Application of signal processing techniques in the assessment of clinical risks in preterm infants

TL;DR

This paper explores advanced signal processing techniques applied to physiological data to improve clinical risk assessment and early detection of serious conditions like IVH in preterm infants.

Contribution

It introduces spectral analysis, cross-spectral transfer function analysis, and fractal analysis of physiological signals, along with multivariate regression models, to identify potential clinical markers for adverse outcomes.

Findings

Spectral measures correlate with clinical risk index.

Cross-spectral analysis distinguishes infants with IVH.

Multivariate regression improves CO estimation accuracy.

Abstract

Preterm infants with very low birth weight suffer from a high risk of intra-ventricular hemorrhage(IVH) and other serious diseases. To improve the clinical risk assessment of preterm infants and develop potential clinically makers for the adverse outcome, the first part of the paper develops the frequency spectral analysis on the non-invasively measured heart rate variability, blood pressure variability and cerebral near-infrared spectroscopy measures. Moderate and high correlations with the clinical risk index for babies were identified from various spectral measures of arterial baroreflex and cerebral autoregulation functions. It was also observed that the cross-spectral transfer function analysis of cerebral NIRS and arterial blood pressure was able to provide a number of parameters that were potentially useful for distinguishing between preterm infants with or without IVH. Furthermore, the detrended fluctuation analysis that quantifies the fractal correlation properties of physiological signals has been examined, to determine whether it could derive markers for the identification of preterm infants with IVH. Cardiac output(CO) and total peripheral resistance(TPR) are two important parameters of the cardiovascular system. Measurement of these two parameters can provide valuable information for the assessment and management of patients needing intensive care, including preterm infants in the neonatal intensive care unit. To further assess the changes in CO and TPR in the preterm infants, the multivariate regression model based on the useful features from ABP method was used to improve the accuracy and robustness of the estimation. The combination of signal analysis and multivariate regression model in estimation of CO has produced some outcomes, and in the future, more effort should be involved in this kind of research to improve the prediction of serious diseases in preterm infants.