Model Performance Metrics Research Articles

A systematic review and meta-analysis of lung cancer risk prediction models.

Lung cancer (LC) remains the leading cause of cancer-related mortality worldwide. Early detection through targeted screening significantly improves patient outcomes. However, identifying high-risk individuals remains a critical challenge. This systematic review evaluates externally validated LC risk prediction models to assess their performance and potential applicability in screening strategies. Of the 11,805 initial studies, 66 met inclusion criteria and 38 published mainly between 2020 and 2024 were included in the final analysis. Model methodologies, validation approaches, and performance metrics were extracted and compared. The review identified 18 models utilising conventional machine learning, six employing neural networks, and 14 comparing different predictive frameworks. The Prostate Lung Colorectal and Ovarian Cancer Screening Trial (PLCOm2012) demonstrated superior sensitivity across diverse populations, while newer models, such as Optimized Early Warning model for Lung cancer risk (OWL) and CanPredict, showed promising results. However, differences in population demographics and healthcare systems may limit the generalisability of these models. While LC risk prediction models have advanced, their applicability to specific healthcare systems, such as Finland's, requires further adaptation and validation. Future research should focus on optimising these models for local contexts to improve clinical impact and cost-effectiveness in targeted screening programmes. PROSPERO CRD42022321391.

Hierarchical diagnosis of breast phyllodes tumors enabled by deep learning of ultrasound images: a retrospective multi-center study

ObjectivePhyllodes tumors (PTs) are rare breast tumors with high recurrence rates, current methods relying on post-resection pathology often delay detection and require further surgery. We propose a deep-learning-based Phyllodes Tumors Hierarchical Diagnosis Model (PTs-HDM) for preoperative identification and grading.MethodsUltrasound images from five hospitals were retrospectively collected, with all patients having undergone surgical pathological confirmation of either PTs or fibroadenomas (FAs). PTs-HDM follows a two-stage classification: first distinguishing PTs from FAs, then grading PTs into benign or borderline/malignant. Model performance metrics including AUC and accuracy were quantitatively evaluated. A comparative analysis was conducted between the algorithm’s diagnostic capabilities and those of radiologists with varying clinical experience within an external validation cohort. Through the provision of PTs-HDM’s automated classification outputs and associated thermal activation mapping guidance, we systematically assessed the enhancement in radiologists’ diagnostic concordance and classification accuracy.ResultsA total of 712 patients were included. On the external test set, PTs-HDM achieved an AUC of 0.883, accuracy of 87.3% for PT vs. FA classification. Subgroup analysis showed high accuracy for tumors < 2 cm (90.9%). In hierarchical classification, the model obtained an AUC of 0.856 and accuracy of 80.9%. Radiologists’ performance improved with PTs-HDM assistance, with binary classification accuracy increasing from 82.7%, 67.7%, and 64.2–87.6%, 76.6%, and 82.1% for senior, attending, and resident radiologists, respectively. Their hierarchical classification AUCs improved from 0.566 to 0.827 to 0.725–0.837. PTs-HDM also enhanced inter-radiologist consistency, increasing Kappa values from − 0.05 to 0.41 to 0.12 to 0.65, and the intraclass correlation coefficient from 0.19 to 0.45.ConclusionPTs-HDM shows strong diagnostic performance, especially for small lesions, and improves radiologists’ accuracy across all experience levels, bridging diagnostic gaps and providing reliable support for PTs’ hierarchical diagnosis.

Predictive Analysis of Remaining Useful Life of Batteries: Result Implementation and Discussion

This paper presents the implementation results of a predictive analysis model for estimating the Remaining Useful Life (RUL) of batteries. The study employs machine learning techniques to forecast battery performance degradation, ensuring reliability and efficiency in energy storage systems. The findings highlight the effectiveness of data-driven models in predicting battery lifespan with high accuracy. The paper details the methodology, results, and implications of the study, providing insights into future improvements. The increasing reliance on battery-powered systems, particularly in electric vehicles and renewable energy storage, has heightened the necessity for accurate battery health assessment. Traditional approaches to battery maintenance often lead to inefficiencies and unplanned downtimes. This research addresses these issues by leveraging predictive analytics to develop robust models capable of estimating battery lifespan with minimal error. The study compares various machine learning models, emphasizing the importance of feature selection and model performance metrics. The findings reveal that neural networks outperform other models in predicting RUL, offering a reliable tool for energy management. This paper further discusses computational efficiency, practical deployment challenges, and future research directions to enhance predictive accuracy. The proposed methodology demonstrates the potential for real-world applications, improving battery usage efficiency and reducing operational costs.

Prediction of Rice Productivity Using the Random Forest Regression Algorithm in Cikaret Subdistrict for the Years 2020-2024

The challenges surrounding rice productivity in Indonesia are growing more complex due to factors like climate change, population growth, and limited agricultural land. As the primary food source and main carbohydrate provider, rice is crucial for the majority of Indonesians. This study focuses on predicting rice productivity using the random forest regression algorithm, incorporating predictor variables such as NDVI, NDMI, land area, land surface temperature, rainfall, fertilizer type, and pests. To ensure the accuracy of the model, multicollinearity tests were conducted to check for strong correlations among the independent variables. The tests confirmed the absence of significant linear relationships, allowing all variables to be included in the model. The prediction model was built using time-series data from 2020 to 2023, resulting in 840 samples after eliminating outliers. The optimization process targeted the mtry parameter and the number of decision trees to reduce prediction error. The optimal model, utilizing 7 predictor features and 150 decision trees, achieved a low out of bag (OOB) error and stable mean square error (MSE). Model performance metrics showed a Mean Absolute Error (MAE) of 0.324 tons/hectare, MSE of 0.158 tons/hectare, Root Mean Square Error (RMSE) of 0.398 tons/hectare, and a coefficient of determination (R²) of 0.87. These results demonstrated that the random forest regression algorithm is highly effective in predicting rice productivity, particularly when dealing with complex data involving multiple predictor variables and potential multicollinearity.

Identifying patients at risk of suicide using data from health information exchanges

BackgroundSuicide is within the top 10 causes of death in the US and rates have increased over 30% in the last two decades. Identifying suicide risk in healthcare settings has become a national priority and predicting risk using statistical models has emerged as a promising approach to achieve this. Numerous previously published models drawing on information in hospital records, insurance claims, or single healthcare networks perform reasonably well. With recent federal mandates established to develop health information exchanges (HIEs) that include patient information from both diverse healthcare settings and multiple healthcare networks, it is imperative to determine whether this newer data source offers an opportunity to model suicide risk with robust yet largely non-curated, real-world patient information. We examined the characteristics and performance of suicide risk models developed from a HIE, the Kansas Health Information Network.MethodsIn this retrospective cohort study, utilizing HIE data between 2012 and 2017, we developed predictive models of suicide attempts in 18–64 year old Kansas residents. Predictors included patient age, gender, and medical diagnosis codes. We evaluated model performance metrics, predictor importance, and characteristics of patients accurately labeled as high risk.ResultsA total of 501,595 18–64 year old Kansas residents were included in the study, with a suicide attempt rate of 0.4% (n = 1,914). Our best model had an average area under the ROC curve (AUROC) of 0.82 that was typical of published models (AUROCs = 0.73–0.85), and predictors of suicide attempts included diagnosis codes found in other published models but also reflected the diversity of clinical settings contributing to the HIE.ConclusionsReal-world data from the diverse clinical settings captured in an HIE may provide unprecedented opportunities to predict suicidal behavior.

Detection and pre-localization of coastal wetlands in Brittany, France using topographical indices from altimetric and remote sensing data

The role of wetlands in coastal area processes must be recognized and strengthened. Unless the hydrological and related functions of coastal wetlands are maintained, the success of sustainable coastal development is uncertain. In this study, an initial methodology section is dedicated to the calculation and normalization of several topographic indices to assess soil moisture susceptibility in coastal wetland areas. By providing detailed information on the elevation and topographic structure of coastal wetlands, this study uses various topographic indices, including the Topographic Wetness Index (TWI), the Topographic Position Index (TPI), the Multi-Resolution Valley Bottom Flatness (MRVBF). Additionally, the study considers valley depth, which can contribute to a better understanding of hydrological dynamics, water level variations, and water flow zones during the process of mapping and monitoring changes in these environments. The second research aim of this work is evaluation of the pre-localization of potential coastal wetland areas and their evolution over time in relation to impervious surface changes in Brittany. The analysis reveals that between 1990 and 2020, the area of potentially impervious wetlands increased by 18.3% from 145.3 km2 to 171.95 km2. By combining these pre-localization results with Corine Land Cover (CLC) data and OCS GE, the study highlights the influence of urbanized and impermeable area on coastal wetlands dynamics between 1990 and 2020. The third aim in of this article focuses on assessing the quality of the “binary classification” of wetlands and non-wetlands. A inventory focus on coastal wetlands (carried out by stakeholders between 2011 and 2019) is used as reference data to check whether the proposed methodology is effective and, if so, to determine the score at which it gives satisfactory results. Model performance metrics show a high recall of 0.948 for non-wetland areas, though with moderate precision (0.798), suggesting occasional misclassification of wetland areas as non-wetlands. For wetland areas, the approach achieved high precision (0.936) but a lower recall (0.759), indicating challenges in detecting all existing wetland areas. The overall accuracy of 0.854 and a Kappa coefficient of 0.708 point to a solid performance of the binary classification methodology.

Risk prediction and effect evaluation of complicated appendicitis based on XGBoost modeling

PurposeThe distinction between complicated appendicitis (CAP) and uncomplicated appendicitis (UAP) remains challenging. The purpose of this study was to construct a safe and economical diagnostic model that can accurately and rapidly differentiate between CAP and UAP.MethodsPatient data from 773 appendectomies were retrospectively collected, important features were selected using random forests, and the data were divided into training and test sets in a 3:1 ratio. An integrated learning algorithm, Extreme Gradient Boosting (XGBoost), was introduced to predict the risk of CAP and compared with Support Vector Machine (SVM), Random Forest (RF), and Decision Tree (CART) algorithms. A comprehensive comparison of the four algorithms was performed using model performance metrics such as the area under the receiver’s operating characteristic curve (AUC), sensitivity, specificity, accuracy, precision, negative predictive value(NPV), positive predictive value(PPV),calibration curves, and clinical decision curve analysis (DCA).ResultThe results show that all four prediction models exhibit some predictive ability. The XGBoost model showed the best prediction with AUC, accuracy, sensitivity, specificity,NPV and PPV of 0.914, 0.855, 0.865, 0.846,0.848 and 0.897, respectively, followed by the SVM model with results of AUC, accuracy, sensitivity, specificity,NPV and PPV of 0.882, 0.819, 0.865, 0.779, 0.770 and 0.871, respectively. XGBoost and SVM models show very good calibration. The XGBoost model showed better net clinical benefit compared to the DCA curves of the other models.ConclusionPredictive models based on the XGBoost algorithm have good performance in predicting the risk of acute appendicitis progressing to complicated appendicitis, which helps to optimize clinical decision making.

Machine Learning-Based Prediction Models for Healthcare Outcomes in Patients Participating in Cardiac Rehabilitation: A Systematic Review.

Cardiac rehabilitation (CR) has been proven to reduce mortality and morbidity in patients with cardiovascular disease. Machine learning (ML) techniques are increasingly used to predict healthcare outcomes in various fields of medicine including CR. This systemic review aims to perform critical appraisal of existing ML-based prognosis predictive model within CR and identify key research gaps in this area. A systematic literature search was conducted in Scopus, PubMed, Web of Science, and Google Scholar from the inception of each database to January 28, 2024. The data extracted included clinical features, predicted outcomes, model development, and validation as well as model performance metrics. Included studies underwent quality assessments using the IJMEDI and Prediction Model Risk of Bias Assessment Tool checklist. A total of 22 ML-based clinical models from 7 studies across multiple phases of CR were included. Most models were developed using smaller patient cohorts from 41 to 227, with one exception involving 2280 patients. The prediction objectives ranged from patient intention to initiate CR to graduate from outpatient CR along with interval physiological and psychological progression in CR. The best-performing ML models reported area under the receiver operating characteristics curve between 0.82 and 0.91, with sensitivity from 0.77 to 0.95, indicating good prediction capabilities. However, none of them underwent calibration or external validation. Most studies raised concerns about bias. Readiness of these models for implementation into practice is questionable. External validation of existing models and development of new models with robust methodology based on larger populations and targeting diverse clinical outcomes in CR are needed.

DFL topology optimization based on peer weighting mechanism and graph neural network in digital twin platform

Decentralized federated learning (DFL) represents a distributed learning framework where participating nodes independently train local models and exchange model updates with proximate peers, circumventing the reliance on a centralized orchestrator. This paradigm effectively mitigates server-induced bottlenecks and eliminates single points of failure, which are inherent limitations of centralized federated learning architectures. However, DFL encounters significant challenges in attaining global model convergence due to inherent statistical heterogeneity across nodes and the dynamic nature of network topologies. For the first time, in this paper, we present a topology optimization framework for DFL that integrates a peer weighting mechanism with graph neural networks (GNNs) within a digital twin platform. The proposed approach leverages local model performance metrics and training latency as input factors to dynamically construct an optimized topology that balances computational efficiency and model performance. Specifically, we employ Particle Swarm Optimization to derive node-specific peer weight matrices and utilize a GNN to refine the underlying mesh topology based on these weights. Comprehensive experimental analyses conducted on benchmark datasets demonstrate the superiority of the proposed framework in achieving accelerated convergence and enhanced accuracy across diverse nodes. Additionally, comparative evaluations under IID and Non-IID data distributions substantiate the robustness and adaptability of the approach in heterogeneous learning environments, underscoring its potential to advance decentralized learning paradigms.

Prevalence and molecular epidemiology of the novel equine parasite Theileria haneyi in China.

Equine piroplasmosis (EP), caused by Theileria equi (T. equi) and Babesia caballi (B. caballi), is a tick-borne disease with significant economic impacts on the equine industry. Theileria haneyi (T. haneyi), a newly identified pathogen of EP, is globally distributed but has not been reported in China, where its epidemiological patterns remain undefined. To develop diagnostic techniques for T. haneyi and to assess its prevalence and risk factors in China. Assay development and cross sectional survey. A nested PCR (nPCR) nucleic acid diagnostic technique targeting the chromosome 1 single-copy (chr1sco) open reading frame (ORF) was developed. This method was applied to 1318 equine nucleic acid samples. Additionally, an iELISA serological diagnostic method was established based on the EMA11 gene. This method was utilised in a cross-sectional analysis of 2627 equine samples. Logistic regression analysis was performed to identify significant risk factors. The nPCR assay showed an 11.76% positive detection rate, while iELISA indicated a 16.41% seroprevalence. Both assays were successfully applied for the epidemiological investigation of T. haneyi. Logistic regression analysis identified host species, age, altitude, soil type, and forest and grassland coverage as the main risk factors influencing seropositivity. Phylogenetic analysis of 18S rRNA from 29 positive samples confirmed the presence of T. haneyi in China. Data on clinical signs were not collected. The logistic regression model's performance metrics were not calculated. This study provides the first evidence of T. haneyi infection in China and establishes a scientific basis for understanding its prevalence and geographical distribution.

Association of the child opportunity index with in-hospital mortality and persistence of organ dysfunction at one week after onset of Phoenix Sepsis among children admitted to the pediatric intensive care unit with suspected infection.

The social determinants of health (SDoH) are fundamental factors that contribute to overall health and health-related outcomes. Children living in lower socioeconomic areas have a higher risk of critical illness and worse outcomes compared to children living in more socioeconomically advantaged areas. In this work, we determine whether the Child Opportunity Index (COI 3.0), a multi-dimensional child-specific indicator of neighborhood environment, is associated with in-hospital mortality or persistence of a Phoenix Sepsis Score ≥ 2 at one week following Phoenix Sepsis onset in children admitted to pediatric intensive care units (PICUs) with suspected infection. We performed a retrospective cohort analysis of 63,824 patients with suspected or confirmed infection admission diagnosis in two PICUs in Atlanta, Georgia with a Georgia residential address that could be geocoded and linked to a census tract. The primary outcome was the composite of in-hospital mortality or persistence of a Phoenix Sepsis Score ≥ 2 at one week following Phoenix Sepsis onset. Model performance measures of interest were the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). Models developed with electronic medical record (EMR) data using Egleston (EG) or Scottish Rite (SR) as the training site achieved AUROCs of 0.81-0.84 (95% CI range: 0.8-0.85) and 0.82-0.82 (95% CI range: 0.81-0.83) and AUPRCs of 0.59-0.68 (95% CI range: 0.58-0.69) and 0.62-0.64 (95% CI range: 0.61-0.65) respectively. Despite significant differences in COI 3.0 characteristics and overall in-hospital mortality of children with Phoenix suspected infection between the EG and SR PICUs, the addition of COI 3.0 did not improve the overall model performance metrics. While children admitted to both PICUs were more often from COI 3.0 neighborhoods in the lowest two quintiles, these neighborhood features had less of an impact on the model's predictive performance compared to patient physiologic and biologic features available in the EMR.

Application of Machine Learning Model for Forecasting of Floating PV Cell

floating solar photovoltaic (FPV) systems are gaining global attention due to their efficient use of water surfaces and natural cooling effects. However, the variability in environmental factors affecting FPV systems presents a significant challenge for accurate power forecasting. This paper investigates short-term forecasting of FPV power output using machine learning (ML) techniques. Models including Support Vector Regression (SVR), Random Forest (RF), and Long Short-Term Memory (LSTM) networks are developed and compared. A detailed analysis of feature importance, model performance, and error metrics is presented. The results demonstrate that LSTM achieves superior accuracy with a Mean Absolute Percentage Error (MAPE) of 2.5%, making it a promising tool for FPV power management. Key word - Floating solar photovoltaic (FPV), short-term forecasting, machine learning, Long Short-Term Memory (LSTM), Support Vector Regression (SVR), Random Forest (RF), renewable energy, solar energy forecasting, time-series analysis, power output prediction.

Domain-Adaptive Pretraining of Transformer-Based Language Models on Medical Texts: A High-Performance Computing Experiment

This research was to investigate the effect of utilizing high-performance computing (HPC) resources to enhance the adaptability and performance of transformer-based language models. The research was done through intensive domain-specific pretraining in the medical domain. The study aimed to answer the question: Can domain-adaptive pretraining on medical texts significantly improve language model performance metrics such as perplexity while maintaining computational efficiency and addressing ethical considerations? The research utilized a corpus of medical texts. These were carefully split into training and evaluation datasets. Initial model training on NVIDIA A30 GPUs, with 96% GPU utilization, calculated an average perplexity of 73.54. Following iterative refinements—including domain-specific tokenizer optimization, data preprocessing, mixed-precision training, and adjusted learning parameters—the final model achieved an average perplexity of 3.39. The evaluation run processed 7103 samples in 98.02 seconds, with a training loss of 2.405 and an evaluation loss of 2.045, indicating strong generalization and the absence of overfitting. The final model and results were saved for reproducibility and future use. This study was justified by the pressing need for accurate and efficient medical natural language processing (NLP) applications. The application areas are in clinical decision support, patient record summarization, and medical research analysis. The research findings highlight that investing in HPC-driven domain-adaptive pretraining delivers substantial improvements in performance. It also equips medical NLP models with abilities to handle the complexities of domain-specific language effectively. The Ethical considerations of this research were based on optimizing GPU utilization to reduce energy consumption and ensure transparency through reproducible methodologies. We recommend future research to explore larger medical datasets, broader clinical specializations, and diverse transformer architectures while also investigating the transferability of learned representations across related medical subdomains. The advancements could further enhance the applicability of specialized language models in medical research and practice.

Prediction of lumbar disc degeneration based on interpretable machine learning models: retrospective cohort study.

Prediction of lumbar disc degeneration based on interpretable machine learning models: retrospective cohort study.

Landslide susceptibility mapping for hazard management along Pakistan’s Balakot-Naran Route

Landslides represent a significant hazard worldwide, leading to substantial loss of life and property. In Pakistan’s Balakot Valley, severe landslides frequently disrupt key roadways, particularly along the Balakot-Naran route, which serves as a crucial artery for tourism—the region’s primary economic activity. This study aimed to create a high-resolution landslide susceptibility map for a 457 km² area along this road, where landslides are driven by factors such as intense rainfall, weak geological formations, seismic activity, and slope destabilization due to road construction. Using Geographic Information Systems (GIS) and remote sensing, we analyzed nine critical landslide-predictive factors: slope, aspect, lithology, land cover/land use, plan curvature, profile curvature, and proximity to faults, roads, and streams. Our methodology applied the frequency ratio (FR) model to 80% of the landslide inventory data for model construction, while 20% of the inventory was reserved for validation. The resulting susceptibility map, classified into low, moderate, high, and very high-risk zones using the natural breaks method in ArcGIS, indicates that slope gradient, lithology, land cover, and stream proximity are the primary contributors to landslide occurrence in this area. Model performance metrics demonstrate high predictive accuracy, with a success rate of 0.93 and a prediction rate of 0.96. The generated susceptibility map provides a robust tool for hazard management, offering valuable insights for targeted mitigation strategies and sustainable infrastructure development in this landslide-prone region. This study advanced landslide susceptibility mapping by integrating high-impact geospatial factors with an innovative, validated FR approach, supporting safer, data-driven land use planning and disaster preparedness.

Identifying Risk and Protective Factors for Attrition Among Recently Enlisted Navy Personnel Using Variable Importance Measures.

Approximately 1 in 4 Navy recruits will not complete their obligated service. Identifying factors associated with Navy attrition is essential to retaining a skilled and motivated fighting force. This study utilized machine learning and variable importance measures (VIM) extracted from random forest (RF) and extreme gradient boosting (XGBoost) decision trees to identify factors that contribute to Navy attrition. These methods are more robust and sensitive when assessing hundreds of predictor variables, offering superior performance relative to regression analyses. Leveraging advanced analytics is crucial to better understand these factors to maximize military retention. The study population included Navy personnel with an accession date in 2016. The outcome of interest was a discharge date before the end of obligated service (i.e., attrition). The analysis incorporated 542 independent variables, including demographic factors, medical outpatient visits, and outpatient pharmacy medications. RF, XGBoost, and logistic regression models were trained with different subsets of features to determine which set of variables best predicted attrition. Model performance metrics were generated describing the sensitivity, specificity, positive and negative predictive values, area under the curve, and classification accuracy for the validation data. Overall, there were 39,866 Navy personnel with accessions in 2016 and, of these, 28.15% (n = 11,177) did not complete their obligated service. The RF model provided the best accuracy for predicting attrition (81.7%) and area under the curve (90.0%). Both the RF and XGBoost models outperformed the logistic regression model. VIMs were then assessed, including mean decrease accuracy, mean decrease Gini impurity, gain, and cover. The resulting VIM identified five groups described as: mental health, occupations, demographics/sex-related issues, pain management, and medical compliance. The top 3 VIM and their overall relative risk on attrition were adjustment disorders (RR = 1.39 higher), Seaman Specialists (RR = 3.01 higher), and Electronic Equipment Repairers (RR = 0.44 lower). Five groups of variables were found to be predictive of Navy attrition: Mental health (MH) disorder, alcohol-related problems, occupations, sex, medical appointment compliance, and pain management. These results demonstrate the utility of machine learning models in predicting attrition compared with regression analyses. VIM is a valuable tool that could be used in decision-making processes in the context of military personnel management and retention. Furthermore, ensemble approaches, compared with a single decision tree, improve overall predictive performance and result in a more robust model that is resistant to overfitting.

Study on Skin Cancer (SC) Detection Using Deep Learning (DL)Techniques: A Review

Skin cancer is one of the most prevalent and life-threatening diseases, necessitating early and accurate detection for effective treatment. Deep Learning (DL) techniques, particularly Convolutional Neural Networks (CNNs), have emerged as powerful tools in the automated detection and classification of skin cancer. This review examines the latest advancements in DL-based skin cancer detection, focusing on model architectures, performance metrics, and dataset utilization. Additionally, the study highlights key challenges such as data scarcity, model interpretability, and generalization issues. Future research directions, including multimodal data integration, transfer learning, and enhanced data augmentation techniques, are explored to improve diagnostic accuracy. The findings suggest that DL has significant potential in revolutionizing dermatological diagnostics by offering high precision, reduced subjectivity, and faster analysis.

Scale dependency in modeling nivo-glacial hydrological systems: the case of the Arolla basin, Switzerland

Abstract. Hydrological modeling in alpine catchments poses unique challenges due to the complex interplay of meteorological, topographical, geological, and glaciological drivers with streamflow generation. A significant issue arises from the limited availability of streamflow data due to the scarcity of high-elevation gauging stations. Consequently, there is a pressing need to assess whether streamflow models that are calibrated with moderate-elevation streamflow data can be effectively transferred to higher-elevation catchments, notwithstanding differences in the relative importance of different streamflow-generation processes. Here, we investigate the spatial transferability of calibrated temperature-index melt model parameters within a semi-lumped modeling framework. We focus on evaluating the melt model transferability from the main catchment to nested and neighboring subcatchments in the Arolla valley, southwestern Swiss Alps. We use the Hydrobricks modeling framework to simulate streamflow, implementing three variants of a temperature-index snow and ice melt model (the classical degree-day model, the aspect-related model, and the Hock temperature-index model). Through an analysis of streamflow simulations, benchmark metrics consisting of resampled and bootstrapped discharge time series, and model performance metrics, we demonstrate that robust parameter transferability and accurate streamflow simulation are possible across diverse spatial scales. This finding is conditional upon the melt model applied, with melt models using more spatial information leading to convergence of the model parameters until we observe overparameterization. We conclude that simple semi-lumped models can be used to extend hydrological simulations to ungauged catchments in alpine regions and improve high-elevation water resource management and planning efforts, especially in the context of climate change.

Radiomics in thymic epithelial tumors: a scoping review of current status and advances

BackgroundThis scoping review aims to evaluate the current status and recent advances in the use of radiomics for the diagnosis, risk stratification, and staging of thymic epithelial tumors (TETs). The review also explores radiomics' potential in predicting the risk of myasthenia gravis (MG), an associated autoimmune condition in TETs patients.MethodsA comprehensive literature search was conducted using PubMed and Web of Science to identify studies published since 2012 on the application of radiomics in managing TETs. The studies were assessed for their methodologies, including imaging protocols, feature extraction techniques, and model performance metrics. The Radiomics Quality Score (RQS) was used to evaluate study quality.ResultsA total of 23 studies, including 4701 patients, were analyzed. Radiomics-based models showed high accuracy in distinguishing TETs from other mediastinal masses, predicting risk subtypes, and improving the accuracy of TNM and Masaoka-Koga (MK) staging. Additionally, radiomics demonstrated potential in predicting the risk of MG in thymoma patients. However, all studies were retrospective, and only 6 studies included external validation, with an average RQS of 13.87, accounting for 38.52% of the maximum score.ConclusionRadiomics shows great potential in advancing the diagnosis, risk stratification, and staging of TETs. However, its clinical implementation requires overcoming challenges in standardization, validation, and interpretability. Future research should focus on multi-center prospective studies, external validation, and integrating multi-modal imaging and molecular biomarkers to improve risk assessment and treatment strategies.

Predicting Inpatient Admissions From Emergency Department Triage Using Machine Learning: A Systematic Review.

Predicting Inpatient Admissions From Emergency Department Triage Using Machine Learning: A Systematic Review.