Explanation of machine learning models using shapley additive explanation and application for real data in hospital

Amoeba community dynamics and assembly mechanisms in full-scale drinking water distribution networks under various disinfectant regimens

Free-living amoebae (FLA) are prevalent in drinking water distribution networks (DWDNs), yet our understanding of FLA community dynamics and assembly mechanisms in DWDNs remains limited. This study characterized the occurrence patterns of amoeba communities and identified key factors influencing their assembly across four full-scale DWDNs in three Chinese cities, each utilizing different disinfectants (chlorine, chloramine, and chlorine dioxide). High-throughput sequencing of full-length 18S rRNA genes revealed highly diverse FLA communities and an array of rare FLA species in DWDNs. Unique FLA community structures and higher gene copy numbers of three amoeba taxa of concern (Vermamoeba vermiformis, Acanthamoeba, and Naegleria fowleri) were observed in the chloraminated DWDN, highlighting the distinct impact of chloramine on shaping the amoeba community. The FLA communities in DWDNs were primarily driven by deterministic processes, with disinfectant and nitrogen compounds (nitrate, nitrite, and ammonia) identified as the main influencing factors. Machine learning models revealed high SHapley Additive exPlanations (SHAP) values of dominant amoeba genera (e.g., Vannella and Vermamoeba), indicating their critical ecological roles in shaping broader bacterial and eukaryotic communities. Correlation analyses between amoeba genera and bacterial taxa revealed that 82 % of the bacterial taxa exhibiting a negative correlation with amoebae were gram-negative, suggesting the preferred predation of amoebae toward gram-negative bacteria. Network analysis revealed the presence of only one to two amoebae in distinct modules, suggesting that individual amoebae might be selective in grazing. These findings provide insight into the amoeba community dynamics, assembly mechanisms and ecological roles of amoebae in drinking water, which can aid in risk assessments and mitigation strategies within DWDNs.

The development of classification-based machine-learning models for the toxicity assessment of chemicals associated with plastic packaging

Assessing chemical toxicity in materials like plastic packaging is critical to safeguarding public health. This study presents the development of classification-based machine learning models to predict the toxicity of chemicals associated with plastic packaging. Using an extensive dataset of chemical structures, we trained multiple machine learning models-Random Forest, Support Vector Machine, Linear Discriminant Analysis, and Logistic Regression-targeting endpoints such as Neurotoxicity, Hepatotoxicity, Dermatotoxicity, Carcinogenicity, Reproductive Toxicity, Skin Sensitization, and Toxic Pneumonitis. The dataset was pre-processed by selecting 2D molecular descriptors as feature inputs, with resampling methods (ADASYN, Borderline SMOTE, Random Over-sampler, SVMSMOTE Cluster Centroid, Near Miss, Random Under Sampler) applied to balance classes for accurate classification. A five-fold cross-validation technique was used to optimize model performance, with model parameters fine-tuned using grid search. The model performance was evaluated using accuracy (Acc), sensitivity (Se), specificity (Sp), and area under the receiver operating characteristic curve (AUC-ROC) metrics. In most of the cases, the model accuracy was 0.8 or above for both training and test sets. Additionally, SHAP (SHapley Additive exPlanations) values were utilized for feature importance analysis, highlighting significant descriptors contributing to toxicity predictions. The models were ranked using the Sum of Ranking Differences (SRD) method to systematically select the most effective model. The optimal models demonstrated high predictive accuracy and interpretability, providing a scalable and efficient solution for toxicity assessment compared to traditional methods. This approach offers a valuable tool for rapidly screening potentially hazardous chemicals in plastic packaging.

Improving stroke risk prediction by integrating XGBoost, optimized principal component analysis, and explainable artificial intelligence

The relevance of the study is due to the growing number of diseases of the cerebrovascular system, in particular stroke, which is one of the leading causes of disability and mortality in the world. To improve stroke risk prediction models in terms of efficiency and interpretability, we propose to integrate modern machine learning algorithms and data dimensionality reduction methods, in particular XGBoost and optimized principal component analysis (PCA), which provide data structuring and increase processing speed, especially for large datasets. For the first time, explainable artificial intelligence (XAI) is integrated into the PCA process, which increases transparency and interpretation, providing a better understanding of risk factors for medical professionals. The proposed approach was tested on two datasets, with accuracy of 95% and 98%. Cross-validation yielded an average value of 0.99, and high values of Matthew's correlation coefficient (MCC) metrics of 0.96 and Cohen's Kappa (CK) of 0.96 confirmed the generalizability and reliability of the model. The processing speed is increased threefold due to OpenMP parallelization, which makes it possible to apply it in practice. Thus, the proposed method is innovative and can potentially improve forecasting systems in the healthcare industry.

Machine learning algorithms for predicting future curve using first and second visit data in female adolescent idiopathic scoliosis patients

PURPOSE

This study was designed to develop a machine learning (ML) model that predicts future Cobb angle in patients with adolescent idiopathic scoliosis (AIS) using minimal radiographs and simple questionnaires during the first and second visits.

METHODS

Our study focused on 887 female patients with AIS who were initially consulted at a specialized scoliosis center from July 2011 to February 2023. Patient data, including demographic and radiographic data based on anterior-posterior and lateral whole-spine radiographs, were collected at the first, second, and final visits. ML algorithms were employed to develop individual regression models for future Cobb angles of each curve type (proximal thoracic: PT, main thoracic: MT, and thoracolumbar/lumbar: TLL) using PyCaret in Python. Multiple models were explored and analyzed, with the selection of optimal models based on the coefficient of determination (R2) and median absolute error (MAE).

RESULTS

For the future curve of PT, MT, and TLL, the top-performing models exhibit R2 of 0.73, 0.63, and 0.61 and achieve MAE of 2.3°, 4.0°, and 4.2°.

CONCLUSIONS

The ML-based model using items commonly evaluated at the first and second visits accurately predicted future Cobb angles in female patients with AIS.

Enhancing meteorological data reliability: An explainable deep learning method for anomaly detection

Accurate meteorological observation data is of great importance to human production activities. Meteorological observation systems have been advancing toward automation, intelligence, and informatization. Yet, instrumental malfunctions and unstable sensor node resources could cause significant deviations of data from the actual characteristics it should reflect. To achieve greater data accuracy, early detections of data anomalies, continuous collections and timely transmissions of data are essential. While obvious anomalies can be readily identified, the detection of systematic and gradually emerging anomalies requires further analyses. This study develops an interpretable deep learning method based on an autoencoder (AE), SHapley Additive exPlanations (SHAP) and Bayesian optimization (BO), in order to facilitate prompt and accurate anomaly detections of meteorological observational data. The proposed method can be unfolded into four parts. Firstly, the AE performs anomaly detections based on multidimensional meteorological datasets by marking the data that shows significant reconstruction errors. Secondly, the model evaluates the importance of each meteorological element of the flagged data via SHapley Additive exPlanation (SHAP). Thirdly, a K-sigma based threshold automatic delineation method is employed to obtain reasonable anomaly thresholds that are subject to the data characteristics of different observation sites. Finally, the BO algorithm is adopted to fine-tune difficult hyperparameters, enhancing the model's structure and thus the accuracy of anomaly detection. The practical implication of the proposed model is to inform agricultural production, climate observation, and disaster prevention.

Deep learning and explainable artificial intelligence for investigating dental professionals' satisfaction with CAD software performance

PURPOSE

This study aimed to examine the satisfaction of dental professionals, including dental students, dentists, and dental technicians, with computer-aided design (CAD) software performance using deep learning (DL) and explainable artificial intelligence (XAI)-based behavioral analysis concepts.

MATERIALS AND METHODS

This study involved 436 dental professionals with diverse CAD experiences to assess their satisfaction with various dental CAD software programs. Through exploratory factor analysis, latent factors affecting user satisfaction were extracted from the observed variables. A multilayer perceptron artificial neural network (MLP-ANN) model was developed along with permutation feature importance analysis (PFIA) and the Shapley additive explanation (Shapley) method to gain XAI-based insights into individual factors' significance and contributions.

RESULTS

The MLP-ANN model outperformed a standard logistic linear regression model, demonstrating high accuracy (95%), precision (84%), and recall rates (84%) in capturing complex psychological problems related to human attitudes. PFIA revealed that design adjustability was the most important factor impacting dental CAD software users' satisfaction. XAI analysis highlighted the positive impacts of features supporting the finish line and crown design, while the number of design steps and installation time had negative impacts. Notably, finish-line design-related features and the number of design steps emerged as the most significant factors.

CONCLUSIONS

This study sheds light on the factors influencing dental professionals' decisions in using and selecting CAD software. This approach can serve as a proof-of-concept for applying DL-XAI-based behavioral analysis in dentistry and medicine, facilitating informed software selection and development.

Machine learning-based prediction model integrating ultrasound scores and clinical features for the progression to rheumatoid arthritis in patients with undifferentiated arthritis

OBJECTIVES

Predicting rheumatoid arthritis (RA) progression in undifferentiated arthritis (UA) patients remains a challenge. Traditional approaches combining clinical assessments and ultrasonography (US) often lack accuracy due to the complex interaction of clinical variables, and routine extensive US is impractical. Machine learning (ML) models, particularly those integrating the 18-joint ultrasound scoring system (US18), have shown potential to address these issues but remain underexplored. This study aims to evaluate ML models integrating US18 with clinical data to improve early identification of high-risk patients and support personalized treatment strategies.

METHODS

In this prospective cohort, 432 UA patients were followed for 1 year to track progression to RA. Four ML algorithms and one deep learning model were developed using baseline clinical and US18 data. Comparative experiments on a testing cohort identified the optimal model. SHAP (SHapley Additive exPlanations) analysis highlighted key variables, validated through an ablation experiment.

RESULTS

Of the 432 patients, 152 (35.2%) progressed to the RA group, while 280 (64.8%) remained in the non-RA group. The Random Forest (RnFr) model demonstrated the highest accuracy and sensitivity. SHAP analysis identified joint counts at US18 Grade 2, total US18 score, and swollen joint count as the most influential variables. The ablation experiment confirmed the importance of US18 in enhancing early RA detection.

CONCLUSIONS

Integrating the US18 assessment with clinical data in an RnFr model significantly improves early detection of RA progression in UA patients, offering potential for earlier and more personalized treatments. Key Points • A machine learning model integrating clinical and ultrasound features effectively predicts rheumatoid arthritis progression in undifferentiated arthritis patients. • The 18-joint ultrasound scoring system (US18) enhances predictive accuracy, particularly when incorporated with clinical variables in a Random Forest model. • SHAP analysis underscores that joint severity levels in US18 contribute significantly to early identification of high-risk patients. • This study offers a feasible and efficient approach for clinical implementation, supporting more personalized and timely RA treatment strategies.

An explainable transformer model integrating PET and tabular data for histologic grading and prognosis of follicular lymphoma: a multi-institutional digital biopsy study

BACKGROUND

Pathological grade is a critical determinant of clinical outcomes and decision-making of follicular lymphoma (FL). This study aimed to develop a deep learning model as a digital biopsy for the non-invasive identification of FL grade.

METHODS

This study retrospectively included 513 FL patients from five independent hospital centers, randomly divided into training, internal validation, and external validation cohorts. A multimodal fusion Transformer model was developed integrating 3D PET tumor images with tabular data to predict FL grade. Additionally, the model is equipped with explainable modules, including Gradient-weighted Class Activation Mapping (Grad-CAM) for PET images, SHapley Additive exPlanations analysis for tabular data, and the calculation of predictive contribution ratios for both modalities, to enhance clinical interpretability and reliability. The predictive performance was evaluated using the area under the receiver operating characteristic curve (AUC) and accuracy, and its prognostic value was also assessed.

RESULTS

The Transformer model demonstrated high accuracy in grading FL, with AUCs of 0.964-0.985 and accuracies of 90.2-96.7% in the training cohort, and similar performance in the validation cohorts (AUCs: 0.936-0.971, accuracies: 86.4-97.0%). Ablation studies confirmed that the fusion model outperformed single-modality models (AUCs: 0.974 - 0.956, accuracies: 89.8%-85.8%). Interpretability analysis revealed that PET images contributed 81-89% of the predictive value. Grad-CAM highlighted the tumor and peri-tumor regions. The model also effectively stratified patients by survival risk (P < 0.05), highlighting its prognostic value.

CONCLUSIONS

Our study developed an explainable multimodal fusion Transformer model for accurate grading and prognosis of FL, with the potential to aid clinical decision-making.

Machine Learning Analysis of Nutrient Associations with Peripheral Artery Disease: Insights from NHANES 1999-2004

OBJECTIVE

To investigate the relationship between nutrient intake and peripheral artery disease (PAD) using machine learning (ML) methods.

METHODS

Data from NHANES (1999-2004) were analyzed, including demographic, clinical, and dietary information. Nutrient intake was assessed through 24-hour dietary recalls using CADI and AMPM methods. PAD was defined as an Ankle-Brachial Index (ABI) <0.9. Six ML models-Extreme Gradient Boosting (XGBoost), Random Forest (RF), Naive Bayes Classifier (NB), Support Vector Machine (SVM), Logistic Regression (LR), and Decision Tree (DT)-were trained on a 70/30 train-test split, with missing data imputed and sample imbalance addressed via SMOTE. Model performance was evaluated using AUROC, accuracy, sensitivity, specificity, precision, recall, and F1 score. SHAP analysis was used to identify key features. In addition, to further enhance the interpretability of the model, we applied SHAP analysis to identify the features that have a significant impact on PAD prediction. This approach allowed us to determine the contribution of different variables to the model's output, providing deeper insights into how each feature influences the prediction of PAD outcomes.

RESULTS

Of 31,126 participants, 4,520 met inclusion criteria (mean age 61.2 ± 13.5 years; 48.8% male), and 441 (9.7%) had ABI <0.9. XGBoost outperformed other models, achieving an AUROC of 0.913 (95% CI, 0.891-0.936) and F1 score of 0.932. With SMOTE, its AUROC improved to 0.926 (95% CI, 0.889-0.936) and F1 score to 0.937. SHAP analysis identified Vitamin C, saturated fatty acids, selenium, phosphorus, and protein intake as key predictors of PAD.

CONCLUSIONS

This is the first study to apply ML algorithms to examine nutrient intake and PAD in a general population. Vitamin C and phosphorus showed negative correlations with PAD, while saturated fatty acids, protein, and selenium exhibited positive associations.

Machine learning meta-analysis identifies individual characteristics moderating cognitive intervention efficacy for anxiety and depression symptoms

Cognitive training is a promising intervention for psychological distress; however, its effectiveness has yielded inconsistent outcomes across studies. This research is a pre-registered individual-level meta-analysis to identify factors contributing to cognitive training efficacy for anxiety and depression symptoms. Machine learning methods, alongside traditional statistical approaches, were employed to analyze 22 datasets with 1544 participants who underwent working memory training, attention bias modification, interpretation bias modification, or inhibitory control training. Baseline depression and anxiety symptoms were found to be the most influential factor, with individuals with more severe symptoms showing the greatest improvement. The number of training sessions was also important, with more sessions yielding greater benefits. Cognitive trainings were associated with higher predicted improvement than control conditions, with attention and interpretation bias modification showing the most promise. Despite the limitations of heterogeneous datasets, this investigation highlights the value of large-scale comprehensive analyses in guiding the development of personalized training interventions.

Machine learning-assisted cancer diagnosis in patients with paraneoplastic autoantibodies

PURPOSE

Paraneoplastic syndromes (PNS) are a group of rare disorders triggered by an immune response to malignancy, characterized by diverse neurological, muscular, and systemic symptoms. This study aims to leverage machine learning to develop a predictive model for cancer diagnosis in patients with paraneoplastic autoantibodies.

METHODS

Demographic data included age and sex, and presenting symptoms were recorded. Laboratory data comprised serum or cerebrospinal fluid (CSF) paraneoplastic autoantibody panels. The study included participants who tested positive for at least one autoantibody. Naive Bayes model was used to predict cancer presence. Model performance was evaluated using sensitivity, specificity, likelihood ratios, predictive values, AUC-ROC, Brier score, and overall accuracy. Feature importance was assessed using SHapley Additive exPlanations (SHAP) values. A graphical user interface (GUI)-based application was developed to facilitate model use.

RESULTS

The study included 116 participants, with an average age of 57.1 years and a higher proportion of females (53.4%). The most common presenting symptom was ''Motor'' (40.5%), followed by ''Cognitive'' (17.2%) and ''Bulbar'' (15.5%) symptoms. Cancer was present in 23 participants (19.8%). The Naive Bayes model demonstrated high performance with a sensitivity of 85.71% and specificity of 100.00%. The AUC-ROC was 0.9795, indicating excellent diagnostic capability. Age and the presence or absence of specific autoantibodies were significant predictors of cancer.

CONCLUSION

Machine learning models, such as the Naive Bayes classifier developed in this study, can accurately stratify cancer risk in patients with positive paraneoplastic autoantibodies.

Construction of a prognostic prediction model for colorectal cancer based on 5-year clinical follow-up data

Colorectal cancer (CRC) is a prevalent malignant tumor that presents significant challenges to both public health and healthcare systems. The aim of this study was to develop a machine learning model based on five years of clinical follow-up data from CRC patients to accurately identify individuals at risk of poor prognosis. This study included 411 CRC patients who underwent surgery at Yixing Hospital and completed the follow-up process. A modeling dataset containing 73 characteristic variables was established by collecting demographic information, clinical blood test indicators, histopathological results, and additional treatment-related information. Decision tree, random forest, support vector machine, and extreme gradient boosting (XGBoost) models were selected for modeling based on the features identified through recursive feature elimination (RFE). The Cox proportional hazards model was used as the baseline for model comparison. During the model training process, hyperparameters were optimized using a grid search method. The model performance was comprehensively assessed using multiple metrics, including accuracy, F1 score, Brier score, sensitivity, specificity, positive predictive value, negative predictive value, receiver operating characteristic curve, calibration curve, and decision curve analysis curve. For the selected optimal model, the decision-making process was interpreted using the SHapley Additive exPlanations (SHAP) method. The results show that the optimal RFE-XGBoost model achieved an accuracy of 0.83 (95% CI 0.76-0.90), an F1 score of 0.81 (95% CI 0.72-0.88), and an area under the receiver operating characteristic curve of 0.89 (95% CI 0.82-0.94). Furthermore, the model exhibited superior calibration and clinical utility. SHAP analysis revealed that increased perioperative transfusion quantity, higher tumor AJCC stage, elevated carcinoembryonic antigen level, elevated carbohydrate antigen 19-9 (CA19-9) level, advanced age, and elevated carbohydrate antigen 125 (CA125) level were correlated with increased individual mortality risk. The RFE-XGBoost model demonstrated excellent performance in predicting CRC patient prognosis, and the application of the SHAP method bolstered the model's credibility and utility.

Development of an interpretable machine learning model based on CT radiomics for the prediction of post acute pancreatitis diabetes mellitus

This study sought to establish and validate an interpretable CT radiomics-based machine learning model capable of predicting post-acute pancreatitis diabetes mellitus (PPDM-A), providing clinicians with an effective predictive tool to aid patient management in a timely fashion. Clinical and imaging data from 271 patients who had undergone enhanced CT scans after first-episode acute pancreatitis from March 2017-June 2023 were retrospectively analyzed. Patients were classified into PPDM-A (n = 109) and non-PPDM-A groups (n = 162), and split into training (n = 189) and testing (n = 82) cohorts at a 7:3 ratio. 1223 radiomic features were extracted from CT images in the plain, arterial and venous phases, respectively. The radiomics model was developed based on the optimal features retained after dimensionality reduction, utilizing the extreme gradient boosting (XGBoost) algorithm. Five-fold cross-validation of the model was used to assess the performance of the model in the training and testing cohorts. The clinical performance of the model was assessed through a decision curve analysis, while insight into the predictions derived from this model was derived from Shapley additive explanations (SHAP). The final model incorporated five key radiomic features, and achieved area under the curve values in the training and testing cohorts of 0.947 (95% CI 0.915-0.979) and 0.901 (95% CI 0.838-0.964), respectively. SHAP analyses indicated that textural features were key features relevant to the prediction of PPDM-A incidence. The interpretable CT radiomics-based model developed in this study demonstrated good performance, enabling timely and effective interventions with the potential to improve patient outcomes.

Using machine learning models to identify severe subjective cognitive decline and related factors in nurses during the menopause transition: a pilot study

OBJECTIVE

This study aims to develop and validate a machine learning model for identifying individuals within the nursing population experiencing severe subjective cognitive decline (SCD) during the menopause transition, along with their associated factors.

METHODS

A secondary analysis was performed using cross-sectional data from 1,264 nurses undergoing the menopause transition. The data set was randomly split into training (75%) and validation sets (25%), with the Bortua algorithm employed for feature selection. Seven machine learning models were constructed and optimized. Model performance was assessed using area under the receiver operating characteristic curve, accuracy, sensitivity, specificity, and F1 score. Shapley Additive Explanations analysis was used to elucidate the weights and characteristics of various factors associated with severe SCD.

RESULTS

The average SCD score among nurses in the menopause transition was (5.38 ± 2.43). The Bortua algorithm identified 13 significant feature factors. Among the seven models, the support vector machine exhibited the best overall performance, achieving an area under the receiver operating characteristic curve of 0.846, accuracy of 0.789, sensitivity of 0.753, specificity of 0.802, and an F1 score of 0.658. The two variables most strongly associated with SCD were menopausal symptoms and the stage of menopause.

CONCLUSIONS

The machine learning models effectively identify individuals with severe SCD and the related factors associated with severe SCD in nurses during the menopause transition. These findings offer valuable insights for the management of cognitive health in women undergoing the menopause transition.

Interpretable machine learning models for predicting skip metastasis in cN0 papillary thyroid cancer based on clinicopathological and elastography radiomics features

Background

Skip lymph node metastasis (SLNM) in papillary thyroid cancer (PTC) involves cancer cells bypassing central nodes to directly metastasize to lateral nodes, often undetected by standard preoperative ultrasonography. Although multiple models exist to identify SLNM, they are inadequate for clinically node-negative (cN0) patients, resulting in underestimated metastatic risks and compromised treatment effectiveness. Our study aims to develop and validate a machine learning (ML) model that combines elastography radiomics with clinicopathological data to predict pre-surgical SLNM risk in cN0 PTC patients with increased risk of lymph node metastasis (LNM), improving their treatment strategies.

Methods

Our study conducted a retrospective analysis of 485 newly diagnosed primary PTC patients, divided into training and external validation cohorts. Patients were categorized into SLNM and non-SLNM groups based on follow-up outcomes and postoperative pathology. We collected preoperative clinicopathological data and extracted, standardized radiomics features from elastography imaging to develop various ML models. These models were internally validated using radiomics and clinicopathological data, with the optimal model's feature importance analyzed through the Shapley Additive Explanations (SHAP) approach and subsequently externally validated.

Results

In our study of 485 patients, 67 (13.8%) exhibited SLNM. The extreme gradient boosting (XGBoost) model, integrating elastography radiomics with clinicopathological data, demonstrated superior performance in both internal and external validations. SHAP analysis identified five key determinants of SLNM: three radiomics features from elastography images, one clinical variable, and one pathological variable.

Conclusion

Our evaluation highlights the XGBoost model, which integrates elastography radiomics and clinicopathological data, as the most effective ML approach for the prediction of SLNM in cN0 PTC patients with increased risk of LNM. This innovative model significantly enhances the accuracy of risk assessments for SLNM, enabling personalized treatments that could reduce postoperative metastases in these patients.

Enhanced phase prediction of high-entropy alloys through machine learning and data augmentation

The phase structure information of high-entropy alloys (HEAs) is critical for their design and application, as different phase configurations are associated with distinct chemical and physical properties. However, the broad range of elements in HEAs presents significant challenges for precise experimental design and rational theoretical modeling and simulation. To address these challenges, machine learning (ML) methods have emerged as powerful tools for phase structure prediction. In this study, we use a dataset of 544 HEA configurations to predict phases, including 248 intermetallic, 131 solid solution, and 165 amorphous phases. To mitigate the limitations imposed by the small dataset size, we employ a Generative Adversarial Network (GAN) to augment the existing data. Our results show a significant improvement in model performance with data augmentation, achieving an average accuracy of 94.77% across ten random seeds. Validation on an independent dataset confirms the model's reliability and real-world applicability, achieving 100% prediction accuracy. We also predict FCC and BCC phases for SS HEAs based on elemental composition, achieving a peak accuracy of 98%. Furthermore, feature importance analysis identifies correlations between compositional features and phase formation tendencies, which are consistent with experimental observations. This work proposes an effective strategy to enhance the accuracy and generalizability of machine learning models in phase structure prediction, thus promoting the accelerated design of HEAs for a wide range of applications.

Improving Clinical Decisions in IR: Interpretable Machine Learning Models for Predicting Ascites Improvement after Transjugular Intrahepatic Portosystemic Shunt Procedures

PURPOSE

To evaluate the potential of interpretable machine learning (ML) models to predict ascites improvement in patients undergoing transjugular intrahepatic portosystemic shunt (TIPS) placement for refractory ascites.

MATERIALS AND METHODS

In this retrospective study, 218 patients with refractory ascites who underwent TIPS placement were analyzed. Data on 29 demographic, clinical, and procedural features were collected. Ascites improvement was defined as reduction in the need of paracentesis by 50% or more at the 1-month follow-up. Univariate statistical analysis was performed. Data were split into train and test sets. Feature selection was performed using a wrapper-based sequential feature selection algorithm. Two ML models were built using support vector machine (SVM) and CatBoost algorithms. Shapley additive explanations values were calculated to assess interpretability of ML models. Performance metrics were calculated using the test set.

RESULTS

Refractory ascites improved in 168 (77%) patients. Higher sodium (Na; 136 mEq/L vs 134 mEq/L; P = .001) and albumin (2.91 g/dL vs 2.68 g/dL; P = .03) levels, lower creatinine levels (1.01 mg/dL vs 1.17 mg/dL; P = .04), and lower Model for End-stage Liver Disease (MELD) (13 vs 15; P = .01) and MELD-Na (15 vs 17.5, P = .002) scores were associated with significant improvement, whereas main portal vein puncture was associated with a lower improvement rate (P = .02). SVM and CatBoost models had accuracy ratios of 83% and 87%, with area under the curve values of 0.83 and 0.87, respectively. No statistically significant difference was found between performances of the models in DeLong test (P = .3).

CONCLUSIONS

ML models may have potential in patient selection for TIPS placement by predicting the improvement in refractory ascites.

Acoustic speech features are associated with late-life depression and apathy symptoms: Preliminary findings

BACKGROUND

Late-life depression (LLD) is a heterogenous disorder related to cognitive decline and neurodegenerative processes, raising a need for the development of novel biomarkers. We sought to provide preliminary evidence for acoustic speech signatures sensitive to LLD and their relationship to depressive dimensions.

METHODS

Forty patients (24 female, aged 65-82 years) were assessed with the Geriatric Depression Scale (GDS). Vocal features were extracted from speech samples (reading a pre-written text) and tested as classifiers of LLD using random forest and XGBoost models. Post hoc analyses examined the relationship between these acoustic features and specific depressive dimensions.

RESULTS

The classification models demonstrated moderate discriminative ability for LLD with receiver operating characteristic = 0.78 for random forest and 0.84 for XGBoost in an out-of-sample testing set. The top classifying features were most strongly associated with the apathy dimension (R 2 = 0.43).

DISCUSSION

Acoustic vocal features that may support the diagnosis of LLD are preferentially associated with apathy.

Highlights

The depressive dimensions in late-life depression (LLD) have different cognitive correlates, with apathy characterized by more pronounced cognitive impairment.Acoustic speech features can predict LLD. Using acoustic features, we were able to train a random forest model to predict LLD in a held-out sample.Acoustic speech features that predict LLD are preferentially associated with apathy. These results indicate a predominance of apathy in the vocal signatures of LLD, and suggest that the clinical heterogeneity of LLD should be considered in development of acoustic markers.

Use of Machine Learning to Predict the Incidence of Type 2 Diabetes Among Relatively Healthy Adults: A 10-Year Longitudinal Study in Taiwan

Background: The prevalence of diabetes is increasing worldwide, particularly in the Pacific Ocean island nations. Although machine learning (ML) models and data mining approaches have been applied to diabetes research, there was no study utilizing ML models to predict diabetes incidence in Taiwan. We aimed to predict the onset of diabetes in order to raise health awareness, thereby promoting any necessary lifestyle modifications and help mitigate disease burden. Methods: The research dataset used in the study was retrieved from the Clinical Data Center of Taichung Veterans General Hospital. We collected data from the available electronic health records with a total of 33 items being employed for model construction. Individuals with diabetes and those with missing data were excluded. Ultimately, 6687 adults were included in the final analysis, where we implemented three different ML algorithms, including logistic regression (LR), random forest (RF) and extreme gradient boosting (XGBoost) in order to predict diabetes. Results: The top five important factors involved in the prediction model were glycated hemoglobin (HbA1c), fasting blood glucose, weight, free thyroxine (fT4), and triglycerides (TG). Notably, random forest, logistic regression, and XGBoost reached 99%, 99%, and 98% accuracy, respectively. fT4 seems to be one of the significant features in predicting the onset of diabetes. Moreover, this would be the first study using machine learning models to predict diabetes that has demonstrated the importance of thyroid hormone. Conclusions: A total of 33 items were able to be put into the machine learning model in order to predict diabetes with promising accuracy. In comparison to prior studies on machine learning models, this study not only identified similar key factors for predicting diabetes but also highlighted the significance of thyroid hormones, a factor that was previously overlooked. Moreover, it highlighted the relevance of predicting type 2 diabetes using more affordable methods, which would be useful for clinical healthcare professionals and endocrinologists who apply the models to clinical practice.

Predictive modelling of hospital-acquired infection in acute ischemic stroke using machine learning

Hospital-acquired infections (HAIs) are serious complication for patients with acute ischemic stroke (AIS), often resulting in poor functional outcomes. However, no existing model can specifically predict HAI in AIS patients. Therefore, we employed the Gradient Boosting matching learning algorithm to establish predictive models for HAI occurrence in AIS patients and poor 30-day functional outcomes (modified Rankin Scale > 2) in AIS patients with HAI by analyzing electronic health records from 6560 AIS patients. Model performance was evaluated through internal cross-validation and external validation using an independent cohort of 3521 AIS patients. The established models demonstrated robust predictive performance for HAI in AIS patients, achieving area under the receiver operating characteristic curves (AUROCs) of 0.857 ± 0.008 during internal validation and 0.825 ± 0.002 during external validation. For AIS patients with HAI, the second model effectively predict poor 30-day functional outcomes, with AUROCs of 0.905 ± 0.009 during internal validation and 0.907 ± 0.002 during external validation. In conclusion, machine learning models effectively identify the HAI occurrence and predict poor 30-day functional outcomes in AIS patients with HAI. Future prospective studies are crucial for validating and refining these models for clinical application, as well as for developing an accessible flowchart or scoring system to enhance clinical practices.

Analyzing the impact of socioeconomic indicators on gender inequality in Sri Lanka: A machine learning-based approach

This study conducts a comprehensive analysis of gender inequality in Sri Lanka, focusing on the relationship between key socioeconomic factors and the Gender Inequality Index (GII) from 1990 to 2022. By applying machine learning techniques, including Decision Trees and Ensemble methods, the study investigates the influence of economic indicators such as GDP per capita, government expenditure, government revenue, and unemployment rates on gender disparities. The analysis reveals that higher GDP and government revenues are associated with reduced gender inequality, while greater unemployment rates exacerbate disparities. Explainable AI techniques (SHAP) further highlight the critical role of government policies and economic development in shaping gender equality. These findings offer specific insights for policymakers to design targeted interventions aimed at reducing gender gaps in Sri Lanka, particularly by prioritizing economic growth and inclusive public spending.

Study on SHP2 Conformational Transition and Structural Characterization of Its High-Potency Allosteric Inhibitors by Molecular Dynamics Simulations Combined with Machine Learning

The src-homology 2 domain-containing phosphatase 2 (SHP2) is a human cytoplasmic protein tyrosine phosphatase that plays a crucial role in cellular signal transduction. Aberrant activation and mutations of SHP2 are associated with tumor growth and immune suppression, thus making it a potential target for cancer therapy. Initially, researchers sought to develop inhibitors targeting SHP2's catalytic site (protein tyrosine phosphatase domain, PTP). Due to limitations such as conservativeness and poor membrane permeability, SHP2 was once considered a challenging drug target. Nevertheless, with the in-depth investigations into the conformational switch mechanism from SHP2's inactive to active state and the emergence of various SHP2 allosteric inhibitors, new hope has been brought to this target. In this study, we investigated the interaction models of various allosteric inhibitors with SHP2 using molecular dynamics simulations. Meanwhile, we explored the free energy landscape of SHP2 activation using enhanced sampling technique (meta-dynamics simulations), which provides insights into its conformational changes and activation mechanism. Furthermore, to biophysically interpret high-dimensional simulation trajectories, we employed interpretable machine learning methods, specifically extreme gradient boosting (XGBoost) with Shapley additive explanations (SHAP), to comprehensively analyze the simulation data. This approach allowed us to identify and highlight key structural features driving SHP2 conformational dynamics and regulating the activity of the allosteric inhibitor. These studies not only enhance our understanding of SHP2's conformational switch mechanism but also offer crucial insights for designing potent allosteric SHP2 inhibitors and addressing drug resistance issues.

Robust and interpretable deep learning system for prognostic stratification of extranodal natural killer/T-cell lymphoma

PURPOSE

Extranodal natural killer/T-cell lymphoma (ENKTCL) is an hematologic malignancy with prognostic heterogeneity. We aimed to develop and validate DeepENKTCL, an interpretable deep learning prediction system for prognosis risk stratification in ENKTCL.

METHODS

A total of 562 patients from four centers were divided into the training cohort, validation cohort and test cohort. DeepENKTCL combined a tumor segmentation model, a PET/CT fusion model, and prognostic prediction models. RadScore and TopoScore were constructed using radiomics and topological features derived from fused images, with SHapley Additive exPlanations (SHAP) analysis enhancing interpretability. The final prognostic models, termed FusionScore, were developed for predicting progression-free survival (PFS) and overall survival (OS). Performance was assessed using area under the receiver operator characteristic curve (AUC), time-dependent C-index, clinical decision curves (DCA), and Kaplan-Meier (KM) curves.

RESULTS

The tumor segmentation model accurately delineated the tumor lesions. RadScore (AUC: 0.908 for PFS, 0.922 for OS in validation; 0.822 for PFS, 0.867 for OS in test) and TopoScore (AUC: 0.756 for PFS, 0.805 for OS in validation; 0.689 for PFS, 0.769 for OS in test) both exhibited potential prognostic capability. The time-dependent C-index (0.897 for PFS, 0.928 for OS in validation; 0.894 for PFS, 0.868 for OS in test) and DCA indicated that FusionScore offers significant prognostic performance and superior net clinical benefits compared to existing models. KM survival analysis showed that higher FusionScores correlated with poorer PFS and OS across all cohorts.

CONCLUSION

DeepENKTCL provided a robust and interpretable framework for ENKTCL prognosis, with the potential to improve patient outcomes and guide personalized treatment.

The anesthesiologist's guide to critically assessing machine learning research: a narrative review

Artificial Intelligence (AI), especially Machine Learning (ML), has developed systems capable of performing tasks that require human intelligence. In anesthesiology and other medical fields, AI applications can improve the precision and efficiency of daily clinical practice, and can also facilitate a personalized approach to patient care, which can lead to improved outcomes and quality of care. ML has been successfully applied in various settings of daily anesthesiology practice, such as predicting acute kidney injury, optimizing anesthetic doses, and managing postoperative nausea and vomiting. The critical evaluation of ML models in healthcare is crucial to assess their validity, safety, and clinical applicability. Evaluation metrics allow an objective statistical assessment of model performance. Tools such as Shapley Values (SHAP) help interpret how individual variables contribute to model predictions. Transparency in reporting is key in maintaining trust in these technologies and to ensure their use follows ethical principles, aiming to reduce safety concerns while also benefiting patients. Understanding evaluation metrics is essential, as they provide detailed information on model performance and their ability to discriminate between individual class rates. This article offers a comprehensive framework in assessing the validity, applicability, and limitations of models, guiding responsible and effective integration of ML technologies into clinical practice. A balance between innovation, patient safety and ethical considerations must be pursued.

NETosis Genes and Pathomic Signature: A Novel Prognostic Marker for Ovarian Serous Cystadenocarcinoma

To evaluate the prognostic significance and molecular mechanism of NETosis markers in ovarian serous cystadenocarcinoma (OSC), we constructed a machine learning-based pathomic model utilizing hematoxylin and eosin (H&E) slides. We analyzed 333 patients with OSC from The Cancer Genome Atlas for prognostic-related neutrophil extracellular trap formation (NETosis) genes through bioinformatics analysis. Pathomic features were extracted from 54 cases with complete pathological images, genetic matrices, and clinical information. Two pathomic prognostic models were constructed using support vector machine (SVM) and logistic regression (LR) algorithms. Additionally, we established a predictive scoring system that integrated pathomic scores based on the NETcluster subtypes and clinical signature. We identified four NETosis genes significantly correlated with OSC prognosis, which were functionally associated with immune response, somatic mutations, tumor invasion, and metastasis. Five robust pathomic features were selected for overall survival prediction. The LR and SVM pathomic models demonstrated strong predictive performance for the NETcluster subtype classification through five-fold cross-validation. Time-dependent ROC analysis revealed excellent prognostic capability of the LR pathomic model's score for the overall survival (AUC values of 0.658, 0.761, and 0.735 at 36, 48, and 60 months, respectively), further validated by Kaplan-Meier analysis. The expression levels of NETosis genes greatly affected OSC patients' prognoses. The pathomic analysis of H&E slide pathological images provides an effective approach for predicting both NETcluster subtype and overall survival in OSC patients.

Machine Learning Models for Efficient Property Prediction of ABX3 Materials: A High-Throughput Approach

Recently, ABX3 materials have garnered significant attention due to their diverse applications in photovoltaics, catalysis, and optoelectronics as well as their remarkable efficiency in energy conversion. However, progress has been somewhat slow due to the high expenses of the experiment or the time-consuming density functional theory (DFT) calculation. In this study, we utilized the extreme gradient boosting (XGBoost) algorithm to facilitate the discovery and characterization of ABX3 compounds based on vast data sets generated by DFT calculations. While the XGBoost algorithm provides a powerful tool for accelerating the discovery of ABX3 compounds, it is crucial to acknowledge that different DFT approximation levels can significantly impact the predicted band gaps, potentially introducing discrepancies when compared with experimental values. In the first step, we predict the space group of 13947 oxides and halides using the Open Quantum Materials Database and elemental features. Our analysis yields classification accuracies ranging from 82.39% to 99.14% across these materials. Following this, XGBoost regression algorithms are employed to interrogate the data set, enabling predictions of volume (achieving an optimal accuracy of 98.41%, with a mean absolute error (MAE) of 2.395 Å3 and a root-mean-square error (RMSE) of 4.416 Å3), formation energy (an optimal accuracy of 97.36%, with an MAE of 0.075 eV/atom and an RMSE of 0.132 eV/atom), and band gap energy (an optimal accuracy of 87.00%, an MAE of 0.391 eV, and an RMSE of 0.574 eV). Finally, these prediction models are employed to identify the possible space groups for each of the 1252 new ABX3 formulas. Then, we predict the volume, the formation energy, and the band gap energy for each candidate space group. Through these predictive models, machine learning accelerates the exploration of new materials with enhanced performance and functionality.

Cytokine profiles as predictors of HIV incidence using machine learning survival models and statistical interpretable techniques

HIV remains a critical global health issue, with an estimated 39.9 million people living with the virus worldwide by the end of 2023 (according to WHO). Although the epidemic's impact varies significantly across regions, Africa remains the most affected. In the past decade, considerable efforts have focused on developing preventive measures, such as vaccines and pre-exposure prophylaxis, to combat sexually transmitted HIV. Recently, cytokine profiles have gained attention as potential predictors of HIV incidence due to their involvement in immune regulation and inflammation, presenting new opportunities to enhance preventative strategies. However, the high-dimensional, time-varying nature of cytokine data collected in clinical research, presents challenges for traditional statistical methods like the Cox proportional hazards (PH) model to effectively analyze survival data related to HIV. Machine learning (ML) survival models offer a robust alternative, especially for addressing the limitations of the PH model's assumptions. In this study, we applied survival support vector machine (SSVM) and random survival forest (RSF) models using changes or means in cytokine levels as predictors to assess their association with HIV incidence, evaluate variable importance, measure predictive accuracy using the concordance index (C-index) and integrated Brier score (IBS) and interpret the model's predictions using Shapley additive explanations (SHAP) values. Our results indicated that RSFs models outperformed SSVMs models, with the difference covariate model performing better than the mean covariate model. The highest C-index for SSVM was 0.7180 under the difference covariate model, while for RSF, it reached 0.8801 under the difference covariate model using the log-rank split rule. Key cytokines identified as positive predictors of HIV incidence included TNF-A, BASIC-FGF, IL-5, MCP-3, and EOTAXIN, while 29 cytokines were negative predictors. Baseline factors such as condom use frequency, treatment status, number of partners, and sexual activity also emerged as significant predictors. This study underscored the potential of cytokine profiles for predicting HIV incidence and highlighted the advantages of RSFs models in analyzing high-dimensional, time-varying data over SSVMs. It further through ablation studies emphasized the importance of selecting key features within mean and difference based covariate models to achieve an optimal balance between model complexity and predictive accuracy.

Machine learning-based prediction of in-hospital mortality for critically ill patients with sepsis-associated acute kidney injury

OBJECTIVES

This study aims to develop and validate a prediction model in-hospital mortality in critically ill patients with sepsis-associated acute kidney injury (SA-AKI) based on machine learning algorithms.

METHODS

Patients who met the criteria for inclusion were identified in the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database and divided according to the validation (n = 2440) and development (n = 9756, 80%) queues. Ensemble stepwise feature selection method was used to screen for effective features. The prediction models of short-term mortality were developed by seven machine learning algorithms. Ten-fold cross-validation was used to verify the performance of the algorithm in the development queue. The area under the receiver operating characteristic curve (ROC-AUC) was used to evaluate the differentiation accuracy and performance of the prediction model in the validation queue. The best-performing model was interpreted by Shapley additive explanations (SHAP).

RESULTS

A total of 12,196 patients were enrolled in this study. Eleven variables were finally chosen to develop the prediction model. The AUC of the random forest (RF) model was the highest value both in the Ten-fold cross-validation and evaluation (AUC: 0.798, 95% CI: 0.774-0.821). According to the SHAP plots, old age, low Glasgow Coma Scale (GCS) score, high AKI stage, reduced urine output, high Simplified Acute Physiology Score (SAPS II), high respiratory rate, low temperature, low absolute lymphocyte count, high creatinine level, dysnatremia, and low body mass index (BMI) increased the risk of poor prognosis.

CONCLUSIONS

The RF model developed in this study is a good predictor of in-hospital mortality for patients with SA-AKI in the intensive care unit (ICU), which may have potential applications in mortality prediction.

Prediction of the End of a Romantic Relationship in Peruvian Youth and Adults: A Machine Learning Approach

This study explores the effectiveness of machine learning models in predicting the end of romantic relationships among Peruvian youth and adults, considering various socioeconomic and personal attributes. The study implements logistic regression, gradient boosting, support vector machines, and decision trees on SMOTE-balanced data using a sample of 429 individuals to improve model robustness and accuracy. Using stratified random sampling, the data is split into training (80%) and validation (20%) sets. The models are evaluated through 10-fold cross-validation, focusing on accuracy, F1-score, AUC, sensitivity, and specificity metrics. The Random Forest model is the preferred algorithm because of its superior performance in all evaluation metrics. Hyperparameter tuning was conducted to optimize the model, identifying key predictors of relationship dissolution, including negative interactions, desire for emotional infidelity, and low relationship satisfaction. SHAP analysis was utilized to interpret the directional impact of each variable on the prediction outcomes. This study underscores the potential of machine learning tools in providing deep insights into relationship dynamics, suggesting their application in personalized therapeutic interventions to enhance relationship quality and reduce the incidence of breakups. Future research should incorporate larger and more diverse datasets to further validate these findings.

Quantification of intratumoral heterogeneity using habitat-based MRI radiomics to identify HER2-positive, -low and -zero breast cancers: a multicenter study

BACKGROUND

Human epidermal growth factor receptor 2-targeted (HER2) therapy with antibody-drug conjugates has proven effective for patients with HER2-low breast cancer. However, intratumoral heterogeneity (ITH) poses a great challenge in identifying HER2-low tumors. ITH signatures were developed by quantifying ITH to differentiate HER2-positive, -low and -zero breast cancers.

METHODS

This retrospective study included 614 patients from two institutions. The study was structured into two primary tasks: task 1 was to differentiate between HER2-positive and -negative tumors, followed by task 2 to differentiate HER2-low and -zero tumors. Whole-tumor radiomics features and habitat radiomics features were extracted from MRI to construct the radiomics and ITH signatures. Multivariable logistic regression analysis was used to determine significant independent predictors. A combined model integrating significant clinicopathologic variables, radiomics signature, and ITH signature was developed for task (1) Subsequently, the better-performing model was established using the same approach for task (2) The area under the receiver operating characteristic curve (AUC) was used to assess the performance of each model.

RESULTS

Task 1 comprised 614 patients (training, n = 348; validation, n = 149; and test cohorts, n = 117). Task 2 encompassed 501 patients (training, n = 283; validation, n = 122; and test cohorts, n = 96). For task1, the ITH signature showed outstanding performance, achieving AUCs of 0.81, 0.81, and 0.81 in the training, validation and test cohorts, respectively. The combined model achieved improved performance, with AUCs of 0.83, 0.84 and 0.83 across the three cohorts, respectively. For task2, the ITH signature maintained superior performance, with AUCs of 0.94, 0.93 and 0.84 across the training, validation and test cohorts, respectively. Multivariable logistic regression analysis indicated that none of the clinicopathologic characteristics were retained as predictors associated with odds of HER2-low tumors.

CONCLUSIONS

Our study developed ITH signatures that quantified ITH using habitat-based MRI radiomics, achieving outstanding performance in differentiating HER2-postive and -negative tumors, and further differentiating HER2-low and -zero breast cancers.

Machine Learning Research Trends in Traditional Chinese Medicine: A Bibliometric Review

Background

Integrating Traditional Chinese Medicine (TCM) knowledge with modern technology, especially machine learning (ML), has shown immense potential in enhancing TCM diagnostics and treatment. This study aims to systematically review and analyze the trends and developments in ML applications in TCM through a bibliometric analysis.

Methods

Data for this study were sourced from the Web of Science Core Collection. Data were analyzed and visualized using Microsoft Office Excel, Bibliometrix, and VOSviewer.

Results

474 documents were identified. The analysis revealed a significant increase in research output from 2000 to 2023, with China leading in both the number of publications and research impact. Key research institutions include the Shanghai University of Traditional Chinese Medicine and the China Academy of Chinese Medical Sciences. Major research hotspots identified include ML applications in TCM diagnosis, network pharmacology, and tongue diagnosis. Additionally, chemometrics with ML are highlighted for their roles in quality control and authentication of TCM products.

Conclusion

This study provides a comprehensive overview of ML applications' development trends and research landscape in TCM. The integration of ML has led to significant advancements in TCM diagnostics, personalized medicine, and quality control, paving the way for the modernization and internationalization of TCM practices. Future research should focus on improving model interpretability, fostering international collaborations, and standardized reporting protocols.

Predictive value of triglyceride-glucose index for the occurrence of acute respiratory failure in asthmatic patients of MIMIC-IV database

This study aims to investigate the association between the triglyceride-glucose (TyG) index and the occurrence of acute respiratory failure in asthma patients. This retrospective observational cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV 2.2) database. The primary outcome was the development of acute respiratory failure in asthma patients. Initially, the Boruta algorithm and SHapley Additive exPansions were applied to preliminarily determine the feature importance of the TyG index, and a risk prediction model was constructed to evaluate its predictive ability. Secondly, Logistic regression proportional hazards models were employed to assess the association between the TyG index and acute respiratory failure in asthma patients. Finally, subgroup analyses were conducted for sensitivity analyses to explore the robustness of the results. A total of 751 asthma patients were included in the study. When considering the TyG index as a continuous variable, logistic regression analysis revealed that in the unadjusted Model 1, the odds ratio (OR) was 2.381 (95% CI: 1.857-3.052; P < 0.001), in Model II, the OR was 2.456 (95% CI: 1.809-3.335; P < 0.001), and in the multivariable-adjusted model, the OR was 1.444 (95% CI: 1.029-2.028; P = 0.034). A consistent association was observed between the TyG index and the risk of acute respiratory failure in asthma patients. No significant interaction was found between the TyG index and various subgroups (P > 0.05). Furthermore, machine learning results indicated that an elevated TyG index was a significant feature predictive of respiratory failure in asthma patients. The baseline risk model achieved an AUC of 0.743 (95% CI: 0.679-0.808; P < 0.05), whereas the combination of the baseline risk model with the TyG index yielded an AUC of 0.757 (95% CI: 0.694-0.821; P < 0.05). The TyG index can serve as a predictive indicator for acute respiratory failure in asthma patients, albeit confirmation of these findings requires larger-scale prospective studies.

Machine learning for predicting in-hospital mortality in elderly patients with heart failure combined with hypertension: a multicenter retrospective study

BACKGROUND

Heart failure combined with hypertension is a major contributor for elderly patients (≥ 65 years) to in-hospital mortality. However, there are very few models to predict in-hospital mortality in such elderly patients. We aimed to develop and test an individualized machine learning model to assess risk factors and predict in-hospital mortality in in these patients.

METHODS

From January 2012 to December 2021, this study collected data on elderly patients with heart failure and hypertension from the Chongqing Medical University Medical Data Platform. Least absolute shrinkage and the selection operator was used for recognizing key clinical variables. The optimal predictive model was chosen among eight machine learning algorithms on the basis of area under curve. SHapley Additive exPlanations and Local Interpretable Model-agnostic Explanations was employed to interpret the outcome of the predictive model.

RESULTS

This study ultimately comprised 4647 elderly individuals with hypertension and heart failure. The Random Forest model was chosen with the highest area under curve for 0.850 (95% CI 0.789-0.897), high accuracy for 0.738, recall 0.837, specificity 0.734 and brier score 0.178. According to SHapley Additive exPlanations results, the most related factors for in-hospital mortality in elderly patients with heart failure and hypertension were urea, length of stay, neutrophils, albumin and high-density lipoprotein cholesterol.

CONCLUSIONS

This study developed eight machine learning models to predict in-hospital mortality in elderly patients with hypertension as well as heart failure. Compared to other algorithms, the Random Forest model performed significantly better. Our study successfully predicted in-hospital mortality and identified the factors most associated with in-hospital mortality.

Identifying the toxic mechanisms of emerging electronic contaminations liquid crystal monomers and the construction of a priority control list for graded control

Liquid crystal monomers (LCMs) are identified as emerging organic contaminations with largely unexplored health impacts. To elucidate their toxic mechanisms, support the establishment of environmental discharge and management standards, and promote effective LCMs control, this study constructs a database covering 20,545 potential targets of 1431 LCMs, highlighting 9 key toxic target proteins that disrupt the nervous system and metabolic functions. GO and KEGG pathway analysis suggests LCMs severely affect nervous system, linked to neurodegenerative diseases and mental health disorders, with toxicity variations driven by electronegativity and structural complexity of LCM terminal groups. To achieve tiered control of LCMs, construct toxicity risk control lists for 9 key toxic target proteins, suitable for the graded control of LCMs, management recommendations are provided based on toxicity levels. These lists were validated for reliability and offer reliable toxicity predictions for LCMs. SHAP analysis points to electronic properties, molecular shape, and structural characteristics of LCMs as primary health impact factors. As the first study integrating machine learning with computational toxicology to outline LCMs health impacts, it aims to enhance public understanding of LCM toxicity risks and support the development of environmental standards, effective management of LCM production and emissions, and reduction of public exposure risks.

Development and validation of machine learning models for intraoperative blood transfusion prediction in severe lumbar disc herniation

Lumbar disc herniation (LDH) is a common cause of lower back pain and sciatica, and posterior lumbar interbody fusion (PLIF) is always employed. This multicenter retrospective study investigates predicting intraoperative blood transfusion for LDH patients undergoing PLIF in China. The research includes 6,241 patients from 22 medical centers and employs 8 feature selection methods and 10 machine learning models, including an integrated stacking model. The optimal predictive model was selected based on the receiver operating characteristic area under the curve, clinical applicability, and computational efficiency. Among the evaluated combinations, the simulated annealing support vector machine recursive + stacking model achieved the highest performance with an area under the curve of 0.884, supported by robust calibration and decision curve analyses. A publicly accessible web calculator was developed to assist clinicians in decision-making. This work significantly enhances intraoperative transfusion predictions, providing valuable tools for improving patient management.

Development and Validation of a Machine Learning-Based Nomogram for Prediction of Unplanned Reoperation Postspinal Surgery Within 30 Days

BACKGROUND

Unplanned reoperation postspinal surgery (URPS) leads to prolonged hospital stays, higher costs, decreased patient satisfaction, and adversely affects postoperative rehabilitation. This study aimed to develop and validate prediction models (nomograms) for early URPS risk factors using machine learning methods, aiding spine surgeons in designing prevention strategies, promoting early recovery, reducing complications, and improving patient satisfaction.

METHODS

Medical records of 639 patients who underwent reoperation postspinal surgery from the First Affiliated Hospital of Air Force Medical University (2018-2022) were collected, including baseline indicators, perioperative indicators, and laboratory indicators. After applying inclusion and exclusion criteria, 122 URPS and 155 non-URPS patients were identified and randomly divided into training (82 URPS and 111 non-URPS) and validation (40 URPS and 44 non-URPS) cohorts. Three machine learning methods (least absolute shrinkage and selection operator regression, Random Forest, and Support Vector Machine Recursive Feature Elimination) were used to select feature variables, and their intersection was used to develop the prediction model, tested on the validation cohort.

RESULTS

Six factors-implant, postoperative suction drainage, gelatin sponge, anticoagulants, antibiotics, and disease type-were identified to construct a nomogram diagnostic model. The area under the curve of this nomogram was 0.829 (95% confidence interval 0.771-0.886) in the training cohort and 0.854 (95% confidence interval 0.775-0.933) in the validation cohort. Calibration curves demonstrated satisfactory agreement between predictions and actual probabilities. The decision curve indicated clinical usefulness with a threshold between 1% and 90%.

CONCLUSIONS

The established model can effectively predict URPS in patients and can assist spine surgeons in devising personalized and rational clinical prevention strategies.

Robust Radiomics Models for Predicting HIFU Prognosis in Uterine Fibroids Using SHAP Explanations: A Multicenter Cohort Study

This study sought to develop and validate different machine learning (ML) models that leverage non-contrast MRI radiomics to predict the degree of nonperfusion volume ratio (NVPR) of high-intensity focused ultrasound (HIFU) treatment for uterine fibroids, equipping clinicians with an early prediction tool for decision-making. This study conducted a retrospective analysis on 221 patients with uterine fibroids who received HIFU treatment and were divided into a training set (N = 117), internal validation (N = 49), and an external test set (N = 55). The 851 radiomics features were extracted from T2-weighted imaging (T2WI), and the max-relevance and min-redundancy (mRMR) and the least absolute shrinkage and selection operator (LASSO) regression were applied for feature selection. Several ML models were constructed by logistic regression (LR), decision tree (DT), random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and light gradient boosting machine (LGBM). These models underwent internal and external validation, and the best model's feature significance was assessed via the Shapley additive explanations (SHAP) method. Four significant non-contrast MRI radiomics features were identified, with the SVM model outperforming others in both internal and external validations, and the AUCs of the T2WI models were 0.860, 0.847, and 0.777, respectively. SHAP analysis highlighted five critical predictors of postoperative NVPR degree, encompassing two radiomics features from non-contrast MRI and three clinical data indicators. The SVM model combining radiomics features and clinical parameters effectively predicts NVPR degree post-HIFU, which enables timely and effective interventions of HIFU.

Utilizing machine learning techniques to identify severe sleep disturbances in Chinese adolescents: an analysis of lifestyle, physical activity, and psychological factors

Background

Adolescents often experience difficulties with sleep quality. The existing literature on predicting severe sleep disturbance is limited, primarily due to the absence of reliable tools.

Methods

This study analyzed 1966 university students. All participants were classified into a training set and a validation set at the ratio of 8:2 at random. Participants in the training set were utilized to establish models, and the logistic regression (LR) and five machine learning algorithms, including the eXtreme Gradient Boosting Machine (XGBM), Naïve Bayesian (NB), Support Vector Machine (SVM), Decision Tree (DT), CatBoosting Machine (CatBM), were utilized to develop models. Whereas, those in the validation set were used to validate the developed models.

Results

The incidence of severe sleep disturbance was 5.28% (104/1969). Among all developed models, the XGBM model performed best in AUC (0.872 [95%CI: 0.848-0.896]), followed by the CatBM model (0.853 [95% CI: 0.821-0.878]) and DT model (0.843 [95% CI: 0.801-0.870]), whereas the AUC of the logistic regression model was only 0.822 (95% CI: 0.777-0.856). Additionally, the XGBM model had the best accuracy (0.792), precision (0.780), F1 score (0.796), Brier score (0.143), and log loss (0.444).

Conclusions

The XGBM model may be a useful tool to estimate the risk of experiencing severe sleep disturbance among adolescents.

Machine learning predicts peak oxygen uptake and peak power output for customizing cardiopulmonary exercise testing using non-exercise features

PURPOSE

Cardiopulmonary exercise testing (CPET) is considered the gold standard for assessing cardiorespiratory fitness. To ensure consistent performance of each test, it is necessary to adapt the power increase of the test protocol to the physical characteristics of each individual. This study aimed to use machine learning models to determine individualized ramp protocols based on non-exercise features. We hypothesized that machine learning models will predict peak oxygen uptake ( V ˙ O2peak) and peak power output (PPO) more accurately than conventional multiple linear regression (MLR).

METHODS

The cross-sectional study was conducted with 274 (♀168, ♂106) participants who performed CPET on a cycle ergometer. Machine learning models and multiple linear regression were used to predict V ˙ O2peak and PPO using non-exercise features. The accuracy of the models was compared using criteria such as root mean square error (RMSE). Shapley additive explanation (SHAP) was applied to determine the feature importance.

RESULTS

The most accurate machine learning model was the random forest (RMSE: 6.52 ml/kg/min [95% CI 5.21-8.17]) for V ˙ O2peak prediction and the gradient boosting regression (RMSE: 43watts [95% CI 35-52]) for PPO prediction. Compared to the MLR, the machine learning models reduced the RMSE by up to 28% and 22% for prediction of V ˙ O2peak and PPO, respectively. Furthermore, SHAP ranked body composition data such as skeletal muscle mass and extracellular water as the most impactful features.

CONCLUSION

Machine learning models predict V ˙ O2peak and PPO more accurately than MLR and can be used to individualize CPET protocols. Features that provide information about the participant's body composition contribute most to the improvement of these predictions.

TRIAL REGISTRATION NUMBER

DRKS00031401 (6 March 2023, retrospectively registered).

Interpretable machine learning models based on shear-wave elastography radiomics for predicting cardiovascular disease in diabetic kidney disease patients

BACKGROUND

The risk of cardiovascular complications is significantly elevated in patients with diabetic kidney disease (DKD). Recognizing the link between the progression of DKD and an increased risk of cardiovascular disease (CVD), it is crucial to focus on the early prediction and management of CVD risk factors among these patients to potentially enhance their health outcomes.

OBJECTIVE

This study sought to bridge the existing gap by developing and validating machine learning (ML) models that utilize clinical data and shear wave elastography (SWE) radiomics features to identify patients at risk of CVD, ultimately aiming to improve the management of DKD.

MATERIALS AND METHODS

This study conducted a retrospective analysis of 586 patients with DKD, dividing them into training and external validation cohorts. We categorized patients based on the presence or absence of CVD. Utilizing SWE imaging, we extracted and standardized radiomics features to develop multiple ML models. These models underwent internal validation using radiomics features alone, clinical data, or a combination thereof. The optimal model was then identified, and its feature importance was assessed through the Shapley Additive Explanations (SHAP) method, before proceeding to external validation.

RESULTS

Among the 586 patients analyzed, 30.7% (180/586) were identified as at risk for CVD. The study pinpointed six significant radiomics features related to CVD, alongside six critical pieces of clinical data. The Support Vector Machine (SVM) model outperformed others in both internal and external validations. Further, SHAP analysis highlighted five principal determinants of CVD risk, comprising three clinical indicators and two SWE radiomics features.

CONCLUSIONS

This study highlights the effectiveness of an SVM model that combines clinical and radiomics features in predicting CVD risk among DKD patients. It enables early prediction of CVD in this patient group, thereby supporting the implementation of timely and suitable interventions.

Can machine learning models improve the prediction of surgical site infection in abdominal surgery than traditional statistical models?

OBJECTIVE

To externally validate by revision and update the study on the efficacy of nosocomial infection control (SENIC) model of surgical site infection (SSI) using logistic regression (LR) and machine learning (ML) approaches.

METHODS

A retrospective analysis of hospital database-derived data from patients that had undergone gastrointestinal, colorectal and hernia surgeries (identified by ICD-9-CM). The SENIC index was calculated and fitted in an LR. MLs were developed using decision-tree (DT), random forest (RF), extreme-gradient-boosting (XGBoost) and Naïve Bayes (NB).

RESULTS

The prevalence of an SSI was 3.21% (404 of 12 596 surgeries; 95% confidence interval [CI] 2.91%, 3.53%). The C-statistic for the original SENIC model was 0.668 (95% CI 0.648, 0.688) with an observed/expected (O/E) ratio of 0.998 (interquartile range [IQR] 0.750, 1.047). An updated-SENIC-LR model with six predictors had a C-statistic of 0.768 (95% CI 0.745, 0.790) and O/E ratio of 0.999 (IQR 0.976, 1.004). The performance of MLs considering 14 predictors was poorer than the updated-SENIC-LR with C-statistics of 0.679, 0.675, 0.656 and 0.651 for NB, XGBoost, RF and DT, respectively. Overfitting was detected for ML approaches, particularly for DT, RF and XGBoost.

CONCLUSION

The updated-SENIC-LR model and NB may be useful for monitoring SSI risk following abdominal surgery.

Physiological model-based machine learning for classifying patients with binge-eating disorder (BED) from the Oral Glucose Tolerance Test (OGTT) curve

BACKGROUND AND OBJECTIVE

Binge eating disorder (BED) is the most frequent eating disorder, often confused with obesity, with which it shares several characteristics. Early identification could enable targeted therapeutic interventions. In this study, we propose a hybrid pipeline that, starting from plasma glucose data acquired during the Oral Glucose Tolerance Test (OGTT), allows us to classify the two types of patients through computational modeling and artificial intelligence.

METHODS

The proposed hybrid pipeline integrates a classical mechanistic model of delayed differential equations (DDE) that describes glucose-insulin dynamics with machine learning (ML) methods. Ad hoc techniques, including structural identifiability analysis, have been employed for refining and evaluating the mathematical model. Additionally, a dedicated pipeline for identifying and optimizing model parameters has been applied to obtain reliable estimates. Robust feature extraction and classifier selection processes were developed to ensure the optimal choice of the best-performing classifier.

RESULTS

By leveraging parameters estimated from the mechanistic model alongside easily obtainable patient information (such as glucose levels at 30 and 120 min post-OGTT, glycated hemoglobin (Hb1Ac), body mass index (BMI), and waist circumference), our approach facilitates accurate classification of patients, enabling tailored therapeutic interventions.

CONCLUSION

Initial findings, focusing on correctly categorizing patients with BED based on metabolic data, demonstrate promising outcomes. These results suggest significant potential for refinement, including exploration of alternative mechanistic models and machine learning algorithms, to enhance classification accuracy and therapeutic strategies.

Optimizing feature selection with gradient boosting machines in PLS regression for predicting moisture and protein in multi-country corn kernels via NIR spectroscopy

Differences in moisture and protein content impact both nutritional value and processing efficiency of corn kernels. Near-infrared (NIR) spectroscopy can be used to estimate kernel composition, but models trained on a few environments may underestimate error rates and bias. We assembled corn samples from diverse international environments and used NIR with chemometrics and partial least squares regression (PLSR) to determine moisture and protein. The potential of five feature selection methods to improve prediction accuracy was assessed by extracting sensitive wavelengths. Gradient boosting machines (GBMs), particularly CatBoost and LightGBM, were found to effectively select crucial wavelengths for moisture (1409, 1900, 1908, 1932, 1953, 2174 nm) and protein (887, 1212, 1705, 1891, 2097, 2456 nm). SHAP plots highlighted significant wavelength contributions to model prediction. These results illustrate GBMs' effectiveness in feature engineering for agricultural and food sector applications, including developing multi-country global calibration models for moisture and protein in corn kernels.

Development and validation of a machine learning-based diagnostic model for Parkinson's disease in community-dwelling populations: Evidence from the China health and retirement longitudinal study (CHARLS)

BACKGROUND

Parkinson's disease (PD) is a major neurodegenerative disorder in Middle-aged and elderly people.There is a pressing need for effective predictive models, particularly in chinese population.

OBJECTIVE

This study aims to develop and validate a machine learning-based diagnostic model to identify individuals with PD in community-dwelling populations using data from the China Health and Retirement Longitudinal Study (CHARLS).

METHODS

We utilized data from 19,134 individuals aged 45 and above from the CHARLS dataset, with 265 adults reported to have PD. The external validation cohort included 1500 individuals, with 21 (1.4 %) having PD.The random forest (RF) algorithm was used to develop an interpretable PD prediction model, which was internally validated using 10-fold cross-validation and externally validated with a dataset from Northern Jiangsu People's Hospital. SHapley Additive exPlanation (SHAP) values were employed to elucidate the model's predictions.

RESULTS

The RF model demonstrated robust performance with an Area Under the Curve (AUC) of 0.884 and high sensitivity, specificity, and F1 scores. The model's performance in external validation cohort, highlighting an AUC of 0.82 and an accuracy of 0.99. The model's performance remained consistent across internal and external validation cohorts. SHAP analysis provided insights into the importance and interaction of various predictors, enhancing model interpretability.

CONCLUSION

The study presents a highly accurate and interpretable machine learning-based diagnostic model to identify individuals with PD in middle-aged and older Chinese adults. By combined with predictive risk factors and chronic disease information, the model offers valuable insights for early identification and intervention, potentially mitigating PD progression.

Modeling Temperature-Dependent Photoluminescence Dynamics of Colloidal CdS Quantum Dots Using Long Short-Term Memory (LSTM) Networks

This study addresses the challenge of modeling temperature-dependent photoluminescence (PL) in CdS colloidal quantum dots (QD), where PL properties fluctuate with temperature, complicating traditional modeling approaches. The objective is to develop a predictive model capable of accurately capturing these variations using Long Short-Term Memory (LSTM) networks, which are well suited for managing temporal dependencies in time-series data. The methodology involved training the LSTM model on experimental time-series data of PL intensity and temperature. Through numerical simulation, the model's performance was assessed. Results demonstrated that the LSTM-based model effectively predicted PL trends under different temperature conditions. This approach could be applied in optoelectronics and quantum dot-based sensors for enhanced forecasting capabilities.

Employing machine learning techniques for prediction of micronutrient supplementation status during pregnancy in East African Countries

Micronutrient deficiencies, known as "hidden hunger" or "hidden malnutrition," pose a significant health risk to pregnant women, particularly in low-income countries like the East Africa region. This study employed eight advanced machine learning algorithms to predict the status of micronutrient supplementation among pregnant women in 12 East African countries, using recent demographic health survey (DHS) data. The analysis involved 138,426 study samples, and algorithm performance was evaluated using accuracy, area under the ROC curve (AUC), specificity, precision, recall, and F1-score. Among the algorithms tested, the random forest classifier emerged as the top performer in predicting micronutrient supplementation status, exhibiting excellent evaluation scores (AUC = 0.892 and accuracy = 94.0%). By analyzing mean SHAP values and performing association rule mining, we gained valuable insights into the importance of different variables and their combined impact, revealing hidden patterns within the data. Key predictors of micronutrient supplementation were the mother's education level, employment status, number of antenatal care (ANC) visits, access to media, number of children, and religion. By harnessing the power of machine learning algorithms, policymakers and healthcare providers can develop targeted strategies to improve the uptake of micronutrient supplementation. Key intervention components involve enhancing education, strengthening ANC services, and implementing comprehensive media campaigns that emphasize the importance of micronutrient supplementation. It is also crucial to consider cultural and religious sensitivities when designing interventions to ensure their effectiveness and acceptance within the specific population. Furthermore, researchers are encouraged to explore and experiment with various techniques to optimize algorithm performance, leading to the identification of the most effective predictors and enhanced accuracy in predicting micronutrient supplementation status.

Interpretable machine learning models for predicting short-term prognosis in AChR-Ab+ generalized myasthenia gravis using clinical features and systemic inflammation index

Background

Myasthenia Gravis (MG) is an autoimmune disease that causes muscle weakness in 80% of patients, most of whom test positive for anti-acetylcholine receptor (AChR) antibodies (AChR-Abs). Predicting and improving treatment outcomes are necessary due to varying responses, ranging from complete relief to minimal improvement.

Objective

Our study aims to develop and validate an interpretable machine learning (ML) model that integrates systemic inflammation indices with traditional clinical indicators. The goal is to predict the short-term prognosis (after 6 months of treatment) of AChR-Ab+ generalized myasthenia gravis (GMG) patients to guide personalized treatment strategies.

Methods

We performed a retrospective analysis on 202 AChR-Ab+ GMG patients, dividing them into training and external validation cohorts. The primary outcome of this study was the Myasthenia Gravis Foundation of America (MGFA) post-intervention status assessed after 6 months of treatment initiation. Prognoses were classified as "unchanged or worse" for a poor outcome and "improved or better" for a good outcome. Accordingly, patients were categorized into "good outcome" or "poor outcome" groups. In the training cohort, we developed and internally validated various ML models using systemic inflammation indices, clinical indicators, or a combination of both. We then carried out external validation with the designated cohort. Additionally, we assessed the feature importance of our most effective model using the Shapley Additive Explanations (SHAP) method.

Results

In our study of 202 patients, 28.7% (58 individuals) experienced poor outcomes after 6 months of standard therapy. We identified 11 significant predictors, encompassing both systemic inflammation indexes and clinical metrics. The extreme gradient boosting (XGBoost) model demonstrated the best performance, achieving an area under the receiver operating characteristic (ROC) curve (AUC) of 0.944. This was higher than that achieved by logistic regression (Logit) (AUC: 0.882), random forest (RF) (AUC: 0.917), support vector machines (SVM) (AUC: 0.872). Further refinement through SHAP analysis highlighted five critical determinants-two clinical indicators and three inflammation indexes-as crucial for assessing short-term prognosis in AChR-Ab+ GMG patients.

Conclusion

Our analysis confirms that the XGBoost model, integrating clinical indicators with systemic inflammation indexes, effectively predicts short-term prognosis in AChR-Ab+ GMG patients. This approach enhances clinical decision-making and improves patient outcomes.

Advances and prospects of multi-modal ophthalmic artificial intelligence based on deep learning: a review

BACKGROUND

In recent years, ophthalmology has emerged as a new frontier in medical artificial intelligence (AI) with multi-modal AI in ophthalmology garnering significant attention across interdisciplinary research. This integration of various types and data models holds paramount importance as it enables the provision of detailed and precise information for diagnosing eye and vision diseases. By leveraging multi-modal ophthalmology AI techniques, clinicians can enhance the accuracy and efficiency of diagnoses, and thus reduce the risks associated with misdiagnosis and oversight while also enabling more precise management of eye and vision health. However, the widespread adoption of multi-modal ophthalmology poses significant challenges.

MAIN TEXT

In this review, we first summarize comprehensively the concept of modalities in the field of ophthalmology, the forms of fusion between modalities, and the progress of multi-modal ophthalmic AI technology. Finally, we discuss the challenges of current multi-modal AI technology applications in ophthalmology and future feasible research directions.

CONCLUSION

In the field of ophthalmic AI, evidence suggests that when utilizing multi-modal data, deep learning-based multi-modal AI technology exhibits excellent diagnostic efficacy in assisting the diagnosis of various ophthalmic diseases. Particularly, in the current era marked by the proliferation of large-scale models, multi-modal techniques represent the most promising and advantageous solution for addressing the diagnosis of various ophthalmic diseases from a comprehensive perspective. However, it must be acknowledged that there are still numerous challenges associated with the application of multi-modal techniques in ophthalmic AI before they can be effectively employed in the clinical setting.

Enhancing machine learning models for total organic carbon prediction by integrating geospatial parameters in river watersheds

This study utilizes machine learning (ML) algorithms to develop a robust total organic carbon (TOC) prediction model for river waters in the Geumho River sub-basins, South Korea, considering both non-rain and rain events. The model incorporates geospatial parameters such as land use, slope, flow rate, and basic water quality metrics including biochemical oxygen demand (BOD), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), and suspended solids (SS). A key aspect of this research is examining how land use information enhances the model's predictive accuracy. We compared two ML algorithms-extreme gradient boosting (XGBoost) and deep neural networks (DNN)-with a traditional multiple linear regression (MLR) approach. XGBoost outperformed the others, achieving an R2 value between 0.61 and 0.68 in the test dataset and demonstrating significant improvement during rain events with an R2 of 0.77 when including land use data. In contrast, this enhancement was not observed with the MLR model. Feature importance analysis using Shapley values highlighted COD as the primary predictor for non-rain events, while during rain events, COD, TP, TN, SS and agricultural land collectively influenced TOC levels. This study significantly advances understanding of TOC variability across different land use scenarios in river systems and underscores the importance of integrating geospatial and water quality parameters to enhance TOC prediction, particularly during rain events. This methodology provides a valuable framework for developing river management strategies and monitoring long-term TOC trends, especially in scenarios with gaps in essential monitoring data.

Construction of metal interpretable scoring system and identification of tungsten as a novel risk factor in COPD

Numerous studies have highlighted the correlation between metal intake and deteriorated pulmonary function, emphasizing its pivotal role in the progression of Chronic Obstructive Pulmonary Disease (COPD). However, the efficacy of traditional models is often compromised due to overfitting and high bias in datasets with low-level exposure, rendering them ineffective in delineating the contemporary risk trends associated with pulmonary diseases. To address these limitations, we embarked on developing advanced, interpretable models, crucial for elucidating the intricate mechanisms of metal toxicity and enriching the domain knowledge embedded in toxicity models. In this endeavor, we scrutinized extensive, long-term metal exposure datasets from NHANES to explore the interplay between metal and pulmonary functionality. Employing a variety of machine-learning approaches, we opted for the "Mixer of Experts" model for its proficiency in identifying a myriad of toxicological trends and sensitivities. We conceptualized and illustrated the TSAP (Toxicity Score at Population-level), a metal interpretable scoring system offering performance nearly equivalent to the amalgamation of standard interpretable methods addressing the "black box" conundrum. This streamlined, bifurcated procedural analysis proved instrumental in discerning established risk factors, thereby uncovering Tungsten as a novel contributor to COPD risk. SYNOPSIS: TSAP achieved satisfied performance with transparent interpretability, suggesting tungsten intake need further action for COPD prevention.