Lightweight Privacy-preserving Medical Diagnosis in Edge Computing

ObjectivesA radiomics-based explainable eXtreme Gradient Boosting (XGBoost) model was developed to predict central cervical lymph node metastasis (CCLNM) in patients with papillary thyroid carcinoma (PTC), including positive and negative effects.MethodsA total of 587 PTC patients admitted at Binzhou Medical University Hospital from 2017 to 2021 were analyzed retrospectively. The patients were randomized into the training and test cohorts with an 8:2 ratio. Radiomics features were extracted from ultrasound images of the primary PTC lesions. The minimum redundancy maximum relevance algorithm and the least absolute shrinkage and selection operator regression were used to select CCLNM positively-related features and radiomics scores were constructed. Clinical features, ultrasound features, and radiomics score were screened out by the Boruta algorithm, and the XGBoost model was constructed from these characteristics. SHapley Additive exPlanations (SHAP) was used for individualized and visualized interpretation. SHAP addressed the cognitive opacity of machine learning models.ResultsEleven radiomics features were used to calculate the radiomics score. Five critical elements were used to build the XGBoost model: capsular invasion, radiomics score, diameter, age, and calcification. The area under the curve was 91.53% and 90.88% in the training and test cohorts, respectively. SHAP plots showed the influence of each parameter on the XGBoost model, including positive (i.e., capsular invasion, radiomics score, diameter, and calcification) and negative (i.e., age) impacts. The XGBoost model outperformed the radiologist, increasing the AUC by 44%.ConclusionsThe radiomics-based XGBoost model predicted CCLNM in PTC patients. Visual interpretation using SHAP made the model an effective tool for preoperative guidance of clinical procedures, including positive and negative impacts.

ObjectiveTo compare an Extreme Gradient Boosting (XGboost) model with a multivariable logistic regression (LR) model for their ability to predict sepsis after extremely severe burns.MethodsFor this observational study, patient demographic and clinical information were collected from medical records. The two models were evaluated using area under curve (AUC) of the receiver operating characteristic (ROC) curve.ResultsOf the 103 eligible patients with extremely severe burns, 20 (19%) were in the sepsis group, and 83 (81%) in the non-sepsis group. The LR model showed that age, admission time, body index (BI), fibrinogen, and neutrophil to lymphocyte ratio (NLR) were risk factors for sepsis. Comparing AUC of the ROC curves, the XGboost model had a higher predictive performance (0.91) than the LR model (0.88). The SHAP visualization tool indicated fibrinogen, NLR, BI, and age were important features of sepsis in patients with extremely severe burns.ConclusionsThe XGboost model was superior to the LR model in predictive efficacy. Results suggest that, fibrinogen, NLR, BI, and age were correlated with sepsis after extremely severe burns.

Background This study aimed to establish and validate a machine learning (ML) model for predicting in-hospital mortality in critically ill patients with chronic kidney disease (CKD). Methods This study collected data on CKD patients from 2008 to 2019 using the Medical Information Mart for Intensive Care IV. Six ML approaches were used to build the model. Accuracy and area under the curve (AUC) were used to choose the best model. In addition, the best model was interpreted using SHapley Additive exPlanations (SHAP) values. Results There were 8527 CKD patients eligible for participation; the median age was 75.1 (interquartile range: 65.0–83.5) years, and 61.7% (5259/8527) were male. We developed six ML models with clinical variables as input factors. Among the six models developed, the eXtreme Gradient Boosting (XGBoost) model had the highest AUC, at 0.860. According to the SHAP values, the sequential organ failure assessment score, urine output, respiratory rate, and simplified acute physiology score II were the four most influential variables in the XGBoost model. Conclusions In conclusion, we successfully developed and validated ML models for predicting mortality in critically ill patients with CKD. Among all ML models, the XGBoost model is the most effective ML model that can help clinicians accurately manage and implement early interventions, which may reduce mortality in critically ill CKD patients with a high risk of death.

The diagnosis of schizophrenia and bipolar disorder is often delayed several years despite illness typically emerging in late adolescence or early adulthood, which impedes initiation of targeted treatment. To investigate whether machine learning models trained on routine clinical data from electronic health records (EHRs) can predict diagnostic progression to schizophrenia or bipolar disorder among patients undergoing treatment in psychiatric services for other mental illness. This cohort study was based on data from EHRs from the Psychiatric Services of the Central Denmark Region. All patients aged 15 to 60 years with at least 2 contacts (at least 3 months apart) with the Psychiatric Services of the Central Denmark Region between January 1, 2013, and November 21, 2016, were included. Analysis occurred from December 2022 to November 2024. Predictors based on EHR data, including medications, diagnoses, and clinical notes. Diagnostic transition to schizophrenia or bipolar disorder within 5 years, predicted 1 day before outpatient contacts by means of elastic net regularized logistic regression and extreme gradient boosting (XGBoost) models. The area under the receiver operating characteristic curve (AUROC) was used to determine the best performing model. The study included 24 449 patients (median [Q1-Q3] age at time of prediction, 32.2 [24.2-42.5] years; 13 843 female [56.6%]) and 398 922 outpatient contacts. Transition to the first occurrence of either schizophrenia or bipolar disorder was predicted by the XGBoost model, with an AUROC of 0.70 (95% CI, 0.70-0.70) on the training set and 0.64 (95% CI, 0.63-0.65) on the test set, which consisted of 2 held-out hospital sites. At a predicted positive rate of 4%, the XGBoost model had a sensitivity of 9.3%, a specificity of 96.3%, and a positive predictive value (PPV) of 13.0%. Predicting schizophrenia separately yielded better performance (AUROC, 0.80; 95% CI, 0.79-0.81; sensitivity, 19.4%; specificity, 96.3%; PPV, 10.8%) than was the case for bipolar disorder (AUROC, 0.62, 95% CI, 0.61-0.63; sensitivity, 9.9%; specificity, 96.2%; PPV, 8.4%). Clinical notes proved particularly informative for prediction. These findings suggest that it is possible to predict diagnostic transition to schizophrenia and bipolar disorder from routine clinical data extracted from EHRs, with schizophrenia being notably easier to predict than bipolar disorder.

Development and validation of interpretable machine learning model for pre-treatment predicting the response to targeted and immune therapy in hepatocellular carcinoma.

Glass forming ability prediction of bulk metallic glasses based on fused strategy

ObjectivesHuman brucellosis is a public health problem endangering health and property in China. Predicting the trend and the seasonality of human brucellosis is of great significance for its prevention. In...

Recently, edge computing has attracted significant interest due to its ability to extend cloud computing utilities and services to the network edge with low response times and communication costs. In general, edge computing requires mobile users to upload their raw data to a centralized data server for further processing. However, these data usually contain sensitive information about mobile users that the users do not want to reveal, such as sexual orientation, political stance, health status, and service access history. The transmission of user data increases the leakage risk of data privacy since many extra devices can get access to these data. In this article, we attempt to keep the data of edge devices and end users on their local storage to resist the leakage of user privacy. To this end, we integrate federated learning and edge computing to propose P2FEC, a privacy-preserving framework that can construct a unified deep learning model across multiple users or devices without uploading their data to a centralized server. Furthermore, we use membership inference attacks as a case study for the privacy analysis of edge computing. The experiments show that the model constructed by our framework can achieve similar prediction performance and stricter protection of data privacy, compared to the model trained by standard edge computing.

Based on the statistics, advertisements (ads) generate more than 80% of companies' revenues. However, the complexity of the ads ecosystem blurs the boundaries of responsibility between companies. It can not analyze privacy and security issues in such an ecosystem. For example, collecting user information without user consent for data analysis and displaying ads will cause privacy leakage. But we can not explain which types of the company need to take measures to protect privacy. In our paper, to clarify the responsibility of companies in the advertisement ecosystem, we divide them into six types of entities according to their needs and functions. Then, we design a lightweight simulation platform to illustrate the advertisement ecosystem and support security and privacy analysis. Finally, we take personal ads recommendations based on federated learning as an example to verify the feasibility for privacy and security analysis in this platform.

Sepsis-related mortality rates are high among elderly patients, especially those in intensive care units (ICUs). Early prediction of the prognosis of sepsis is critical, as prompt and effective treatment can improve prognosis. Researchers have predicted mortality and the development of sepsis using machine learning algorithms; however, few studies specifically focus on elderly patients with sepsis. This paper proposes a viable model for early prediction of in-hospital mortality among elderly patients diagnosed with sepsis. We extracted patient information from the Medical Information Mart for Intensive Care IV database. We employed several machine learning algorithms to predict the in-hospital mortality of elderly ICU patients with sepsis. The performance of the model was evaluated by using the AUROC and F1 score. Furthermore, the SHAP algorithm was used to explain the model, analyze how the individual features affect the model output, and visualize the Shapley value for a single individual. Our study included 18522 elderly patients, with a mortality of 15.4%. After screening, 59 clinical variables were extracted to develop models. Feature importance analysis showed that age, PO2, RDW, SPO2, WBC, and urine output were significantly related to the in-hospital mortality. According to the results of AUROC (0.871 (95% CI: 0.854–0.888)) and F1 score (0.547 (95% CI: 0.539–0.661)) analyses, the extreme gradient boosting (XGBoost) model outperformed the other models (i.e., LGBM, LR, RF, DT, and KNN). Furthermore, SHAP force analysis illustrated how the constructed model visualized the individualized prediction of death. XGBoost machine learning framework gives good in-hospital mortality prediction of elderly patients with sepsis and can maximize prediction model accuracy. The XGBoost model could be an effective tool to assist doctors in identifying high-risk cases of in-hospital mortality among elderly patients with sepsis. This could be used to create a clinical decision support system in the future.

BackgroundPrior research has demonstrated that nutrition plays a crucial role in the establishment and maturation of the reproductive lifetime. Although the specific dietary components involved in preventing or postponing the reproductive lifespan are still unknown, a healthy diet can affect the reproductive lifespan. Here, the study aimed to explore the relationship between reproductive lifespan and diet quality by utilizing the Healthy Eating Index-2015 (HEI-2015).MethodsIn this study, a total of 2761 postmenopausal women were selected from the National Health and Nutrition Examination Survey (NHANES) from 2005 to 2016. Diet quality was determined using HEI-2015 based on two 24-hour dietary recalls. Reproductive lifespan was defined as the number of years between self-reported age at menarche and menopause. Weighted linear regression and eXtreme Gradient Boosting (XGBoost) models were used to analyze the relationship between HEI-2015 and reproductive lifespan. Subsequently, the impact of various components of HEI-2015 on reproductive lifespan was assessed through weighted quantile sum (WQS) regression models.ResultsAmong 2761 postmenopausal women, the mean age was 63.7 years. 41.5% were obese, and 49.7% were non-Hispanic white. After adjusting for sociodemographic characteristics, lifestyle factors, and medical history, individuals in the highest tertile of HEI-2015 had a 4.81% (95% CI: 1.82–7.79%) longer reproductive time life. Higher HEI-2015 was also significantly associated with a higher likelihood of late menopause (p for trend < 0.05). Based on XGBoost models, the relative importance of HEI-2015 on reproductive lifespan was determined. Whole fruits, whole grains, total protein foods, and greens and beans significantly contributed to extending age at menopause and reproductive time life in the HEI-2015. The weights of the WQS index for age at menopause were 27.1%, 23.2%, 10.1%, and 7.5% respectively, while the weights of the WQS index for reproductive time life were 30.2%, 14.6%, 9.3%, and 14.0% respectively.ConclusionThere is a positive association between the HEI-2015 and reproductive lifespan. This underscores the significance of enhancing adherence to healthy dietary patterns in preventing a shorter reproductive lifespan.

In the monitoring and analysis of physiological and biochemical indicators of athletes, traditional data mining (DM) technology cannot extract compelling features and laws when processing high-dimensional and complex multivariate data, and the accuracy of the analysis results is low. The lack of real-time monitoring of the dynamically changing physiological state makes it impossible to detect athletes’ overtraining or fatigue in time, which affects the training effect and the health of athletes. This paper constructs an improved XGBoost (eXtreme Gradient Boosting) model to clean and normalize the collected physiological and biochemical data, remove outliers and fill in missing values, and construct a variable set representing the characteristics of different training periods to provide high-quality input data for subsequent model analysis. This paper combines the SHAP (SHapley Additive exPlanations) method to quantify the importance of each feature, selects the variables that contribute most to the recognition of the training state to optimize the model input, reduce the model complexity, and improve the computational efficiency. Based on the original XGBoost model, the loss function can be adjusted and the adaptive learning rate mechanism can be added to enable the model better to capture the dynamic changes of physiological and biochemical indicators and improve the prediction accuracy. Combined with the prediction results of the improved model, a real-time monitoring system was designed to track the changes in the physiological state of athletes during different training periods, and to issue an alarm when abnormal trends were detected to assist coaches in adjusting training plans. The experimental results show that in the feature evaluation, three key physiological indicators, namely blood oxygen saturation, blood lactate concentration, and heart rate, are extracted, which reduces the computational complexity of the subsequent model. In the four training stages of the basic period, load period, high-intensity period and recovery period, the loss values of the XGBoost model were approximately 0.5, 0.42, 0.4 and 0.35 respectively. In the monitoring data of 4 batches of football players, with 100 players in each batch, the accuracy rate remained above 0.83 and the response time was below 2 s. The experiment proved the effectiveness of the research model in the monitoring and analysis of physiological and biochemical indicators.

A new FCM-XGBoost system for predicting Pavement Condition Index

This study aims at the financial decision making problems faced by enterprises in the process of financing and capital operation, and uses the extreme gradient boosting (XGBoost) model to provide enterprises with decision support for the optimal financing channels. First, financial data from enterprises of different industries and sizes are collected, and key features related to financing and capital operation are selected according to the financial decision making needs of enterprises. Subsequently, the XGBoost model is used to predict the impact of different financing methods on corporate financial indicators (including capital structure, profitability, liquidity, debt repayment ability, etc.), and the best model parameters are selected through cross-validation and hyperparameter tuning.

River flow forecasting has been the focus of many researchers for many years.&amp;#160; The methods evolved from simple statistical methods to highly sophisticated mathematical models.&amp;#160; In recent years, due to the advancement of computers and artificial algorithms, new methods have become increasingly reliable and easier to use.&amp;#160; One of the promising artificial intelligence methods is the Extreme Gradient Boosting (XGBoost) model.&amp;#160; XGBoost is a scalable, distributed gradient-boosting decision tree machine learning library.&amp;#160; It provides parallel tree boosting and is the leading machine learning library for regression, classification, and ranking problems.&amp;#160; Three different algorithms of XGBoost were used in this research and the results were compared.&amp;#160; These algorithms were Random Search, Grid Search, and CatBoost. The proposed models were conducted in a station located P&amp;#242; River&amp;#160;basin which is the longest river in Italy, and it flows from the Cottian Alps and ends at a delta projecting into the Adriatic Sea new Venice.&amp;#160; The data were divided into training and validation sets.&amp;#160; The statistical indicators included mean square error, Nash-Sutcliffe efficiency, and mean absolute error were calculated for each set to compare the efficiency of each algorithm.&amp;#160; These indicators showed that XGBoost using random search algorithm had better performance, although the other algorithms were also acceptable predictions.&amp;#160; In general, the XGBoost model could be used as a reliable tool to forecast the river flow at locations with enough historical data.