THE AUTOMATIC ANALYSIS OF BRAIN ELECTRICAL ACTIVITY AND ITS PREDICTION OF PARTIAL PRESSURE OF BLOOD CARBON DIOXIDE IN PREMATURE NEWBORN BABIES

Abstract

Currently the gold standard of monitoring blood carbon dioxide in premature babies is through blood sampling at regular time intervals. Close monitoring of the partial pressure of blood carbon dioxide (PCO2) in a newborn preterm baby is important. Hypo/hypercarbia has been associated with brain and lung damage. Continuous monitoring of PCO2 in preterm babies is not currently used in practice in part due to lack of a reliable technology. Software has been developed at University of Manchester for the automatic quantification of preterm EEG. Preterm EEG has been shown to change with changes in PCO2 with slowing of EEG at low levels of PCO2 and suppression of EEG at high levels of PCO2. EEG changes have also been demonstrated with blood pH. In this study, I aimed to:1. Develop linear regression equations for the prediction of PaCO2 and blood pH. 2. Determine the levels of agreement between PaCO2 and pH measured by blood gas analysis and the carbon dioxide and pH predicted by automatic analysis of EEG respectively.EEG was acquired for 36 hours starting from soon after birth by placing 7 hydrogel scalp electrodes. Babies born before 30 weeks' gestational age and with arterial catheters were included in the study. Arterial blood gases were measured using ABL 835 flex (Radiometer, Copenhagen, Denmark) and corrected for bias. EEG was quantified for interburst intervals and relative powers of delta, theta, alpha and beta EEG frequency bands using in-house developed software. Linear regression equations were developed using automatic linear modelling technique supplied by SPSS version 19. Bias and repeatability was calculated using Bland Altman method. Two linear regression equations for the prediction of PCO2 and one linear regression equation for the prediction of pH were developed using 46 blood gas measurements performed on 16 babies. The bias (mean difference: PaCO2 - PeegCO2) was 0.79 kPa and the precision (standard deviation of differences) was 0.71 kPa using forward stepwise method. The bias (mean difference: blood pH - EEG pH) was 0.02 kPa and the precision was 0.047 kPa using best subset method. PeegCO2 monitoring is a new development in the field of continuous monitoring in neonatal intensive care. Bias and precision data indicate that it can be a valuable clinical tool for monitoring trends of PCO2. However, further work is needed in development of artefact rejection and validation before introducing it into clinical practice.

Date of Award	1 Aug 2015
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Peter Clayton (Supervisor), Patrick Gaydecki (Supervisor) & Suresh Victor (Supervisor)