Classification and feature extraction for neonatal sepsis detection

Abstract

<p> In this original study, we investigate the performances of machine learning algorithms on a neonatal sepsis detection task. We consider this work to be of great interest to both engineers and clinicians, as it uses non-invasive, already existing, vital signs monitoring signals in a population of very low birth weight infants at high risk of sepsis. Vital sign variability may indeed represent a general indicator of health and wellbeing and be helpful in the early detection of systematic inflammation such as sepsis. We used state of the art feature extraction technics and evaluate a large variety of binary classification models among which a neural network based generative model. The models were chosen from two main families: discriminative and generative. This enables a comprehensive study of different kinds of traditional and advanced binary classification algorithms.</p> <p> Our study reveals that advanced machine learning models are more robust to changes in the feature extraction pipeline, although linear classifiers have a comparable performance when the feature extraction is tuned. The advanced model performing the best is a neural network based generative model which is a hybrid generative and discriminative model. A large window length when computing the features is beneficial to almost all algorithms, indicating the relevance of frequency domain related features for the neonatal sepsis detection task.</p> <p> Overall we obtain a classification AUROC above 0.85, which makes our prediction models potentially relevant in clinical practice. This will enable earlier therapeutic interventions and thereby reduce morbidity and mortality in infants.</p>

Weight a minute: The smaller and more immature, the more predictable the autonomic regulation?

“…A detailed description of the method as well as the algorithm used is published. 13,14 In summary, in this longitudinal cohort study we prospectively collected high-frequency monitor data and performed retrospective annotation of clinical events. Inclusion criteria was admission to the NICUs of Karolinska University Hospitals in Solna and Huddinge, Sweden between February 2016, and June 2020.…”

“…A detailed description of the applied sepsis detection algorithm was published before. 13 In summary we split the high frequency monitoring data (IBI, RR, SpO 2 ) into windows of 45 min. Every window was marked as "1" if it was between 24 h before and 4 h after a sepsis event.…”

“…The described changes of HRV preceding major inflammatory events, such as sepsis, have been utilised in different prediction algorithms. 7,10,12,13 The fact that body weight and degree of prematurity are associated with input variables of these algorithms and simultaneously are independent risk factors for developing complications as sepsis summarises the complexity of the problem.…”

“…We therefore included characteristics of RR and SpO 2 in addition to HRC, resulting in improved algorithm performance. 13,14 In addition, other studies investigated the possibility to include these additional vital signs in the algorithm with promising results. 11,32 Other groups found that the cross-correlation between HR and SpO 2 was most valuable in detecting sepsis and necrotizing enterocolitis in infants.…”