A simple EEG-based decision tool for neonatal therapeutic hypothermia in hypoxic-ischemic encephalopathy

• Main Authors: Fiammante, Marc, Vermersch, Anne-Isabelle, Vidailhet, Marie, Chavez, Mario
• Format: Journal Article
• Language: English
• Published: 29-03-2024
• Subjects: Quantitative Biology - Neurons and Cognition
• DOI: 10.48550/arxiv.2403.20239

Indication of therapeutic hypothermia needs an accurate identification of
brain injury in the early neonatal period. Here, we aim to provide a simple
hypothermia decision-making tool for the term neonates with hypoxic-ischemic
encephalopathy (HIE) based on features of conventional electroencephalogram
(EEG) taken less than 6 hours from birth. EEG recordings from one hundred
full-term babies with HIE were included in the study. Each EEG recording was
graded by pediatric neurologists for HIE severity. Amplitude of each EEG
segment was analyzed in the slow frequency bands. Temporal fluctuations of
spectral power in delta (0.5 - 4 Hz) frequency band was used to characterize
each HIE grade. For each grade of abnormality, we estimated level and duration
(number of consecutive segments above a given level) probability densities for
power of delta oscillations. These 2D representation of EEG dynamics can
identify mild HIE group from those of requiring hypothermia. Our discrimination
system yielded an accuracy, recall, positive predictive value (precision),
negative predictive value, false alarm ratio and F1-score of 98%, 99%, 99%,
0.94%, 0.06 and 99%, respectively. These results provided an accurate
discrimination of mild versus moderate or severe HIE, and only one mild case
was erroneously detected as relevant for hypothermia. Quantized probability
densities of slow spectral features (delta power) from early conventional EEG
(withing 6 hours of birth) revealed significant differences in slow spectral
dynamics between infants with mild HIE grades and those relevant for
hypothermia.