A Deep Learning Based Approach for Detecting Fetal Stress Using CTG Signal

Read full text Summarise with VIRAT
Authors: S.M.Seeni Mohamed Aliar Maraikkayar, Rajendran Tamilselvi, M. Parisa Beham
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Detecting fetal stress accurately and promptly is critical for preventing adverse perinatal outcomes. Cardiotocography (CTG) is a standard monitoring technique widely used during pregnancy to assess fetal health by recording fetal heart rate (FHR) and uterine contraction (UC) signals. However, traditional methods of interpreting CTG signals are often subject to observer variability and may result in delayed or inaccurate diagnoses. Also, traditional feature-based classification methods require manual reading of the morphological characteristics from the fetal heart rate (FHR) curve to capture the signal's features. This process is time-consuming, costly, and prone to calibration bias. To address this, we propose a deep learning-based approach for detecting fetal stress using CTG signal data. This study introduces a novel Enhanced-VGGNet based method for classifying FHR signals, which reduces human error. This algorithm learns superficial features directly from the data, using AlexNet and GoogleNet that enabling the real-time diagnosis of FHR data. The dataset employed in this study is from kaggle dataset which includes numerous attributes for feature extraction. The proposed enhanced VGGNet serves as the classification model, categorizing the signal into Normal, Suspicious, and Pathological classes. This automated approach outperforms traditional machine learning classification methods, which struggle to incorporate a greater number of attributes for more accurate feature extraction. The efficiency of the proposed methodology is then assessed using several performance parameters. It provides sensitivity, specificity and accuracy of 95.8%, 91.2% & 96.6% respectively which shows its outperformance compared with other methods

Fields of Study & Subjects

ECG Monitoring and Analysis Artificial Intelligence in Healthcare Heart Rate Variability and Autonomic Control

Publication details

Journal / source
Traitement du signal
Volume
43
Issue
1
Pages
171–183
Publication date
28 February 2026
Language
EN