Explainable machine learning model for predicting decline in platelet count after interventional closure in children with patent ductus arteriosus DOI Creative Commons

Songyue Zhang,

Yidong Zhang, Hao Li

et al.

Frontiers in Pediatrics, Journal Year: 2025, Volume and Issue: 13

Published: Feb. 6, 2025

This study aimed to apply four machine learning algorithms develop the optimal model predict decline in platelet count (DPC) after interventional closure children with patent ductus arteriosus (PDA). Data from PDA who underwent successful transcatheter at Second Affiliated Hospital of Wenzhou Medical University and Yuying Children's January 2016, December 2022, were collected. The cohort data split into training testing sets. DPC following intervention is defined as a percentage ≥25% [(baseline count-nadir count)/baseline count]. extra tree algorithm was used for feature selection ML [random forest (RF), adaptive boosting, extreme gradient logistic regression] established. Moreover, SHapley Additive exPlanation (SHAP) explain importance features models. included 330 PDA, which 113 (34.2%) experienced DPC. After 62 clinical considered, selected six build Amongst algorithms, RF achieved greatest AUC. SHAP analysis revealed that pulmonary artery systolic pressure, size defect weight top three most important model. Furthermore, descriptions two accurate predictions, explanations prediction results provided. In this study, an (RF) capable predicting post-intervention undergoing

Language: Английский

Explainable machine learning model for predicting decline in platelet count after interventional closure in children with patent ductus arteriosus DOI Creative Commons

Songyue Zhang,

Yidong Zhang, Hao Li

et al.

Frontiers in Pediatrics, Journal Year: 2025, Volume and Issue: 13

Published: Feb. 6, 2025

This study aimed to apply four machine learning algorithms develop the optimal model predict decline in platelet count (DPC) after interventional closure children with patent ductus arteriosus (PDA). Data from PDA who underwent successful transcatheter at Second Affiliated Hospital of Wenzhou Medical University and Yuying Children's January 2016, December 2022, were collected. The cohort data split into training testing sets. DPC following intervention is defined as a percentage ≥25% [(baseline count-nadir count)/baseline count]. extra tree algorithm was used for feature selection ML [random forest (RF), adaptive boosting, extreme gradient logistic regression] established. Moreover, SHapley Additive exPlanation (SHAP) explain importance features models. included 330 PDA, which 113 (34.2%) experienced DPC. After 62 clinical considered, selected six build Amongst algorithms, RF achieved greatest AUC. SHAP analysis revealed that pulmonary artery systolic pressure, size defect weight top three most important model. Furthermore, descriptions two accurate predictions, explanations prediction results provided. In this study, an (RF) capable predicting post-intervention undergoing

Language: Английский

Citations

0