Related Experiment Video
Updated: May 16, 2026
07:27
How to Administer Near-Infrared Spectroscopy in Critically ill Neonates, Infants, and Children
Published on: August 19, 2020
Applying data mining techniques to improve diagnosis in neonatal jaundice.
Duarte Ferreira1, Abílio Oliveira, Alberto Freitas
1Centro Hospitalar Tâmega e Sousa, EPE, Lugar do Tapadinho, Penafiel, 4564-007, Portugal. duarte83@gmail.com
BMC Medical Informatics and Decision Making
|December 11, 2012
Summary
Data mining techniques accurately predict neonatal jaundice, a common newborn condition. This approach enhances diagnostic capabilities, preventing potential neurological complications in infants.
Area of Science:
- Neonatal Medicine
- Data Science
- Medical Informatics
Background:
- Neonatal hyperbilirubinemia (jaundice) is increasingly prevalent due to shorter hospital stays.
- While often benign, poorly evaluated jaundice can cause severe neurological damage.
- Data mining offers improved diagnostic methodologies in various medical fields.
Purpose of the Study:
- To enhance the diagnostic accuracy of neonatal jaundice.
- To apply data mining techniques for predicting hyperbilirubinemia in newborns.
Main Methods:
- An observational study involving 227 healthy newborns (≥35 weeks gestation).
- Collected over 70 variables, including transcutaneous bilirubin levels, using a noninvasive bilirubinometer.
- Trained and tested classification models (decision trees, neural networks) using Weka software.
Main Results:
- Data mining algorithms achieved high accuracy in predicting subsequent hyperbilirubinemia.
- At 24 hours of life, prediction accuracy reached 89%.
- Naive Bayes, multilayer perceptron, and simple logistic algorithms yielded the best results.
Conclusions:
- Data mining approaches can effectively support medical decision-making.
- These techniques offer a promising method for improving neonatal jaundice diagnosis.
- Early and accurate diagnosis can prevent severe neurological consequences.
