Multi-objective Evaluation and Optimization of Stochastic Gradient Boosting Machines for Genomic Prediction and Selection in Wheat (Triticum aestivum) Breeding

Although the rate of pediatric postoperative mortality is low, the development and validation of perioperative risk assessment models have allowed for the stratification of those at highest risk, including the Pediatric Risk Assessment (PRAm) score. The clinical application of such tools requires manual data entry, which may be inaccurate or incomplete, compromise efficiency, and increase physicians' clerical obligations. We aimed to create an electronically derived, automated PRAm score and to evaluate its agreement with the original American College of Surgery National Surgical Quality Improvement Program (ACS NSQIP)-derived and validated score. We performed a retrospective observational study of children <18 years who underwent noncardiac surgery from 2017 through 2021 at Boston Children's Hospital (BCH). An automated PRAm score was developed via electronic derivation of International Classification of Disease (ICD) -9 and -10 codes. The primary outcome was agreement and correlation among PRAm scores obtained via automation, NSQIP data, and manual physician entry from the same BCH cohort. The secondary outcome was discriminatory ability of the 3 PRAm versions. Fleiss Kappa, Spearman correlation (rho), and intraclass correlation coefficient (ICC) and receiver operating characteristic (ROC) curve analyses with area under the curve (AUC) were applied accordingly. Of the 6014 patients with NSQIP and automated PRAm scores (manual scores: n = 5267), the rate of 30-day mortality was 0.18% (n = 11). Agreement and correlation were greater between the NSQIP and automated scores (rho = 0.78; 95% confidence interval [CI], 0.76-0.79; P <.001; ICC = 0.80; 95% CI, 0.79-0.81; Fleiss kappa = 0.66; 95% CI, 0.65-0.67) versus the NSQIP and manual scores (rho = 0.73; 95% CI, 0.71-0.74; P < .001; ICC = 0.78; 95% CI, 0.77-0.79; Fleiss kappa = 0.56; 95% CI, 0.54-0.57). ROC analysis with AUC showed the manual score to have the greatest discrimination (AUC = 0.976; 95% CI, 0.959,0.993) compared to the NSQIP (AUC = 0.904; 95% CI, 0.792-0.999) and automated (AUC = 0.880; 95% CI, 0.769-0.999) scores. Development of an electronically derived, automated PRAm score that maintains good discrimination for 30-day mortality in neonates, infants, and children after noncardiac surgery is feasible. The automated PRAm score may reduce the preoperative clerical workload and provide an efficient and accurate means by which to risk stratify neonatal and pediatric surgical patients with the goal of improving clinical outcomes and resource utilization.

To assess which of the formulae for estimating renal function (Cockcroft-Gault (CG), Chronic Kidney Disease Epidemiology Collaboration CKD-EPI) and Modification of Diet in Renal Disease (MDRD)) provides the most accurate prediction of minimum vancomycin concentration (Cmin) and to evaluate whether they can be interchanged to optimise vancomycin dosage. An observational and retrospective study was undertaken in hospitalised adult patients treated with intravenous vancomycin. Patients with serum creatinine (Scr) >2 mg/dL and <0.5 mg/dL, body mass index >40 kg/m2, need for extracorporeal clearance techniques and unstable renal function were excluded. Bayesian analysis was used to obtain individual pharmacokinetic parameters. Vancomycin clearance (CLvan) was calculated by means of CG (eCLvanCG), CKD-EPI (eCLvanCKD-EPI) and MDRD (eCLvanMDRD) and used to obtain Cmin estimates (eCmin). eCmin and observed Cmin were compared using an intraclass correlation coefficient (ICC). A post-hoc analysis by subgroups (age, sex, weight, Scr and estimated glomerular filtration rate (eGFR)) was performed. From each eCLvan, the area under the curve (AUC) was calculated and categorised as AUC <400 mg*hour/L, AUC 400-600 mg*hour/L and AUC >600 mg*hour/L. The kappa coefficient was applied to study AUC concordance. A total of 228 patients (69.3% men) were included. eCminCG had a statistically significant better agreement with Cmin (ICC >0.7) and showed good agreement in almost all subgroups. Patients with Scr >1.1 mg/dL were the only subgroup in which eCminMDRD and eCminCKD-EPI had an adequate ICC with no statistically significant differences compared with eCminCG. eCminMDRD had a similar ICC to eCminCG in the eGFR <60 mL/min and age 46-75 years subgroups. Kappa values showed regular agreement in all subgroups: 0.32 (AUC <400 mg*hour/L), 0.24 (AUC 400-600 mg*hour/L) and 0.41 (AUC >600 mg*hour/L). The CG formula provides the most accurate prediction of vancomycin Cmin. In patients with eGFR <60 mL/min and aged 46-75 years, MDRD also shows a good predictive capacity. However, in low weight and elderly patients, Cmin predictions are superior with CG. Therefore, renal function estimation equations should not be considered interchangeable for vancomycin dose adjustments.

Predicting traditional Chinese medicine constitutions in adults aged ≥ 65 years: A machine learning approach.

To determine if there is a comparative measurement on bilateral sunrise view radiographs that can help predict acute transient patellar dislocation (ATPD). A retrospective chart review from a single institution was conducted of two patient groups, a case group with ATPD diagnosed by MRI, and a control group with knee injury not involving the patella or medial retinaculum. Three readers blinded to the MRI diagnosis reviewed the sunrise view radiographs and reported three values for both knees: (1) medial trochlea - medial patella distance (MT-MP); (2) lateral trochlea -medial patella distance (LT-MP); and (3) medial patellofemoral angle (MPFA), (Image 2). Diagnostic accuracy was assessed utilizing empirical ROC curves and their corresponding area under the curves (AUC). Inter-reader reliability was determined using intraclass correlation coefficients (ICC). Sensitivity and specificity were calculated for varying differences in MPFA and MT-MP between affected knee and unaffected knee. Of the three measurements, increased MPFA measured on the injured knee was the most accurate predictor of ATPD with substantial inter-reader reliability (AUC 0.765, ICC 0.681). MT-MP (AUC 0.707, ICC 0.586) and LT-MP (AUC 0.593, 0.189) distances were less accurate and less reliable predictors. Asymmetric difference in the MPFA or MT-MP between affected knee and unaffected knee was a very specific, albeit not sensitive indicator of ATPD. Comparative sunrise view knee radiographs can help predict the diagnosis of ATPD with high specificity, prompting early MRI evaluation for definitive diagnosis, helping expedite treatment planning.

Background: The World Health Organization declared COVID-19 “a pandemic” on March 11, 2020. An essential preventive and screening strategy adopted was temperature screening at various public places such as airports and hospitals with devices like handheld, noncontact infrared thermometers. Before this pandemic, the industrial thermometers have never been used in health-care settings. The study was conducted to compare temperature readings of these industrial-grade infrared noncontact handheld thermometers with a digital thermometer. Materials and Methods: Three Testo® industrial-grade infrared noncontact thermometer (IGNCT) and Dr. Trust digital thermometer were used on 117 individuals for recording temperature. The average of three readings was taken for each instrument by a single investigator. Results: Out of 117 individuals, 14 (11.9%) were febrile. The Pearson correlation values between the digital thermometers varied from 0.3 to 0.42. The reliability, as measured by the intraclass correlation, was poor (0.16-“0.4). The area under the curve for IGNCT devices varied from 76% to 84%. Conclusion: Industrial-grade infrared noncontact temperature (IGNCT) recording devices with its no-touch technique to measure temperature while maintaining distance may be useful in a pandemic situation. However, the higher the area under the curve, the better it is. In a large cohort, 20% will be misclassified, and in absolute terms, they may be translated into large numbers. Hence, there is a case for better devices having an area under the curve of nearly 100% for the screening of population. The role of IGNCT devices for the screening of individuals to identify cases of febrile illness is limited owing to its poor reliability and unacceptable area under the curve for screening large populations.

In this paper, we propose new listwise learning-to-rank models that mitigate the shortcomings of existing ones. Existing listwise learning-to-rank models are generally derived from the classical Plackett-Luce model, which has three major limitations. (1) Its permutation probabilities overlook ties, i.e., a situation when more than one document has the same rating with respect to a query. This can lead to imprecise permutation probabilities and inefficient training because of selecting documents one by one. (2) It does not favor documents having high relevance. (3) It has a loose assumption that sampling documents at different steps is independent. To overcome the first two limitations, we model ranking as selecting documents from a candidate set based on unique rating levels in decreasing order. The number of steps in training is determined by the number of unique rating levels. We propose a new loss function and associated four models for the entire sequence of weighted classification tasks by assigning high weights to the selected documents with high ratings for optimizing Normalized Discounted Cumulative Gain (NDCG). To overcome the final limitation, we further propose a novel and efficient way of refining prediction scores by combining an adapted Vanilla Recurrent Neural Network (RNN) model with pooling given selected documents at previous steps. We encode all of the documents already selected by an RNN model. In a single step, we rank all of the documents with the same ratings using the last cell of the RNN multiple times. We have implemented our models using three settings: neural networks, neural networks with gradient boosting, and regression trees with gradient boosting. We have conducted experiments on four public datasets. The experiments demonstrate that the models notably outperform state-of-the-art learning-to-rank models.

Postherpetic neuralgia (PHN), a complication of herpes zoster, significantly impacts the quality of life of affected patients. Research indicates that early intervention for pain can reduce the occurrence or severity of PHN. This study aims to develop a predictive model and scoring table to identify patients at risk of developing PHN following acute herpetic neuralgia, facilitating informed clinical decision-making. We conducted a retrospective review of 524 hospitalized patients with herpes zoster at The First Affiliated Hospital of Zhejiang Chinese Medical University from December 2020 to December 2023 and classified them according to whether they had PHN, collecting a comprehensive set of 30 patient characteristics and disease-related indicators, 5 comorbidity indicators, 2 disease score values, and 10 serological indicators. Relevant features associated with PHN were identified using the least absolute shrinkage and selection operator (LASSO). Then, the patients were divided into a training set and a test set in a 4:1 ratio, with comparability tested using univariate analysis. Six models were established in the training set using machine learning methods: support vector machines, logistic regression, random forest, k-nearest neighbor, gradient boosting, and neural network. The performance of these models was evaluated in the test set, and a nomogram based on logistic regression was used to create a PHN prediction score table. Eight non-zero characteristic variables selected from the LASSO regression results were included in the model, including age [area under the curve (AUC)=0.812, p<0.001], Numerical Rating Scale (NRS) (AUC=0.792, p<0.001), receiving treatment time (AUC=0.612, p<0.001), rash recovery time (AUC=0.680, p<0.001), history of malignant tumor (AUC=0.539, p<0.001), history of diabetes (AUC=0.638, p<0.001), varicella-zoster virus immunoglobulin M (AUC=0.620, p<0.001), and serum nerve-specific enolase (AUC=0.659, p<0,001). The gradient boosting model outperformed other classifier models on the test set with an AUC of 0.931, 95% confidence interval (CI) (0.882-0.980), accuracy of 0.886 (95% CI 0.809-0.940). In the test set, our predictive scoring table achieved an AUC of 0.820 (95% CI 0.869-0.970) with accuracy of 0.790 (95% CI 0.700-0.864). This study presents a methodology for predicting the development of postherpetic neuralgia in shingles patients by analyzing historical case data, employing various machine learning techniques, and selecting the optimal model through comparative analysis. In addition, a logistic regression model has been used to create a scoring table for predicting the postherpetic neuralgia.

To determine whether radiomics features derived from diffusion-weighted imaging (DWI) and arterial spin labeling (ASL) can improve the differentiation between radiation-induced brain injury (RIBI) and tumor recurrence (TR) in glioma patients. A total of 4199 radiomics features were extracted from conventional MRI, apparent diffusion coefficient (ADC), and cerebral blood flow (CBF) maps, obtained from 96 pathologically confirmed WHO grade 2~4 gliomas with enhancement after standard treatment. The intraclass correlation coefficient (ICC) was used to test segmentation stability between two doctors. Radiomics features were selected using the Mann-Whitney U test, LASSO regression, and RFE algorithms. Four machine learning classifiers were adopted to establish radiomics models. The diagnostic performance of multiparameter, conventional, and single-parameter MRI radiomics models was compared using the area under the curve (AUC). The models were evaluated in the subsequent independent validation set (n = 30). Eight important radiomics features (3 from conventional MRI, 1 from ADC, and 4 from CBF) were selected. Support vector machine (SVM) was chosen as the optimal classifier. The diagnostic performance of the multiparameter MRI radiomics model (AUC 0.96) was higher than that of the conventional MRI (AUC 0.88), ADC (AUC 0.91), and CBF (AUC 0.95) radiomics models. For subgroup analysis, the multiparameter MRI radiomics model showed similar performance, with AUCs of 0.98 in WHO grade 2~3 and 0.96 in WHO grade 4. The incorporation of noninvasive DWI and ASL into the MRI radiomics model improved the diagnostic performance in differentiating RIBI from TR; ASL, especially, played a significant role. • The multiparameter MRI radiomics model was superior to the conventional MRI radiomics model in differentiating glioma recurrence from radiation-induced brain injury. • Diffusion and perfusion MRI could improve the ability of the radiomics model in predicting the progression in patients with glioma. • Arterial spin labeling played an important role in predicting glioma progression using radiomics models.

Hepatocellular carcinoma (HCC) remains a major cause of cancer-related mortality, and transarterial chemoembolization (TACE) is the standard therapy for intermediate-stage disease. However, response to TACE is variable, and reliable quantitative imaging biomarkers are needed to support early treatment decision-making. This study aimed to evaluate the predictive value of the delayed percentage attenuation ratio (DPAR) measured from pre-TACE multiphasic computed tomography (CT) in forecasting early therapeutic response. A retrospective cross-sectional study was conducted involving patients with a definitive diagnosis of HCC who underwent their first TACE session and had complete multiphasic CT imaging before and after treatment. Quantitative washout parameters, delayed percentage attenuation ratio (DPAR), absolute washout (WOAbs), and relative washout (WORel) were measured using standardized region of interest (ROI) placement by three radiologists. Treatment response was assessed four to six weeks post- TACE based on modified Response Evaluation Criteria in Solid Tumors (mRECIST) criteria and classified into responders and non-responders. Diagnostic performance was evaluated using receiver operating characteristic (ROC) analysis, and interobserver reliability was assessed using intraclass correlation coefficient (ICC) and Cohen's κ. A total of 49 HCC patients were included and analyzed. Responders demonstrated significantly higher DPAR values compared with non-responders (median 134.5 vs 113.0; p<0.001). DPAR showed the strongest discriminative performance with an area under the curve (AUC) of 0.898, outperforming WOAbs (AUC 0.689) and WORel (AUC 0.704). The optimal DPAR threshold of ≥120.5 provided 84.4% sensitivity and 88.2% specificity to predict early post-TACE treatment response. Interobserver reliability was excellent for all washout parameters (ICC 0.98–0.99), and agreement for mRECIST classification was also excellent (κ=0.867). In conclusion, pre-TACE DPAR is a robust and reproducible quantitative imaging biomarker that accurately predicts early response to TACE in HCC. A threshold value of ≥120.5 may assist in treatment planning and patient selection in routine clinical practice.

This study aimed to evaluate the diagnostic performances of dual-layer CT (DLCT) for the identification of positive lymph nodes (LNs) in patients with lymphoma and retrospectively included 1165 LNs obtained by biopsy from 78 patients with histologically proven lymphoma, who underwent both pretreatment DLCT and 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT). According to 18F-FDG PET/CT findings as a reference standard, cases were categorized into the LN-negative and LN-positive groups. LNs were then randomly divided at a ratio of 7:3 into the training (n = 809) and validation (n = 356) cohorts. The patients’ clinical characteristics and quantitative parameters including spectral curve slope (λHU), iodine concentration (IC) on arterial phase (AP) and venous phase (VP) images were compared between the LN-negative and LN-positive groups using Chi-square test, t-test or Mann–Whitney U test for categorical variables or quantitative parameters. Multivariate logistic regression analysis with tenfold cross-validation was performed to establish the most efficient predictive model in the training cohort. The area under the curve (AUC) was used to evaluate the diagnostic value of the predictive model, and differences in AUC were determined by the DeLong test. Moreover, the predictive model was validated in the validation cohort. Repeatability analysis was performed for LNs using intraclass correlation coefficients (ICCs). In the training cohort, long diameter (LD) had the highest AUC as an independent factors compared to other parameter in differentiating LN positivity from LN negativity (p = 0.006 to p < 0.001), and the AUC of predictive model jointly involving LD and λHU-AP was significantly elevated (AUC of 0.816, p < 0.001). While the AUC of predictive model in the validation cohort was 0.786. Good to excellent repeatability was observed for all parameters (ICC > 0.75). The combination of DLCT with morphological and functional parameters may represent a potential imaging biomarker for detecting LN positivity in lymphoma.

Background: Accurate subtype identification of lymphoma cancer is crucial for effective diagnosis and treatment planning. Although standard deep learning algorithms have demonstrated robustness, they are still prone to overfitting and limited generalization, necessitating more reliable and robust methods. Objectives: This study presents an autoencoder-augmented stacked ensemble learning (SEL) framework integrating deep feature extraction (DFE) and ensembles of machine learning classifiers to improve lymphoma subtype identification. Methods: Convolutional autoencoder (CAE) was utilized to obtain high-level feature representations of histopathological images, followed by dimensionality reduction via Principal Component Analysis (PCA). Various models were utilized for classifying extracted features, i.e., Random Forest (RF), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), AdaBoost, and Extra Trees classifiers. A Gradient Boosting Machine (GBM) meta-classifier was utilized in an SEL approach to further fine-tune final predictions. Results: All the models were tested using accuracy, area under the curve (AUC), and Average Precision (AP) metrics. The stacked ensemble classifier performed better than all the individual models with a 99.04% accuracy, 0.9998 AUC, and 0.9996 AP, far exceeding what regular deep learning (DL) methods would achieve. Of standalone classifiers, MLP (97.71% accuracy, 0.9986 AUC, 0.9973 AP) and Random Forest (96.71% accuracy, 0.9977 AUC, 0.9953 AP) provided the best prediction performance, while AdaBoost was the poorest performer (68.25% accuracy, 0.8194 AUC, 0.6424 AP). PCA and t-SNE plots confirmed that DFE effectively enhances class discrimination. Conclusion: This study demonstrates a highly accurate and reliable approach to lymphoma classification by using autoencoder-assisted ensemble learning, reducing the misclassification rate and significantly enhancing the accuracy of diagnosis. AI-based models are designed to assist pathologists by providing interpretable outputs such as class probabilities and visualizations (e.g., Grad-CAM), enabling them to understand and validate predictions in the diagnostic workflow. Future studies should enhance computational efficacy and conduct multi-centre validation studies to confirm the model’s generalizability on extensive collections of histopathological datasets.

Driverless artificial intelligence framework for the identification of malignant pleural effusion

ObjectiveTo develop and validate a generalized prediction model that can classify epidermal growth factor receptor (EGFR) mutation status in non–small cell lung cancer patients.MethodsA total of 346 patients (296 in the training cohort and 50 in the validation cohort) from four centers were included in this retrospective study. First, 1085 features were extracted using IBEX from the computed tomography images. The features were screened using the intraclass correlation coefficient, hypothesis tests and least absolute shrinkage and selection operator. Logistic regression (LR), decision tree (DT), random forest (RF), and support vector machine (SVM) were used to build a radiomics model for classification. The models were evaluated using the following metrics: area under the curve (AUC), calibration curve (CAL), decision curve analysis (DCA), concordance index (C-index), and Brier score.ResultsSixteen features were selected, and models were built using LR, DT, RF, and SVM. In the training cohort, the AUCs was .723, .842, .995, and .883; In the validation cohort, the AUCs were .658, 0567, .88, and .765. RF model with the best AUC, its CAL, C-index (training cohort=.998; validation cohort=.883), and Brier score (training cohort=.007; validation cohort=0.137) showed a satisfactory predictive accuracy; DCA indicated that the RF model has better clinical application value.ConclusionMachine learning models based on computed tomography images can be used to evaluate EGFR status in patients with non–small cell lung cancer, and the RF model outperformed LR, DT, and SVM.

Explainable Boosting Machine Model with a Parallel Ensemble Design Predicts Local Failure for Head and Neck Cancer With Clinical, CT, and Delta CBCT Radiomic Features

Neglected Tropical Diseases (NTDs) affect 1.5 billion people worldwide with 39% of the global burden occurring in Africa. In Kenya, NTDs remain endemic despite control efforts, with co-endemicity of soil-transmitted helminths (STH), schistosomiasis (SCH), and lymphatic filariasis (LF) complicating intervention strategies. This study developed machine learning models to predict high-risk co-endemic areas using demographic and Water, Sanitation, and Hygiene (WASH) indicators. The study analyzed Kenya's 2022 NTD co-endemicity data from the Expanded Special Project for Elimination of Neglected Tropical Diseases, incorporating WASH and population variables. Three machine learning algorithms, Random Forest, Gradient Boosting Machine, and Extreme Gradient Boosting (XGBoost) were trained to classify regions by STH prevalence levels and co-endemicity status. Model performance was evaluated using cross-validation, Receiver Operating Characteristic - Area under the Curve (AUC) and variable importance analysis. The RF model achieved the highest predictive performance (AUC = 0.70), followed by XGBoost (AUC = 0.66) and GBM (AUC = 0.62). Key predictors included improved sanitation access (mean importance score: 0.24), population density (0.21), and co-endemicity with LF/SCH (0.18). Spatial analysis identified Eastern and North-Eastern Kenya as persistent hotspots, correlating with low WASH coverage (<40%). Machine learning models effectively identified the high-risk NTD co-endemic areas in Kenya, with RF outperforming other models. These findings support targeted interventions integrating WASH improvements with mass drug administration in identified hotspots. We propose a real-time dashboard for dynamic risk mapping to optimize resource allocation; a strategy aligned with Kenya's NTD Elimination Strategic Plan 2030.