Dynamic ECG signal quality evaluation based on persistent homology and GoogLeNet method.

Area of Science:

• Biomedical Engineering
• Data Science
• Cardiology

Background:

• Continuous Electrocardiograph (ECG) monitoring is crucial for early cardiovascular disease detection.
• Wearable ECG devices are increasingly used but susceptible to signal contamination.
• Reliable ECG signal quality assessment is essential for accurate diagnosis.

Purpose of the Study:

• To propose a novel quality assessment method for wearable dynamic ECG signals.
• To leverage topological data analysis (TDA) with persistent homology for ECG signal evaluation.
• To enhance the reliability of remote cardiovascular monitoring systems.

Main Methods:

• Constructing point clouds from ECG signals.
• Generating complex sequences and persistent barcodes using TDA.
• Training a GoogLeNet classification model with transfer learning and 10-fold cross-validation.
• Validating the method on 12-lead and single-lead ECG datasets.

Main Results:

• The proposed TDA-based method achieved high classification performance.
• For 12-lead ECGs Dataset: mAcc = 98.04%, F1 = 98.40%, Se = 97.15%, Sp = 98.93%.
• For single-lead ECGs Dataset: mAcc = 98.55%, F1 = 98.62%, Se = 98.37%, Sp = 98.85%.

Conclusions:

• The TDA with persistent homology method offers a robust approach for wearable dynamic ECG signal quality assessment.
• This method outperforms traditional techniques based on waveform and time-frequency characteristics.
• The findings support the use of advanced data analysis for improving the accuracy of wearable health monitoring.