A neural network approach to predict early neonatal sepsis. (June 2019)

• Record Type: Journal Article
• Authors: López-Martínez, Fernando; Núñez-Valdez, Edward Rolando; Lorduy Gomez, Jaime; García-Díaz, Vicente
• Is Part Of: Computers & electrical engineering. Volume 76(2019)
• Page Start: 379
• Page End: 388
• Publication Date: 2019-06
• Subjects: Machine learning -- Artificial neural networks -- Sepsis neonatal -- Medical decision support systems -- Smart systems; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2019.04.015
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: Best calculated AUC of 92.5% (95% CI [91.4–93.06]) showing fair result with the final diagnosis. This model predicts correctly positive cases of neonatal sepsis 92.5% better than a randomly selected individual. This study showed better results than previous studies performed by the same authors using logistic regression with the same input variables that showed a calculated AUC of 87%. The miss rate and the fall out rate were not significant for our model due to the imbalance of the sampling data set. The use of a multi-layer artificial neural network help to overcome the non-linearity problem with the risk factor variables used in the study. Abstract: The purpose of this study is to develop a non-invasive neural network classification model for early neonatal sepsis detection. Early neonatal sepsis is a public health issue and one of the leading causes of complications and deaths in neonatal intensive care units. The data used in this study is from Crecer's Hospital center in Cartagena-Colombia. An imbalanced dataset of 555 neonates with (66%) of negative cases and (34%) of positive cases was used for this study. The study results show a sensitivity of 80.32%, a specificity of 90.4%, precision on the positive predicted value of 83.1% in the test sample and a calculated area under the curve of 92.5% (95% Confidence Interval [91.4-93.06]). This neural network model can be used as a smart system's inference engine to support the detection of neonatal sepsis in neonatal intensive care units.