A deep learning approach for fetal QRS complex detection.

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence

Background:

• Non-invasive fetal electrocardiography (NI-FECG) offers potential for diagnosing fetal conditions.
• Accurate detection of fetal QRS complexes is crucial for clinical analysis.
• Maternal ECG (MECG) interference complicates fetal signal interpretation.

Purpose of the Study:

• To develop and evaluate a deep learning approach for fetal QRS complex detection using raw NI-FECG signals.
• To assess the feasibility of detecting fetal QRS complexes without maternal ECG cancellation.
• To investigate the impact of signal quality assessment and activation functions on detection performance.

Main Methods:

• A convolutional neural network (CNN) model was employed for fetal QRS complex detection.
• Raw NI-FECG signal features were normalized before input to the CNN classifier.
• Sample entropy was used for signal quality assessment, with poor-quality signals excluded.
• Performance was evaluated using precision, recall, and F-measure metrics.

Main Results:

• The proposed deep learning method achieved high precision (75.33%), recall (80.54%), and F-measure (77.85%).
• Performance was superior to traditional methods like KNN, naive Bayes, and SVM.
• The Relu activation function demonstrated better performance than Sigmoid and Tanh.

Conclusions:

• Reliable fetal QRS complex detection is achievable using raw NI-FECG signals with a deep learning approach.
• Maternal ECG cancellation is not required for effective fetal QRS detection.
• Signal quality assessment significantly improves classification performance.