GRU-based temporal modeling of ECG derived physiological features for eclampsia outcome classification

Early detection of instability in the neurological and cardiovascular systems is imperative for the prevention of severe maternal complications such as eclampsia. Because of the increased occurrence of eclampsia during pregnancy weeks 20 or later, the importance of physiological monitoring is vital during the second and third trimesters. The findings clearly indicate the ability of temporal deep learning to classify physiological patterns related to eclampsia-related outcomes. In this regard, the suggested approach has been assessed on 10,000 instances of pregnancy-related ECG and clinical data, generated from clinically significant features obtained from various biomedical databases such as Kaggle, as well as the current clinical literature. The employed dataset includes multiple lead ECG signals along with the maternal clinical and demographic features. In the preprocessing step, data normalization and noise management were performed on the dataset. The Adaptive Synthetic Sampling (ADASYN) method was utilized to overcome class imbalance. Some of the significant ECG features, such as RR interval, QRS interval, and ECG axis, were extracted from the ECG signals to be able to use them as a feature vector representing unstable physiology. Furthermore, temporal relations between extracted features were modeled using the Gated Recurrent Unit network. The model presented in this paper was evaluated against Logistic Regression, SVM, Random Forest, XGBoost, and LSTM. The experimental results show that the GRU model outperforms all other comparison methods by scoring 97.2% for accuracy, 96.8% for precision, 97.0% for recall, and 96.9% for F1-score, while the ROC-AUC is 0.989, and PR-AUC scores 0.98. From these results, one can see how efficient temporal deep learning is in detecting physiological behavior and how it can be used in early warnings of instability from eclampsia. Further investigation involving independent subjects, clinical validation of results, and prospective ECG data collection would be necessary for the proposed method to apply to the clinical setting.

Authors

Publication Details

Journal
Scientific Reports
Published
2026-11-13
DOI
https://doi.org/10.1038/s41598-026-77966-9
Primary Topic
ECG Monitoring and Analysis
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

GRU-based temporal modeling of ECG derived physiological features for eclampsia outcome classification

N. Karthikeyan, Sathish Kumar V R, Shajidha R. S.
Scientific Reports
ECG Monitoring and Analysis
article

GRU-based temporal modeling of ECG derived physiological features for eclampsia outcome classification

N. Karthikeyan, Sathish Kumar V R, Shajidha R. S.
article en

Abstract

Early detection of instability in the neurological and cardiovascular systems is imperative for the prevention of severe maternal complications such as eclampsia. Because of the increased occurrence of eclampsia during pregnancy weeks 20 or later, the importance of physiological monitoring is vital during the second and third trimesters. The findings clearly indicate the ability of temporal deep learning to classify physiological patterns related to eclampsia-related outcomes. In this regard, the suggested approach has been assessed on 10,000 instances of pregnancy-related ECG and clinical data, generated from clinically significant features obtained from various biomedical databases such as Kaggle, as well as the current clinical literature. The employed dataset includes multiple lead ECG signals along with the maternal clinical and demographic features. In the preprocessing step, data normalization and noise management were performed on the dataset. The Adaptive Synthetic Sampling (ADASYN) method was utilized to overcome class imbalance. Some of the significant ECG features, such as RR interval, QRS interval, and ECG axis, were extracted from the ECG signals to be able to use them as a feature vector representing unstable physiology. Furthermore, temporal relations between extracted features were modeled using the Gated Recurrent Unit network. The model presented in this paper was evaluated against Logistic Regression, SVM, Random Forest, XGBoost, and LSTM. The experimental results show that the GRU model outperforms all other comparison methods by scoring 97.2% for accuracy, 96.8% for precision, 97.0% for recall, and 96.9% for F1-score, while the ROC-AUC is 0.989, and PR-AUC scores 0.98. From these results, one can see how efficient temporal deep learning is in detecting physiological behavior and how it can be used in early warnings of instability from eclampsia. Further investigation involving independent subjects, clinical validation of results, and prospective ECG data collection would be necessary for the proposed method to apply to the clinical setting.

Scientific Reports
Good health and well-being
ECG Monitoring and Analysis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon