A CNN-RNN unified framework for intrapartum cardiotocograph classification

Area of Science:

• Medical Technology
• Artificial Intelligence in Healthcare
• Fetal Medicine

Background:

• Prenatal fetal monitoring is vital for assessing fetal growth and health.
• Early detection of fetal abnormalities prevents hypoxia and mortality.
• Current methods rely on fetal heart rate monitoring equipment.

Purpose of the Study:

• To develop an advanced model for evaluating fetal health status using fetal heart rate and uterine contraction signals.
• To improve the accuracy and reliability of prenatal fetal monitoring.
• To assist obstetricians in clinical decision-making for better patient outcomes.

Main Methods:

• Utilized a hybrid deep learning model integrating 1D-CNN (One Dimension Convolutional Neural Network) and GRU (Gated Recurrent Unit).
• Implemented data preprocessing and enhancement techniques to balance class distribution in the training set.
• Employed standard evaluation metrics including accuracy, sensitivity, specificity, and ROC analysis.

Main Results:

• Achieved a high accuracy of 95.15% on the test set.
• Demonstrated excellent sensitivity at 96.20% and specificity at 94.09%.
• The model effectively classifies fetal health status based on physiological signals.

Conclusions:

• The proposed 1D-CNN and bidirectional GRU hybrid model is effective for fetal health status evaluation.
• This approach can significantly assist obstetricians in clinical decision-making.
• Establishes a baseline for integrating advanced deep learning models in fetal health monitoring.