Topography of maturational changes in EEG burst spectral power of the preterm infant with a normal follow-up at 2 years of age

Deep Learning for Generalized EEG Seizure Detection after Hypoxia-Ischemia-Preclinical Validation

Brain maturity and many clinical treatments such as therapeutic hypothermia (TH) can significantly influence the morphology of neonatal EEG seizures after hypoxia-ischemia (HI), and so there is a need for generalized automatic seizure identification. This study validates efficacy of advanced deep-learning pattern classifiers based on a convolutional neural network (CNN) for seizure detection after HI in fetal sheep and determines the effects of maturation and brain cooling on their accuracy. The cohorts included HI-normothermia term (n = 7), HI-hypothermia term (n = 14), sham-normothermia term (n = 5), and HI-normothermia preterm (n = 14) groups, with a total of >17,300 h of recordings. Algorithms were trained and tested using leave-one-out cross-validation and k-fold cross-validation approaches. The accuracy of the term-trained seizure detectors was consistently excellent for HI-normothermia preterm data (accuracy = 99.5%, area under curve (AUC) = 99.2%). Conversely, when the HI-normothermia preterm data were used in training, the performance on HI-normothermia term and HI-hypothermia term data fell (accuracy = 98.6%, AUC = 96.5% and accuracy = 96.9%, AUC = 89.6%, respectively). Findings suggest that HI-normothermia preterm seizures do not contain all the spectral features seen at term. Nevertheless, an average 5-fold cross-validated accuracy of 99.7% (AUC = 99.4%) was achieved from all seizure detectors. This significant advancement highlights the reliability of the proposed deep-learning algorithms in identifying clinically translatable post-HI stereotypic seizures in 256Hz recordings, regardless of maturity and with minimal impact from hypothermia.

Quantitative analysis of high-frequency activity in neonatal EEG

OBJECTIVE

To determine the presence and potential utility of independent high-frequency activity recorded from scalp electrodes in the electroencephalogram (EEG) of newborns.

METHODS

We compare interburst intervals and continuous activity at different frequencies for EEGs retrospectively recorded at 256 Hz from 4 newborn groups: 1) 36 preterms (<32 weeks' gestational age, GA); 2) 12 preterms (32-37 weeks' GA); 3) 91 healthy full terms; 4) 15 full terms with hypoxic-ischemic encephalopathy (HIE). At 4 standard frequency bands (delta, 0.5-3 Hz; theta, 3-8 Hz; alpha, 8-15 Hz; beta, 15-30 Hz) and 3 higher-frequency bands (gamma1, 30-48 Hz; gamma2, 52-99 Hz; gamma3, 107-127 Hz), we compared power spectral densities (PSDs), quantitative features, and machine learning model performance. Feature selection and further machine learning methods were performed on one cohort.

RESULTS

We found significant (P < 0.01) differences in PSDs, quantitative analysis, and machine learning modelling at the higher-frequency bands. Machine learning models using only high-frequency features performed best in preterm groups 1 and 2 with a median (95% confidence interval, CI) Matthews correlation coefficient (MCC) of 0.71 (0.12-0.88) and 0.66 (0.36-0.76) respectively. Interburst interval-detector models using both high- and standard-bandwidths produced the highest median MCCs in all four groups. High-frequency features were largely independent of standard-bandwidth features, with only 11/84 (13.1%) of correlations statistically significant. Feature selection methods produced 7 to 9 high-frequency features in the top 20 feature set.

CONCLUSIONS

This is the first study to identify independent high-frequency activity in newborn EEG using in-depth quantitative analysis. Expanding the EEG bandwidths of analysis has the potential to improve both quantitative and machine-learning analysis, particularly in preterm EEG.

EEG effective connectivity during the first year of life mirrors brain synaptogenesis, myelination, and early right hemisphere predominance

INTRODUCTION

The maturation of electroencephalogram (EEG) effective connectivity in healthy infants during the first year of life is described.

METHODS

Participants: A cross-sectional sample of 125 healthy at-term infants, from 0 to 12 months of age, underwent EEG in a state of quiet sleep.

PROCEDURES

The EEG primary currents at the source were described with the sLoreta method. An unmixing algorithm was applied to reduce the leakage, and the isolated effective coherence, a direct and directed measurement of information flow, was calculated.

RESULTS AND DISCUSSION

Initially, the highest indices of connectivity are at the subcortical nuclei, continuing to the parietal lobe, predominantly the right hemisphere, then expanding to temporal, occipital, and finally the frontal areas, which is consistent with the myelination process. Age-related connectivity changes were mostly long-range and bilateral. Connections increased with age, mainly in the right hemisphere, while they mainly decreased in the left hemisphere. Increased connectivity from 20 to 30 Hz, mostly at the right hemisphere. These findings were consistent with right hemisphere predominance during the first three years of life. Theta and alpha connections showed the greatest changes with age. Strong connectivity was found between the parietal, temporal, and occipital regions to the frontal lobes, responsible for executive functions and consistent with behavioral development during the first year. The thalamus exchanges information bidirectionally with all cortical regions and frequency bands.

CONCLUSIONS

The maturation of EEG connectivity during the first year in healthy infants is very consistent with synaptogenesis, reductions in synaptogenesis, myelination, and functional and behavioral development.

Deep Learning for EEG Seizure Detection in Preterm Infants

EEG is the gold standard for seizure detection in the newborn infant, but EEG interpretation in the preterm group is particularly challenging; trained experts are scarce and the task of interpreting EEG in real-time is arduous. Preterm infants are reported to have a higher incidence of seizures compared to term infants. Preterm EEG morphology differs from that of term infants, which implies that seizure detection algorithms trained on term EEG may not be appropriate. The task of developing preterm specific algorithms becomes extra-challenging given the limited amount of annotated preterm EEG data available. This paper explores novel deep learning (DL) architectures for the task of neonatal seizure detection in preterm infants. The study tests and compares several approaches to address the problem: training on data from full-term infants; training on data from preterm infants; training on age-specific preterm data and transfer learning. The system performance is assessed on a large database of continuous EEG recordings of 575[Formula: see text]h in duration. It is shown that the accuracy of a validated term-trained EEG seizure detection algorithm, based on a support vector machine classifier, when tested on preterm infants falls well short of the performance achieved for full-term infants. An AUC of 88.3% was obtained when tested on preterm EEG as compared to 96.6% obtained when tested on term EEG. When re-trained on preterm EEG, the performance marginally increases to 89.7%. An alternative DL approach shows a more stable trend when tested on the preterm cohort, starting with an AUC of 93.3% for the term-trained algorithm and reaching 95.0% by transfer learning from the term model using available preterm data. The proposed DL approach avoids time-consuming explicit feature engineering and leverages the existence of the term seizure detection model, resulting in accurate predictions with a minimum amount of annotated preterm data.

Measuring Cot-Side the Effects of Parenteral Nutrition on Preterm Cortical Function

Early nutritional compromise after preterm birth is shown to affect long-term neurodevelopment, however, there has been a lack of early functional measures of nutritional effects. Recent progress in computational electroencephalography (EEG) analysis has provided means to measure the early maturation of cortical activity. Our study aimed to explore whether computational metrics of early sequential EEG recordings could reflect early nutritional care measured by energy and macronutrient intake in the first week of life. A higher energy or macronutrient intake was assumed to associate with improved development of the cortical activity. We analyzed multichannel EEG recorded at 32 weeks (32.4 ± 0.7) and 36 weeks (36.6 ± 0.9) of postmenstrual age in a cohort of 28 preterm infants born before 32 weeks of postmenstrual age (range: 24.3–32 weeks). We computed several quantitative EEG measures from epochs of quiet sleep (QS): (i) spectral power; (ii) continuity; (iii) interhemispheric synchrony, as well as (iv) the recently developed estimate of maturational age. Parenteral nutritional intake from day 1 to day 7 was monitored and clinical factors collected. Lower calories and carbohydrates were found to correlate with a higher reduction of spectral amplitude in the delta band. Lower protein amount associated with higher discontinuity. Both higher proteins and lipids intake correlated with a more developmental increase in interhemispheric synchrony as well as with better progress in the estimate of EEG maturational age (EMA). Our study shows that early nutritional balance after preterm birth may influence subsequent maturation of brain activity in a way that can be observed with several intuitively reasoned and transparent computational EEG metrics. Such measures could become early functional biomarkers that hold promise for benchmarking in the future development of therapeutic interventions.

Brain maturation in the first 3 months of life, measured by electroencephalogram: A comparison between preterm and term-born infants

OBJECTIVE

Preterm infants are at risk for altered brain maturation resulting in neurodevelopmental impairments. Topographical analysis of high-density electroencephalogram during sleep matches underlying brain maturation. Using such an EEG mapping approach could identify preterm infants at risk early in life.

METHODS

20 preterm (gestational age < 32 weeks) and 20 term-born infants (gestational age > 37 weeks) were recorded by 18-channel daytime sleep-EEG at term age (GA 40 weeks for preterm and 2-3 days after birth for term infants) and 3 months (corrected age for preterm infants).

RESULTS

Preterm infant's power spectrum at term age is immature, leveling off with term infants at 3 months of age. Topographical distribution of maximal power density however, reveals qualitative differences between the groups until 3 months of age. Preterm infants exhibit more temporal than central activation at term age and more occipital than central activation at 3 months of age. Moreover, being less mature at term age predicts being less mature at 3 months of age.

CONCLUSION

Topographical analysis of sleep EEG reveals changes in brain maturation between term and preterm infants early in life.

SIGNIFICANCE

In future, automated analysis tools using topographical power distribution could help identify preterm infants at risk early in life.

Quantitative Preterm EEG Analysis: The Need for Caution in Using Modern Data Science Techniques

Hemodynamic changes during neonatal transition increase the vulnerability of the preterm brain to injury. Real-time monitoring of brain function during this period would help identify the immediate impact of these changes on the brain. Neonatal EEG provides detailed real-time information about newborn brain function but can be difficult to interpret for non-experts; preterm neonatal EEG poses even greater challenges. An objective quantitative measure of preterm brain health would be invaluable during neonatal transition to help guide supportive care and ultimately protect the brain. Appropriate quantitative measures of preterm EEG must be calculated and care needs to be taken when applying the many techniques available for this task in the era of modern data science. This review provides valuable information about the factors that influence quantitative EEG analysis and describes the common pitfalls. Careful feature selection is required and attention must be paid to behavioral state given the variations encountered in newborn EEG during different states. Finally, the detrimental influence of artifacts on quantitative EEG analysis is illustrated.

Visual and Quantitative Electroencephalographic Analysis in Healthy Term Neonates Within the First Six Hours and the Third Day of Life

BACKGROUND

What constitutes a "normal" background electroencephalography (EEG) rhythm immediately after birth is not well understood. We performed video-electroencephalography recordings in the first six hours (first measure) and the third day of life (second measure) for evidence of transient changes in brain function.

METHODS

We performed a cohort study of an incidental sample of healthy term neonates in a single-center nursery. Main outcome measures were as follows: (1) EEG visual analysis, which included sleep-wake cycles, proportions of discontinuity and bursts with delta brushes, and number per hour of alpha/theta rolandic activity, encoches frontales, and transients; and (2) the electroencephalographic spectral analysis, which included power spectrum in the following frequency bands: delta, 0.5 to 4 Hz; theta, 4 to 8 Hz; alpha, 8 to 13 Hz; and beta, 13 to 30 Hz. Theta/delta and alpha/delta ratios were also calculated.

RESULTS

Twenty-two babies were enrolled. Significant findings (P < 0.05) in the first six hours with respect to 48 to 72 hours of life were (1) increased discontinuity, indeterminate sleep, and bursts with delta brushes; (2) higher number of transients, and lower number of alpha/theta rolandic activity and encoches frontales. Minimal changes were found in power spectrum data. However, using receiver operating characteristic curve analysis, theta/delta ratio ≤0.484 was the best cutoff to discriminate between the two measures (positive predictive value, 100.0; 95% confidence interval 71.0 to 100).

CONCLUSIONS

In healthy term neonates, immature electroencephalographic patterns, lack of clearly defined sleep-wake cycles, and frequent transients can be considered normal electroencephalographic findings in the first six hours of life. Normative power spectrum data are provided. These findings suggest that neonatal adaptation immediately after birth leads to transient changes in brain function.

A review of important electroencephalogram features for the assessment of brain maturation in premature infants

This review describes the maturational features of the baseline electroencephalogram (EEG) in the neurologically healthy preterm infant. Features such as continuity, sleep state, synchrony and transient waveforms are described, even from extremely preterm infants and includes abundant illustrated examples. The physiological significance of these EEG features and their relationship to neurodevelopment are highlighted where known. This review also demonstrates the importance of multichannel conventional EEG monitoring for preterm infants as many of the features described are not apparent if limited channel EEG monitors are used.

CONCLUSION

This review aims to provide healthcare professionals in the neonatal intensive care unit with guidance on the more common normal maturational features seen in the EEG of preterm infants.

Large-scale network organization of EEG functional connectivity in newborn infants

The organization of functional brain networks changes across human lifespan. The present study analyzed functional brain networks in healthy full-term infants (N = 139) within 1-6 days from birth by measuring neural synchrony in EEG recordings during quiet sleep. Large-scale phase synchronization was measured in six frequency bands with the Phase Lag Index. Macroscopic network organization characteristics were quantified by constructing unweighted minimum spanning tree graphs. The cortical networks in early infancy were found to be significantly more hierarchical and had a more cost-efficient organization compared with MST of random control networks, more so in the theta and alpha than in other frequency bands. Frontal and parietal sites acted as the main hubs of these networks, the topological characteristics of which were associated with gestation age (GA). This suggests that individual differences in network topology are related to cortical maturation during the prenatal period, when functional networks shift from strictly centralized toward segregated configurations. Hum Brain Mapp 38:4019-4033, 2017. © 2017 Wiley Periodicals, Inc.

An Automated Quiet Sleep Detection Approach in Preterm Infants as a Gateway to Assess Brain Maturation

Sleep state development in preterm neonates can provide crucial information regarding functional brain maturation and give insight into neurological well being. However, visual labeling of sleep stages from EEG requires expertise and is very time consuming, prompting the need for an automated procedure. We present a robust method for automated detection of preterm sleep from EEG, over a wide postmenstrual age (PMA = gestational age + postnatal age) range, focusing first on Quiet Sleep (QS) as an initial marker for sleep assessment. Our algorithm, CLuster-based Adaptive Sleep Staging (CLASS), detects QS if it remains relatively more discontinuous than non-QS over PMA. CLASS was optimized on a training set of 34 recordings aged 27–42 weeks PMA, and performance then assessed on a distinct test set of 55 recordings of the same age range. Results were compared to visual QS labeling from two independent raters (with inter-rater agreement Kappa = 0. 93), using Sensitivity, Specificity, Detection Factor (DF = proportion of visual QS periods correctly detected by CLASS) and Misclassification Factor (MF = proportion of CLASS-detected QS periods that are misclassified). CLASS performance proved optimal across recordings at 31–38 weeks (median DF = 1.0, median MF 0–0.25, median Sensitivity 0.93–1.0, and median Specificity 0.80–0.91 across this age range), with minimal misclassifications at 35–36 weeks (median MF = 0). To illustrate the potential of CLASS in facilitating clinical research, normal maturational trends over PMA were derived from CLASS-estimated QS periods, visual QS estimates, and nonstate specific periods (containing QS and non-QS) in the EEG recording. CLASS QS trends agreed with those from visual QS, with both showing stronger correlations than nonstate specific trends. This highlights the benefit of automated QS detection for exploring brain maturation.

Association between theta power in 6-month old infants at rest and maternal PTSD severity: A pilot study

Compared to infants born to mothers without PTSD, infants born to mothers with active PTSD develop poorer behavioral reactivity and emotional regulation. However, the association between perinatal maternal PTSD and infant neural activation remains largely unknown. This pilot study (N=14) examined the association between perinatal PTSD severity and infant frontal neural activity, as measured by MEG theta power during rest. Results indicated that resting left anterior temporal/frontal theta power was correlated with perinatal PTSD severity (p=0.004). These findings suggest delayed cortical maturation in infants whose mothers had higher perinatal PTSD severity and generate questions regarding perinatal PTSD severity and infant neurophysiological consequences.

Structural damage in early preterm brain changes the electric resting state networks

A robust functional bimodality is found in the long-range spatial correlations of newborn cortical activity, and it likely provides the developmentally crucial functional coordination during the initial growth of brain networks. This study searched for possible acute effects on this large scale cortical coordination after acute structural brain lesion in early preterm infants. EEG recordings were obtained from preterm infants without (n=11) and with (n=6) haemorrhagic brain lesion detected in their routine ultrasound exam. The spatial cortical correlations in band-specific amplitudes were examined within two amplitude regimes, high and low amplitude periods, respectively. Technical validation of our analytical approach showed that bimodality of this kind is a genuine physiological characteristic of each brain network. It was not observed in datasets created from uniform noise, neither is it found between randomly paired signals. Hence, the observed bimodality arises from specific interactions between cortical regions. We found that significant long-range amplitude correlations are found in most signal pairs in both groups at high amplitudes, but the correlations are generally weaker in newborns with brain lesions. The group difference is larger during high mode, however the difference did not have any statistically apparent topology. Graph theoretical analysis confirmed a significantly larger weight dispersion in the newborns with brain lesion. Comparison of graph measures to a child's performance at two years showed that lower clustering coefficient and weight dispersion were both correlated to better neurodevelopmental outcomes. Our findings suggest that the common preterm brain haemorrhage causes diffuse changes in the functional long-range cortical correlations. It has been recently recognized that the high mode network activity is crucial for early brain development. The present observations may hence offer a mechanistic link between early lesion and the later emergence of complex neurocognitive sequelae.

Functional connectivity in preterm infants derived from EEG coherence analysis

OBJECTIVE

To quantify the neuronal connectivity in preterm infants between homologous channels of both hemispheres.

METHODS

EEG coherence analysis was performed on serial EEG recordings collected from preterm infants with normal neurological follow-up. The coherence spectrum was divided in frequency bands: δnewborn(0-2 Hz), θnewborn(2-6 Hz), αnewborn(6-13 Hz), βnewborn(13-30 Hz). Coherence values were evaluated as a function of gestational age (GA) and postnatal maturation.

RESULTS

All spectra show two clear peaks in the δnewborn and θnewborn-band, corresponding to the delta and theta EEG waves observed in preterm infants. In the δnewborn-band the peak magnitude coherence decreases with GA and postnatal maturation for all channels. In the θnewborn-band, the peak magnitude coherence decreases with GA for all channels, but increases with postnatal maturation for the frontal polar channels. In the βnewborn-band a modest magnitude coherence peak was observed in the occipital channels, which decreases with GA.

CONCLUSIONS

Interhemispherical connectivity develops analogously with electrocortical maturation: signal intensities at low frequencies decrease with GA and postnatal maturation, but increase at high frequencies with postnatal maturation. In addition, peak magnitude coherence is a clear trend indicator for brain maturation.

SIGNIFICANCE

Coherence analysis can aid in the clinical assessment of the functional connectivity of the infant brain with maturation.

Early Brain Activity Relates to Subsequent Brain Growth in Premature Infants

Recent experimental studies have shown that early brain activity is crucial for neuronal survival and the development of brain networks; however, it has been challenging to assess its role in the developing human brain. We employed serial quantitative magnetic resonance imaging to measure the rate of growth in circumscribed brain tissues from preterm to term age, and compared it with measures of electroencephalographic (EEG) activity during the first postnatal days by 2 different methods. EEG metrics of functional activity were computed: EEG signal peak-to-peak amplitude and the occurrence of developmentally important spontaneous activity transients (SATs). We found that an increased brain activity in the first postnatal days correlates with a faster growth of brain structures during subsequent months until term age. Total brain volume, and in particular subcortical gray matter volume, grew faster in babies with less cortical electrical quiescence and with more SAT events. The present findings are compatible with the idea that (1) early cortical network activity is important for brain growth, and that (2) objective measures may be devised to follow early human brain activity in a biologically reasoned way in future research as well as during intensive care treatment.