Assessment of cortical and sub-cortical function in neonates by electrophysiological monitoring

The aim of this thesis was the assessment of cortical and sub-cortical function in neonates by electrophysiological monitoring, i.e. to evaluate the function of the neonatal cortex and brainstem through quantitative analysis of signals readily available in the NICU. These signals include amplitude-integrated EEG (aEEG), EEG, ECG and ABP. Important factors influencing neonatal cerebral function are the immaturity of preterm infants, and pathological conditions such as stroke, asphyxia and hemorrhage. Electrophysiological monitoring may be a valuable tool to monitor the development of cortical and sub-cortical function during the intensive care period. However, currently our knowledge of functional maturation and related signal changes is mostly based on visual pattern recognition. Since there are no exact reference values, maturation is difficult to assess and interpretations may vary between investigators. Digital monitoring recordings provide an opportunity towards quantitative signal analysis. This approach may lead to a more objective and uniform description of cortical and sub-cortical function. In the introductory Chapter 1, a brief overview of the mechanisms of perinatal encephalopathy is given, clinical diagnostic protocols are discussed and potential treatment strategies are highlighted. Furthermore, the origin and characteristics of the neonatal monitoring signals analyzed in this thesis are explained. The First part of this thesis focuses on traditional analysis of cortical function, quantifying maturational changes in multichannel aEEG and EEG recordings and quantifying the effects of anti-epileptic drugs on the EEG, to provide an objective and reproducible method for assessing brain maturation and long-term prognosis. In Chapter 2, multi-channel aEEG recordings were quantitatively analyzed and assessed for regional differences in maturation. We investigated 40 preterm infants (PMA 27–37 weeks) with normal follow-up at 24 months of age, at a median postnatal age of 8 days, using 4-h EEG recordings according to the international 10–20 system reduced montage. Nine channels were selected and converted to aEEG registrations. For each aEEG registration, lower margin amplitude (LMA), upper margin amplitude (UMA) and bandwidth (UMA–LMA) were calculated. In all channels, LMA showed a strong positive correlation with PMA. Below 32 weeks of PMA, LMA was =5 µV. Linear regression analysis showed a maximum LMA difference between channels of approximately 2 and 1 µV at 27 and 37 weeks of PMA, respectively. The lowest are LMA values in the occipital channel and the highest values are in centro-occipital channels. In the frontal, centro-temporal and centro-occipital channels, UMA and bandwidth changed with PMA. No differences in LMA, UMA and bandwidth were found between hemispheres. Skewness of LMA values strongly correlated with PMA, with positive skewness related to an immature brain (PMA=32 weeks) and negative skewness to a maturing (PMA>32 weeks) brain. In conclusion, we detected symmetric increase of aEEG characteristics, indicating symmetric brain maturation of the left and right hemispheres. In Chapter 3 we focused on automated power spectral analysis of the EEG in preterm infants, to identify maturation related changes of spectral measures. Weekly 4-hour EEG recordings (10–20 montage) were performed in 18 preterm infants with GA <32 weeks and normal neurological follow-up at 2 years of age, resulting in 79 recordings studied from 27+4 to 36+3 weeks PMA. Automated spectral analysis was performed on the 4-h EEG recordings. The frequency spectrum was divided in d1 (0.5-1 Hz), d2 (1-4 Hz), ¿ (4-8 Hz), a (8-13 Hz) and ß (13-30 Hz) band. Absolute and relative power of each frequency band and spectral edge frequency were calculated. Maturational changes in spectral measures were observed most clearly in the centro-temporal channels. With advancing PMA, absolute powers of d1, d2 and ¿ decreased. With advancing PMA, relative power of d1 decreased and relative powers of a and ß increased, respectively. In conclusion, with maturation, spectral analysis of the EEG showed a decrease in total power and a significant shift from the lower to the higher frequencies. In Chapter 4 we studied limitations of the EEG in a clinical environment. Assessment of the neonatal EEG may be hampered by drug-specific changes in electrocortical activity. To quantify effects of a loading dose of midazolam and lidocaine on the EEG frequency spectrum of full-term neonates with perinatal arterial ischemic stroke, 12 full-term infants with multi-channel aEEG and EEG recordings were retrospectively selected. During recording, midazolam and/or lidocaine were administered as anti-epileptic drug. We performed spectral analysis on 4-hour EEG segments around the loading dose. The frequency spectrum was divided in d (1-4 Hz), ¿ (4-8 Hz), a (8-13 Hz) and ß (13-30 Hz) bands. Midazolam induced immediate suppression of the aEEG background pattern for 30-60 minutes. Spectral EEG analysis showed decreased total and absolute frequency band powers. Relative d power decreased, ¿ power increased while a and ß powers remained constant. Lidocaine induced no aEEG background pattern suppression. Total and absolute EEG band powers were unchanged. Relative d power decreased, ¿ and a power increased and ß power remained constant. Effects were more pronounced in the stroke-affected hemisphere. In conclusion, both drugs induced a shift from low to higher frequency electrocortical activity. Additionally, midazolam reduced total EEG power. These spectral changes differ from those seen in adult studies. The Second part of this thesis focuses on novel methods for analysis of cortical function, i.e. quantifying changes in cortical function related to maturation and pathological condition through extraction and analysis of EEG transients, and through localization of the electrical sources of cortical activity. In Chapter 5, a method for automated detection of bursts, inter-burst-intervals (IBI), and continuous patterns in the EEG was developed in MATLAB® to aid with prognosis and stratification of clinical treatment for preterm infants. Results were evaluated for preterm infants with normal neurological follow-up at 2 years. The detection algorithm for burst, IBI and continuous pattern is based on selection by amplitude, time span, number of channels and numbers of active electrodes. Annotations of two neurophysiologists were used to determine threshold values. The training-set consisted of EEG recordings of 4 preterm infants with PMA of 29-34 weeks. Optimal threshold values were based on overall highest sensitivity. For evaluation, both observers verified detections in an independent dataset of 4 EEG recordings with comparable PMA. Algorithm performance was assessed by calculation of sensitivity and positive predictive value. Results of algorithm evaluation: sensitivity values of 90 ± 6, 80 ± 9 and 97 ± 5 % for burst, IBI and continuous patterns, respectively. Corresponding positive predictive values were 88 ± 8, 96 ± 3 and 85 ± 15 %, respectively. In conclusion, the algorithm showed high sensitivity and positive predictive values for bursts, IBI and continuous patterns in preterm EEG. In Chapter 6 we quantified the EEG burst frequency spectrum of preterm infants by automated analysis and described the topography of maturational change in spectral parameters. With this new method, we aim to increase sensitivity for maturational changes in cortical activity transients compared to non-specific analysis as presented in Chapters 2 and 3. Therefore, 18 preterm infants <32 weeks gestation and normal neurological follow-up at 2 years, underwent weekly 4-hour EEG recordings (10-20 system). The recordings (n=77) represent a large variability in PMA (28 - 36 weeks). We applied the automated burst detection algorithm presented in Chapter 5, and subsequently performed spectral analysis on burst activity. The frequency spectrum was divided in d1 (0.5-1 Hz), d2 (1-4 Hz), ¿ (4-8 Hz), a (8-13 Hz) and ß (13-30 Hz) band. Spectral parameters were evaluated as function of PMA by regression analysis. Results were interpolated and topographically visualized. The majority of spectral parameters show significant change with PMA. Highest correlation is found for d and ¿ band. Absolute band powers decrease with increasing PMA, while relative a and ß powers increase. Maturational change is largest in frontal and temporal region. In conclusion, the topographic distribution of maturational changes in spectral parameters corresponds with studies showing ongoing gyration and postnatal white matter maturation in frontal and temporal lobes. In Chapter 7 it was hypothesized that combining EEG with MRI could give more information about the development of possible brain damage. The study consisted of continuous EEG monitoring of neonates at risk of hypoxia-ischemia (HI), alternated by MRI examinations on predefined time points. Included were 8 term newborns with suspected neonatal encephalopathy caused by perinatal HI. The EEG was measured between day 1 and 8 after birth. MRI data was acquired between day 1 and 9, within 24 hrs from an observed EEG event. Preliminary results indicated an agreement between abnormalities in diffusion weighted MR images and EEG signal changes at corresponding electrode locations. In Chapter 8 we extended the preliminary results presented in Chapter 7 by exploring the feasibility of locating the sources of cortical activity as measured by scalp EEG in neonates. Three term neonates, representative examples of perinatal asphyxia and stroke, were selected for swLORETA analysis. A head model was created based on generic MRI data. Resulting 3-dimensional current source distributions showed good conformity with neonatal MRI images obtained within a 24-hour period from the EEG measurements. This visualization method may provide valuable information for longitudinal monitoring of neonates at risk, for guiding in-detail MRI examinations or for assessing treatment efficacy. The Third part of this thesis focuses on models for analysis of sub-cortical function, and describes modeling of the autonomic cardiovascular regulation of a preterm infant and assessment of the influence of asphyxia on autonomic cardiovascular regulation. In Chapter 9 we presented a model of the autonomic cardiovascular system of a preterm infant of 28 weeks gestation, with a birth weight of 1000 g and a closed ductus arteriosus by the end of the first week. This model describes the complex interactions between heart rate, blood pressure and respiration. The hemodynamic part of the model consists of a pulsatile heart and several vascular compartments, and is regulated by a baroreflex control system. The model is relatively simple to allow for a mathematical analysis of the dynamics, yet sufficiently complex to provide a realistic representation of the underlying physiology. Baroreflex related fluctuations were calculated by transfer function analysis between SBP and R-R fluctuations, using fast Fourier transform, estimated in the low frequency (LF, 0.04-0.15 Hz) band. LF transfer gain (baroreflex sensitivity, BRS, ms/mmHg) and phase (s) were assumed to reflect the baroreflex mediated cardiovascular fluctuations. The model provides (beat-to-beat) values of R-R interval and blood pressure that resemble realistic signals of preterm infants, and is validated with experimental heart rate and blood pressure data obtained in preterm infants. In Chapter 10 we monitored longitudinal changes of baroreflex activity in a fetal sheep model by analysis of the transfer function between low-frequency R-R and systolic blood pressure (SBP) fluctuations after umbilical cord occlusion (UCO) induced asphyxia. Fourteen preterm fetal lambs (UCO, n=5 and controls, n=9) were instrumented at 102 days gestation (term: 146 days) during caesarian section with a femoral arterial catheter, amniotic pressure catheter, 3-lead ECG and vascular occluder around the umbilical cord (UCO group). At 4 days after surgery, intra-uterine asphyxia was achieved by rapidly filling the vascular occluder for 25-min. Baroreflex related fluctuations were calculated by transfer function analysis. Before occlusion, no differences in R-R interval, SBP, and transfer gain and phase were observed between UCO and controls. Median BRS and delay were 7.8 (IQR, 5.6-11.4) and 4 (IQR, 3-6), and 6.2 (IQR, 5.6–8.4) ms/mmHg and 6 (IQR, 3-7) s, respectively. Three days after UCO, BRS gradually decreased in the UCO group, resulting in significantly lower values by the end of the week (UCO median 4.8, IQR 3.1-6.7 versus controls 7.7, IQR 6.7-10.6 ms/mmHg). In UCO group, delay increased to a median value of 7 (IQR, 4-8) s and differed from the controls. In conclusion, intra-uterine asphyxia results in long-lasting effects of baroreflex mediated fluctuations, limiting the possibility to buffer changes in SBP by adapting fetal heart rate. Finally, in Chapter 11 the results of all studies within this thesis are discussed and further research directions are proposed.