Rigor of Neurovascular Coupling (NVC) Assessment in Newborns Using Different Amplitude EEG Algorithms

Abstract

Birth asphyxia constitutes a major global public health burden for millions of infants with a critical need for real time physiological biomarkers. This proof of concept study targets the translational rigor of such biomarkers and aims to examine whether the variability in the amplitude-integrated EEG (aEEG) outputs impact the determination of neurovascular coupling (NVC) in newborns with encephalopathy. A convenience sample with neonatal asphyxia were monitored for twenty hours in the first day of life with EEG and near infrared spectroscopy (NIRS)-based cerebral tissue oxygen saturation (SctO2). NVC between aEEG and NIRS-SctO2 was assessed using wavelet transform coherence (WTC) analysis, specifically by the wavelet total pixel number of significant coherences within 95% confidence interval. The raw EEG was converted to aEEG using three different methods: Method (M1) derives from the algorithm by Zhang and Ding. Method (M2) uses a Neonatal EEG Analysis Toolbox (WU-NEAT). Method (M3) extracts output directly from a commercial platform with an undisclosed algorithm. Our results demonstrate excellent agreement with Bland Altman comparisons for WTC-based NVC irrespective of the algorithms used, despite significant heterogeneities in the aEEG tracings produced by three algorithms. Our findings confirm the robustness of NVC wavelet analysis in Neonatal Encephalopathy related to HIE.

Wavelet-based neurovascular coupling can predict brain abnormalities in neonatal encephalopathy

“…For this specific study, the time series of EEG from a cross-hemisphere electrode pair of C3 and C4 (i.e., two electrodes in the central region) were used for data analysis of all the neonates. The EEG data from C3-C4 channel pair were first passed through an asymmetric band-pass filter (Parks-McClellan linear-phase FIR filter), which strongly attenuated the signal below 2 Hz and above 15 Hz, followed by conversion to aEEG using Washington University-Neonatal EEG Analysis Toolbox (WU-NEAT) ( Vesoulis et al, 2020 , Das et al, 2020 ). Artificial spikes from aEEG data were first detected and interpolated with neighboring data points.…”

“…We used a MATLAB-based software package and analytic Morlet wavelet to perform WTC analysis ( Grinsted et al, 2004 ) between the spontaneous oscillations of artifact-free NIRS-SctO2 and aEEG signals in neonates with encephalopathy, as reported in our recent publications ( Das et al, 2020 , Tian et al, 2016 , Tian, 2020 ). WTC is a time–frequency domain analysis and characterizes the squared cross-wavelet coherence, R 2 , and relative phase, Δ ϕ , between two time series at multiple time scales and over the entire time duration, without a prior assumption of linearity and stationarity.…”

“…All the pixels which have statistically significant coherence values are marked as black contours within the time-scale map. Using a time-scale (equivalent to time–frequency) domain, NVC was assessed by percentage of the number of significant pixels ( Pix sig ) over the number of total pixels ( Pix total ) within the given time-scale map ( Das et al, 2020 , Tian et al, 2016 , Tian, 2020 ), namely, Pix sig / Pix total . Specifically, Pix sig and Pix total were quantified by summing pixel numbers across all wavelet time scales only within the identified contours and all scales respectively.…”

Feasibility of EEG Phase-Amplitude Coupling to Stratify Encephalopathy Severity in Neonatal HIE Using Short Time Window

“…Others have employed the power of conventional EEG and the traces of amplitude EEG to identify HIE [ 27 , 28 , 29 , 30 , 31 ], but the use of tPAC m had not yet been explored. Moreover, our group has recently implemented neurovascular coupling to identify impaired cerebral autoregulation in neonates with HIE utilizing the wavelet coherence between clinical SO 2 and aEEG recordings [ 32 , 33 ]. Although these methods have shown promising results in differentiating severity of HIE, they all require a long recording time of at least 6 h to make accurate clinical decisions, which limits their utility in the modern TH era.…”

Artificial Intelligence Analysis of EEG Amplitude in Intensive Heart Care

“…Felze: et al proposed a probabilistic neural network classification model for EEG signal classification and recognition. Das et al [19] used modular neural networks to classify large-scale EEG signals and achieved good results. Dereymaeker et al [20] used wavelet transform to extract features, and the accuracy of the classification of the four types of ECG signals by the multilayer perceptron network was 94%.…”