A dilated inception CNN-LSTM network for fetal heart rate estimation.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence in Medicine

Background:

• Fetal heart rate (HR) monitoring is crucial for assessing fetal well-being during pregnancy and labor.
• Noninvasive fetal electrocardiogram (ECG) offers a promising alternative to traditional monitoring methods.
• Extracting fetal HR from abdominal ECG signals is challenging due to low signal-to-noise ratio and nonstationarity, hindering accurate R-peak detection.

Purpose of the Study:

• To develop and evaluate a deep learning-based approach for direct fetal HR determination from noninvasive fetal ECG signals.
• To overcome the limitations of traditional R-peak detection methods in fetal ECG analysis.
• To improve the accuracy and robustness of fetal HR extraction in clinical settings.

Main Methods:

• Utilized a combination of dilated inception convolutional neural networks (CNNs) and long short-term memory (LSTM) networks to analyze fetal ECG signals.
• Employed deep learning for direct fetal HR estimation, bypassing traditional R-peak detection.
• Integrated a CNN-based classifier to assess the reliability of the estimated fetal HR outcomes.

Main Results:

• Achieved a positive percent agreement of 97.3% on a labor dataset and 99.6% on the 2013 Physionet challenge set-A.
• Demonstrated superior performance compared to existing state-of-the-art algorithms.
• The integrated reliability classifier enhanced the robustness of the fetal HR estimation.

Conclusions:

• The proposed deep learning method offers a significant advancement in fetal HR extraction accuracy and reliability.
• This approach has the potential to enhance clinical practice for noninvasive fetal monitoring.
• Direct deep learning-based fetal HR determination presents a robust alternative to conventional signal processing techniques.