Authors:
V. Devigasri, M. Mohamed Sirajudeen
Addresses:
Department of Computer Science, Nilgiri College of Arts and Science, Thaloor, The Nilgiris, Tamil Nadu, India. School of Computer Science and Emerging Technologies, Nilgiri College of Arts and Science, Thaloor, The Nilgiris, Tamil Nadu, India.
Abstract:
Continuous fetal monitoring in pregnancy is an important aspect to demonstrate fetal well-being to reduce the risk of perinatal complications. Basic analysis of fetal heart rate (FHR) and uterine contractions (UC) is limited to direct analysis. It cannot account for complex temporalāspatial patterns of these signals, leading to misclassification of fetal distress and increasing false alarm rates in fetal heart rate pattern categorisation. Current approaches in recommended clinical systems typically use variations of single deep learning (DL) architectures, such as convolutional neural networks (CNNs) or long short-term memory (LSTM) networks, that operate on either spatial or temporal features alone; this reduces predictive robustness when classifying fetal heart rate patterns and distress. This study presents a Deep Ensemble Classifier (DEC) to provide multimodal feature extraction from Continuous Fetal Monitoring (CFM) signals, using CNNs, Bidirectional Long Short-Term Memory (Bi-LSTM), and transformer-based attention mechanisms to enhance sensing of spatial and temporal dimensions. The experimental evaluation on the CTU-UHB Intrapartum Cardiotocography dataset shows significant improvement over 19 state-of-the-art methods, achieving 97.8% accuracy, an F1 score of 0.96, and an AUC of 0.98. The comparative analysis using confusion matrices and ROC curves could further enhance the interpretation that the framework can mitigate false positives and negatives, which are immensely important for clinical decision-making. The ensemble framework has the potential for real-time integration into fetal monitoring systems, which would enable clinicians to detect fetal distress earlier and improve prenatal care.
Keywords: Deep Ensemble; Bi-LSTM and Transformer; Convolutional Neural Networks; Prenatal Health Assessment; Multimodal Analysis; Uterine Contractions; Deep Ensemble Classifier.
Received on: 10/04/2025, Revised on: 03/08/2025, Accepted on: 02/11/2025, Published on: 07/06/2026
DOI: 10.64091/ATIHL.2026.000304
AVE Trends in Intelligent Health Letters, 2026 Vol. 3 No. 2 , Pages: 101-115