|
Predicting neonatal respiratory distress syndrome (RDS) using machine learning
|
Fatemeh Khabbazi niasari    , Aliyeh Mohamadzadeh * , Mahdi Esmaeili , Alireza Amirshahi |
| Department of Pediatrics, Faculty of Medicine, Kashan University of Medical Sciences, Kashan, I.R. Iran. , aliyehmo@gmail.com |
|
|
Abstract: (7 Views) |
Introduction: Neonatal respiratory distress syndrome (RDS) is one of the most common causes of mortality in preterm infants and is primarily associated with pulmonary surfactant deficiency. Risk stratification of RDS is important for improving clinical outcomes and reducing mortality. Conventional diagnostic approaches, including clinical examination and chest radiography, may have limitations in the early assessment of affected neonates.
Method: This study aimed to develop a machine learning-based model for risk stratification of neonatal respiratory distress syndrome. In this retrospective study, data were collected from 300 infants admitted to the neonatal intensive care unit of Shahid Beheshti Hospital in Kashan between 2023 and 2024. Data were initially extracted for 23 clinical variables. However, respiratory support and surfactant administration were excluded from the predictor set because of their treatment-related nature and potential risk of information leakage, leaving 21 variables for model development. After data preprocessing, four machine learning algorithms, including Decision Tree, ID3, k-nearest neighbors (KNN), and support vector machine (SVM), were implemented using RapidMiner. The dataset was stratified and divided into training (70%) and testing (30%) sets. Model performance was evaluated using accuracy, sensitivity, specificity, precision, and F1-score.
Results: The Decision Tree demonstrated the best overall performance, with an accuracy of 90.00%, sensitivity (recall) of 89.23%, specificity of 92.00%, precision of 96.67%, and F1-score of 92.80%. The area under the ROC curve (AUC) for the Decision Tree model was 0.868, indicating good discriminative performance.
Conclusion: The findings suggest that the decision tree model, using non-treatment-related clinical variables, may have potential as an auxiliary risk-stratification tool for RDS. However, external and multicenter validation is required before clinical implementation.
|
|
| Keywords: Respiratory Distress Syndrome, Newborn, Machine Learning, Decision Trees, Neonatal Intensive Care Units, Support Vector Machines |
|
|
|
Type of Study: Research |
Subject:
General
Received: 2026/07/2 | Revised: 2026/10/14 | Accepted: 2026/10/14
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
_-_Copy.png)
.png){kind=logo}
