounting for 1,900 out of 4,156 (45.7%).ConclusionsThe ML algorithm effectively classified patients with LAA, SVO, and CE, demonstrating superior classification performance compared to the LR model. The optimal ML model can identify potential CE patients among SUE patients. These newly identified predictive factors may complement the existing etiological classification system, enabling clinicians to promptly categorize stroke patients’ etiology and initiate optimal strategies for secondary prevention.

Bayesian seismic 4D inversion for lithology and fluid prediction

Seismic data acquired at different times over the same area can provide insight into changes in an oil/gas reservoir. Probabilities for pore fluid will typically change while the lithology remains stable over time. This implies significant correlations across the vintages. We have developed a methodology for Bayesian prediction of joint probabilities for discrete lithology-fluid classes (LFCs) for two vintages, simultaneously considering the seismic AVO data of both vintages. By taking into account cross-vintage correlations of elastic and seismic properties, the simultaneous invertion ensures that the individual results for both vintages as well as their differences are consistent and constrained by the seismic data of both vintages. The method relies on prior geological knowledge of stratigraphic layering, possible lithologies and fluids within each layer, and possible cross-vintage changes in lithology and pore fluid. Multiple LFCs can be used to represent different strengths of dynamic cross-vintage changes. We have tested the algorithm on a synthetic data set and on data from the Edvard Grieg field in the central North Sea. Synthetic results demonstrate that the algorithm is able to use dual-vintage data together with a prior model specifying their correlations to calculate joint LFC posterior probabilities for both vintages with a lower degree of uncertainty than independent single-vintage inversions. The Edvard Grieg results indicate that the underlying model is sufficiently general to explain 4D variations in seismic data using a reasonably simple prior model of 4D LFC changes.

SP6.8 - Predicting Determinants of Successful Transplantation Using Supra-Marginal (DRI-Donor Risk Index &amp;gt;1.5) Deceased Donor Kidneys

Abstract Objective To develop machine-learning-models with national registry data to predict factors associated with favourable outcomes using supra-marginal ( DRI -Donor Risk Index &amp;gt;1,5) deceased donor kidneys Design A retrospective cohort study using UK Transplant Registry data from 2000-2019 Setting National registry administered by NHS Blood and Transplant in the United Kingdom. Participants Adult recipients (n = 6254) of first kidney-alone transplants from very supra-marginal deceased-donors (DRI≥1.5) Main Outcome Measures Death-censored graft failure and patient mortality Predictors Comprehensive recipient, donor, and transplant characteristics. Statistical Analysis Bayesian neural networks, gradient boosting machines, random forest, and SMOTE-balanced bagging classifiers tuned using Bayesian optimisation Cox regression, competing risk analysis, and calibration plots Results Overall 5-year graft survival was 81%. The random-forest-model had excellent predictive performance for graft failure (AUC 0.88, 95% CI 0.87-0.89; RMSE 0.29). recipient age &amp;gt; 75-years (SHR 1.02, 95% CI 1.01-1.03), recipient-BMI &amp;gt;30 (SHR 1.04, 95% CI 1.02-1.07), HLA mismatches &amp;gt;4 (SHR 1.09, 95% CI 1.01-1.17), donor-creatinine &amp;gt;120mmol/L (SHR 1.002, 95% CI 1.001-1.003), and rejection- within 3-months (SHR 1.56, 95% CI 1.32-1.85) were key determinants of poor survival. Prolonged cold-ischemia-time &amp;gt;14hrs (SHR 1.01, 95% CI 1.007-1.015) was detrimental Conclusions Supra-marginal deceased-donor-kidneys can achieve excellent 5-year outcomes with careful recipient selection. Machine-learning accurately predicted factors associated with success. Adjusted predictors of graft failureVariableSHR95% CIp-valueRecipient-age &amp;gt; 75 years1.021.01-1.03&amp;lt;0.001Recipient-BMI &amp;gt;301.041.02-1.070.002HLA-mismatches &amp;gt;41.091.01-1.170.03Creatinine at offer &amp;gt;120 mmol/L/ 1.4 mg/dl1.0021.001-1.003&amp;lt;0.001Rejection at 3 month1.561.32-1.85&amp;lt;0.001Cold-ischemia-time &amp;gt;14 hrs1.011.007-1.015&amp;lt;0.001SHR: subdistribution hazard ratio;

Clinical model to predict the risk of nonalcoholic fatty liver disease: A secondary analysis of data from a cross-sectional study.

This study aimed to develop and validate a clinical model for predicting the risk of nonalcoholic fatty liver disease (NAFLD) by using data from a cross-sectional study. This investigation utilized data from the Dryad database and employed multivariable logistic regression analysis, restricted cubic spline, and nomogram analysis to achieve comprehensive insights. The discrimination and calibration of the nomogram were evaluated using the receiver operating characteristic curve and calibration plot. A total of 1072 patients were included in the study, including 456 with non-NAFLD and 616 with NAFLD. Significant differences were observed in terms of sex, body mass index (BMI), tobacco, hypertension, diabetes, alanine aminotransferase (ALT), aspartate aminotransferase (AST), ALT/AST ratio, uric acid (UA), fasting blood glucose (FBG), triglyceride (TG), high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, systolic blood pressure, and diastolic blood pressure (P < .05 for all comparisons). Multivariable logistic regression analysis indicated that sex, BMI, diabetes, ALT/AST ratio, UA, FBG, and TG were associated with an increased risk of NAFLD. Restricted cubic spline indicated a nonlinear relationship between the risk of NAFLD and variables including ALT/AST ratio, FPG, TG, and UA (P for nonlinearity < .01). The variables in the nomogram included BMI, diabetes, ALT/AST ratio, UA, FBG, and TG. The value of area under the curve was 0.790, indicating that the nomogram prediction model exhibited significant discriminatory accuracy. A reliable clinical model for predicting the risk of NAFLD was developed using readily available clinical data. The model can assist clinicians in identifying individuals with an increased risk of NAFLD, enabling early interventions for preventing and managing this prevalent liver disease.

A Vision-Language Model-Based Traffic Sign Detection Method for High-Resolution Drone Images: A Case Study in Guyuan, China.

As a fundamental element of the transportation system, traffic signs are widely used to guide traffic behaviors. In recent years, drones have emerged as an important tool for monitoring the conditions of traffic signs. However, the existing image processing technique is heavily reliant on image annotations. It is time consuming to build a high-quality dataset with diverse training images and human annotations. In this paper, we introduce the utilization of Vision-language Models (VLMs) in the traffic sign detection task. Without the need for discrete image labels, the rapid deployment is fulfilled by the multi-modal learning and large-scale pretrained networks. First, we compile a keyword dictionary to explain traffic signs. The Chinese national standard is used to suggest the shape and color information. Our program conducts Bootstrapping Language-image Pretraining v2 (BLIPv2) to translate representative images into text descriptions. Second, a Contrastive Language-image Pretraining (CLIP) framework is applied to characterize not only drone images but also text descriptions. Our method utilizes the pretrained encoder network to create visual features and word embeddings. Third, the category of each traffic sign is predicted according to the similarity between drone images and keywords. Cosine distance and softmax function are performed to calculate the class probability distribution. To evaluate the performance, we apply the proposed method in a practical application. The drone images captured from Guyuan, China, are employed to record the conditions of traffic signs. Further experiments include two widely used public datasets. The calculation results indicate that our vision-language model-based method has an acceptable prediction accuracy and low training cost.