Predicting SARS-CoV-2 infection among hemodialysis patients using deep neural network methods DOI Creative Commons

Lihao Xiao,

Hanjie Zhang,

Juntao Duan

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Окт. 9, 2024

COVID-19 has a higher rate of morbidity and mortality among dialysis patients than the general population. Identifying infected early with support predictive models helps centers implement concerted procedures (e.g., temperature screenings, universal masking, isolation treatments) to control spread SARS-CoV-2 mitigate outbreaks. We collect data from multiple sources, including demographics, clinical, treatment, laboratory, vaccination, socioeconomic status, surveillance. Previous prediction models, such as logistic regression, SVM, XGBoost, require sophisticated feature engineering need improved performance. create deep learning Recurrent Neural Networks (RNN) Convolutional (CNN), predict infections during incubation. Our study shows minimal can identify those more accurately previously built models. Long Short-Term Memory (LSTM) model consistently performed well, an AUC exceeding 0.80, peaking at 0.91 in August 2021. The CNN also demonstrated strong results above 0.75. Both outperformed previous best XGBoost by over 0.10 AUC. Prediction accuracy declined pandemic evolved, dropping approximately 0.75 between September 2021 January 2022. Maintaining 20% false positive rate, our LSTM identified 66% 64% cases patients, significantly outperforming 42%. key features for calculating gradient output respect input features. By closely monitoring these factors, receive earlier diagnoses care, leading less severe outcomes. research highlights effectiveness neural networks analyzing longitudinal data, especially predicting crucial incubation period. These network approaches surpass traditional methods relying on aggregated variable means, improving accurate identification infections.

Язык: Английский

The Predictive Value of Soluble Fms-Like Tyrosine Kinase-1 for Prognosis in COVID-19 Patients DOI Creative Commons

CKN Lai,

Yingfei Wang,

Flair Donglai Shi

и другие.

Journal of Inflammation Research, Год журнала: 2025, Номер Volume 18, С. 3511 - 3522

Опубликована: Март 1, 2025

Background: Coronavirus Disease 2019 (COVID-19), caused by the novel coronavirus, has posed a significant threat to global public health, leading substantial morbidity, mortality, and strain on healthcare resources. Despite availability of vaccines treatments, effective biomarkers for predicting disease progression remain limited. This study aimed investigate prognostic value soluble fms-like tyrosine kinase-1 (sFlt-1) in COVID-19 patients. Methods: A prospective cohort was conducted involving 154 patients, with comprehensive clinical data laboratory parameters analyzed evaluate effectiveness sFlt-1 determining severity prognosis. Results: The results revealed that levels correlated significantly severity, showing higher severe/critical cases compared mild (P< 0.05). In deceased group, were notably survivors, an area under curve (AUC) 0.840, good predictive power 28-day mortality. Multivariable logistic regression identified sFlt-1, respiratory rate, albumin as independent factors, combined AUC 0.938 (95% CI: 0.886– 0.991) mortality risk. Conclusion: These findings underscore potential valuable biomarker decision-making managing Future studies should focus application explore its underlying mechanisms enhance patient management strategies. Keywords: COVID-19,

Язык: Английский

Процитировано

0
A machine learning-based severity stratification tool for high altitude pulmonary edema DOI Creative Commons

Luobu Gesang,

Yangzong Suona,

Zhuoga Danzeng

и другие.

BMC Medical Informatics and Decision Making, Год журнала: 2025, Номер 25(1)

Опубликована: Апрель 18, 2025

Язык: Английский

Процитировано

0
A machine learning model for predicting severe mycoplasma pneumoniae pneumonia in school-aged children DOI Creative Commons
Yingying Ye, Zhenpeng Gao, Zhiling Zhang

и другие.

BMC Infectious Diseases, Год журнала: 2025, Номер 25(1)

Опубликована: Апрель 21, 2025

To develop an interpretable machine learning (ML) model for predicting severe Mycoplasma pneumoniae pneumonia (SMPP) in order to provide reliable factors the clinical type of disease. We collected data from 483 school-aged children with M. (MPP) who were hospitalized at Children's Hospital Soochow University between September 2021 and June 2024. Difference analysis univariate logistic regression employed identify predictors training features ML. Eight ML algorithms used build models based on selected features, their effectiveness was validated. The area under curve (AUC), accuracy, five-fold cross-validation, decision (DCA) utilized evaluate performance. Finally, best-performing selected, Shapley Additive Explanations (SHAP) method applied rank importance interpret final model. After feature selection, 30 variables remained. constructed eight assessed effectiveness, finding that CatBoost exhibited best predictive performance, AUC 0.934 accuracy 0.9175. DCA compare benefits models, revealing provided greater net than other within threshold probability range 34% 75%. Additionally, we SHAP model, diagram visually show influence predictor outcome. results identified top six risk as number days fever, D-dimer, platelet count (PLT), C-reactive protein (CRP), lactate dehydrogenase (LDH), neutrophil-to-lymphocyte ratio (NLR). can help physicians accurately SMPP. This early identification facilitates better treatment options timely prevention complications. Furthermore, algorithm enhances model's transparency increases its trustworthiness practical applications.

Язык: Английский

Процитировано

0
sTREM-1 as a Predictive Biomarker for Disease Severity and Prognosis in COVID-19 Patients DOI Creative Commons
Nan Geng, Zhipeng Wu, Zhao Liu

и другие.

Journal of Inflammation Research, Год журнала: 2024, Номер Volume 17, С. 3879 - 3891

Опубликована: Июнь 1, 2024

Background: Research on biomarkers associated with the severity and adverse prognosis of COVID-19 can be beneficial for improving patient outcomes. However, there is limited research role soluble TREM-1 (sTREM-1) in predicting patients. Methods: A total 115 patients admitted to emergency department Beijing Youan Hospital from February May 2023 were included study. Demographic information, laboratory measurements, blood samples sTREM-1 levels collected upon admission. Results: Our study found that plasma increased disease (moderate vs mild, p=0.0013; severe moderate, p=0.0195). had good predictive value 28-day mortality (area under ROC curve was 0.762 0.805, respectively). also exhibited significant correlations age, body temperature, respiratory rate, PaO 2 /FiO , PCT, CRP, CAR. Ultimately, through multivariate logistic regression analysis, we determined (OR 1.008, 95% CI: 1.002– 1.013, p=0.005), HGB 0.966, 0.935– 0.998, p=0.036), D-dimer 1.001, 1.000– p=0.009), CAR 1.761, 1.154– 2.688, p=0.009) independent predictors The combination these four markers yielded a strong cases an AUC 0.919 (95% 0.857 − 0.981). Conclusion: demonstrated mortality, serving as prognostic factor In future, anticipate conducting large-scale multicenter studies validate our findings. Keywords: COVID-19, sTREM-1, inflammation-related markers, severity,

Язык: Английский

Процитировано

3
Study on the predictive value of laboratory inflammatory markers and blood count-derived inflammatory markers for disease severity and prognosis in COVID-19 patients: a study conducted at a university-affiliated infectious disease hospital DOI Creative Commons
Zhipeng Wu, Yu Cao, Zhao Liu

и другие.

Annals of Medicine, Год журнала: 2024, Номер 56(1)

Опубликована: Окт. 24, 2024

Background Since the outbreak of coronavirus disease 2019 (COVID-19), studies have found correlations between blood cell count-derived inflammatory markers (BCDIMs) and severity prognosis in COVID-19 patients. However, there is currently a lack systematic comparisons procalcitonin (PCT), C-reactive protein (CRP), protein-to-albumin ratio (CAR) BCDIMs for assessing

Язык: Английский

Процитировано

3
A machine learning model for predicting severe mycoplasma pneumoniae pneumonia in School-Aged children DOI

yingying ye,

Zhaoyao Gao, Z. Zhang

и другие.

Research Square (Research Square), Год журнала: 2025, Номер unknown

Опубликована: Апрель 4, 2025

Abstract Objective To develop an interpretable machine learning (ML) model for predicting severe Mycoplasma pneumoniae pneumonia (SMPP) in order to provide reliable factors the clinical type of disease. Methods We collected data from 483 school-aged children with M. (MPP) who were hospitalized at Children's Hospital Soochow University between September 2021 and June 2024. Difference analysis univariate logistic regression employed identify predictors training features ML. Eight ML algorithms used build models based on selected features, their effectiveness was validated. The area under curve (AUC), accuracy, five-fold cross-validation, decision (DCA) utilized evaluate performance. Finally, best-performing selected, Shapley Additive Explanations (SHAP) method applied rank importance interpret final model. Results After feature selection, 30 variables remained. constructed eight assessed effectiveness, finding that CatBoost exhibited best predictive performance, AUC 0.934 accuracy 0.9175. DCA compare benefits models, revealing provided greater net than other within threshold probability range 34–75%. Additionally, we SHAP model, diagram visually show influence predictor outcome. results identified top six risk as number days fever, D-dimer, platelet count (PLT), C-reactive protein (CRP), lactate dehydrogenase (LDH), neutrophil-to-lymphocyte ratio (NLR). Conclusions can help physicians accurately SMPP. This early identification facilitates better treatment options timely prevention complications. Furthermore, algorithm enhances model's transparency increases its trustworthiness practical applications.

Язык: Английский

Процитировано

0
Predicting SARS-CoV-2 infection among hemodialysis patients using deep neural network methods DOI Creative Commons

Lihao Xiao,

Hanjie Zhang,

Juntao Duan

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Окт. 9, 2024

COVID-19 has a higher rate of morbidity and mortality among dialysis patients than the general population. Identifying infected early with support predictive models helps centers implement concerted procedures (e.g., temperature screenings, universal masking, isolation treatments) to control spread SARS-CoV-2 mitigate outbreaks. We collect data from multiple sources, including demographics, clinical, treatment, laboratory, vaccination, socioeconomic status, surveillance. Previous prediction models, such as logistic regression, SVM, XGBoost, require sophisticated feature engineering need improved performance. create deep learning Recurrent Neural Networks (RNN) Convolutional (CNN), predict infections during incubation. Our study shows minimal can identify those more accurately previously built models. Long Short-Term Memory (LSTM) model consistently performed well, an AUC exceeding 0.80, peaking at 0.91 in August 2021. The CNN also demonstrated strong results above 0.75. Both outperformed previous best XGBoost by over 0.10 AUC. Prediction accuracy declined pandemic evolved, dropping approximately 0.75 between September 2021 January 2022. Maintaining 20% false positive rate, our LSTM identified 66% 64% cases patients, significantly outperforming 42%. key features for calculating gradient output respect input features. By closely monitoring these factors, receive earlier diagnoses care, leading less severe outcomes. research highlights effectiveness neural networks analyzing longitudinal data, especially predicting crucial incubation period. These network approaches surpass traditional methods relying on aggregated variable means, improving accurate identification infections.

Язык: Английский

Процитировано

1