Related Experiment Video
Updated: Dec 23, 2025
08:07
Author Spotlight: Advancing Labor Management Through Electromyometrial Imaging for Understanding Uterine Contractions
Published on: May 26, 2023
1.7K
Evaluation of electrohysterogram measured from different gestational weeks for recognizing preterm delivery: a
Jin Peng1, Dongmei Hao1, Lin Yang1
1College of Life Science and Bioengineering, Beijing University of Technology, Intelligent Physiological Measurement and Clinical Translation, Beijing International Platform for Scientific and Technological Cooperation, Beijing, China.
Biocybernetics and Biomedical Engineering
|April 21, 2020
Summary
A computational method using electrohysterogram (EHG) signals effectively recognizes preterm delivery. This approach, utilizing random forest (RF) classification, shows high accuracy even with EHG recordings taken before 26 weeks of gestation.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Obstetrics
Background:
- Preterm delivery is a significant global health concern requiring early detection.
- Accurate prediction of preterm birth is crucial for timely intervention and improved neonatal outcomes.
- Electrohysterogram (EHG) signals offer a non-invasive method for monitoring uterine activity.
Purpose of the Study:
- To develop and evaluate a computational method for recognizing preterm delivery using electrohysterogram (EHG) signals.
- To assess the efficacy of the random forest (RF) algorithm in classifying preterm and term deliveries based on EHG data.
- To compare the performance of the RF model using EHG signals recorded at different gestational stages.
Main Methods:
- EHG signals from 300 pregnant women were analyzed.
- Signals were categorized into two groups based on recording time: before 26 weeks (PE/TE) and during/after 26 weeks (PL/TL).
- 31 linear and nonlinear features were extracted from EHG signals, and a random forest classifier was employed with adaptive synthetic sampling and cross-validation.
Main Results:
- The random forest model achieved high performance in both groups: for PL/TL, Accuracy (ACC)=0.93, Sensitivity=0.89, Specificity=0.97, AUC=0.80; for PE/TE, ACC=0.92, Sensitivity=0.88, Specificity=0.96, AUC=0.88.
- The model demonstrated effective preterm delivery recognition even with EHG signals recorded prior to 26 weeks of gestation.
- Feature analysis indicated the potential of both linear and nonlinear EHG signal characteristics for classification.
Conclusions:
- Random forest classification of EHG signals is an effective method for recognizing preterm delivery.
- The computational approach demonstrates high accuracy and reliability, applicable even for early gestational EHG recordings.
- This study supports the potential of EHG-based computational methods for early and accurate prediction of preterm birth.
Keywords:
ACC, accuracyADASYN, adaptive synthetic sampling approachANN, artificial neural networkAR, auto-regressive modelAUC, the area under the curveCorrDim, correlation dimensionDT, decision treeEHG, electrohysterogramElectrohysterogram (EHG)Feature extractionGestational weekIUPC, intrauterine pressure catheterK-NN, K-nearestLDA, linear discriminant analysisLE, Lyapunov exponentMDF, median frequencyMNF, mean frequencyPE, preterm delivery before the 26th week of gestationPF, peak frequencyPL, preterm delivery after the 26th week of gestationPreterm deliveryQDA, quadratic discriminant analysisRF, random forestRMS, root mean squareROC, the receiver operating characteristic curveRandom forest (RF).SD, standard deviationSE, energy values in signalSM, maximum values in signalSS, singular values in signalSV, variance values in signalSVM, support vector machineSampEn, sample entropyTE, term delivery before the 26th week of gestationTL, term delivery after the 26th week of gestationTOCO, tocodynamometerTPEHG, term-preterm electrohysterogramTr, time reversibilityτz, zero-crossing
