TY - JOUR
T1 - Quantitative electroencephalogram and machine learning to predict expired sevoflurane concentration in infants
AU - BRAIN Collaborative Investigators
AU - Kumar, Rachit
AU - Skowno, Justin
AU - von Ungern-Sternberg, Britta S.
AU - Davidson, Andrew
AU - Xu, Ting
AU - Zhang, Jianmin
AU - Song, Xing Rong
AU - Zhang, Mazhong
AU - Zhao, Ping
AU - Liu, Huacheng
AU - Jiang, Yifei
AU - Zuo, Yunxia
AU - de Graaff, Jurgen C.
AU - Vutskits, Laszlo
AU - Olbrecht, Vanessa A.
AU - Szmuk, Peter
AU - Simpao, Allan F.
AU - Tsui, Fuchiang Fuchiang Rich
AU - Pratap, Jayant Jayant Nick
AU - Padiyath, Asif
AU - Nelson, Olivia
AU - Kurth, Charles D.
AU - Yuan, Ian
AU - Alex, Gijo
AU - Efune, Proshad
AU - Saynhalath, Rita
AU - Hu, Paula
AU - Lang, Shih Shan
AU - Huh, Jimmy W.
AU - Roden, Emilie
AU - van Hoorn, Camille
AU - Bhuju, Abhusani
AU - Du, Bin
AU - Chen, Panpan
AU - Ding, Mengmeng
AU - Zheng, Jijian
AU - Lei, Dong Xu
AU - Gao, Zheng Zheng
AU - Sommerfield, David
AU - Sheppard, Suzette
AU - Menezes, Minal
N1 - © 2025. The Author(s).
PY - 2025/10
Y1 - 2025/10
N2 - Processed electroencephalography (EEG) indices used to guide anesthetic dosing in adults are not validated in young infants. Raw EEG can be processed mathematically, yielding quantitative EEG parameters (qEEG). We hypothesized that machine learning combined with qEEG can accurately classify expired sevoflurane concentrations in young infants. Knowledge from this may contribute to development of future infant-specific EEG algorithms. Frontal EEG collected from infants ≤ 3 months were time-matched as one-minute epochs to expired sevoflurane (eSevo). Fifteen qEEG parameters were extracted from each epoch and eight machine learning models combined the qEEG to classify each epoch into one of four eSevo levels (%): 0.1–1.0, 1.0–2.1, 2.1–2.9, and > 2.9. 64 epochs formed the post hoc SHAP dataset to determine the qEEG that contributed most to the model. The remaining epochs were randomly split 50 times into 80/20 training/testing sets. Accuracy and F1-score determined model performance. 42 infants provided 4574 epochs. The top classifiers K-nearest neighbors, default multi-layer perceptron, and support vector machine achieved 67.5–68.7% accuracy. Burst suppression ratio and entropy β were the top contributors to the models. Post hoc analysis performed without burst suppression ratio yielded similar prediction performance. In young infants, machine learning applied to qEEG predicted eSevo levels with moderate success. Burst suppression ratio, the most important contributor, represented an efficient EEG feature that encapsulated underlying EEG changes seen on other qEEG features. These results provided insight into EEG parameter selection and optimal machine learning models used for future development of infant-specific EEG algorithms.
AB - Processed electroencephalography (EEG) indices used to guide anesthetic dosing in adults are not validated in young infants. Raw EEG can be processed mathematically, yielding quantitative EEG parameters (qEEG). We hypothesized that machine learning combined with qEEG can accurately classify expired sevoflurane concentrations in young infants. Knowledge from this may contribute to development of future infant-specific EEG algorithms. Frontal EEG collected from infants ≤ 3 months were time-matched as one-minute epochs to expired sevoflurane (eSevo). Fifteen qEEG parameters were extracted from each epoch and eight machine learning models combined the qEEG to classify each epoch into one of four eSevo levels (%): 0.1–1.0, 1.0–2.1, 2.1–2.9, and > 2.9. 64 epochs formed the post hoc SHAP dataset to determine the qEEG that contributed most to the model. The remaining epochs were randomly split 50 times into 80/20 training/testing sets. Accuracy and F1-score determined model performance. 42 infants provided 4574 epochs. The top classifiers K-nearest neighbors, default multi-layer perceptron, and support vector machine achieved 67.5–68.7% accuracy. Burst suppression ratio and entropy β were the top contributors to the models. Post hoc analysis performed without burst suppression ratio yielded similar prediction performance. In young infants, machine learning applied to qEEG predicted eSevo levels with moderate success. Burst suppression ratio, the most important contributor, represented an efficient EEG feature that encapsulated underlying EEG changes seen on other qEEG features. These results provided insight into EEG parameter selection and optimal machine learning models used for future development of infant-specific EEG algorithms.
KW - Infant electroencephalogram EEG
KW - Machine learning
KW - Pediatric EEG anesthesia
KW - Quantitative EEG anesthesia
KW - SHAP EEG analysis
KW - Anesthetics, Inhalation
KW - Reproducibility of Results
KW - Electroencephalography/methods
KW - Humans
KW - Infant
KW - Male
KW - Entropy
KW - Support Vector Machine
KW - Machine Learning
KW - Sevoflurane
KW - Algorithms
KW - Signal Processing, Computer-Assisted
KW - Female
KW - Infant, Newborn
UR - https://www.scopus.com/pages/publications/105005423010
U2 - 10.1007/s10877-025-01301-2
DO - 10.1007/s10877-025-01301-2
M3 - Article
C2 - 40381151
AN - SCOPUS:105005423010
SN - 1387-1307
VL - 39
SP - 999
EP - 1014
JO - Journal of Clinical Monitoring and Computing
JF - Journal of Clinical Monitoring and Computing
IS - 5
M1 - 036031
ER -