Signal Processing Framework for the Detection of Ventricular Ectopic Beat Episodes.

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Long-term electrocardiogram (ECG) analysis for detecting cardiac abnormalities like ventricular ectopic beats (VEBs) is time-consuming due to manual cross-checking.
• Existing Holter monitor technology captures ECG data but relies on manual interpretation, creating a bottleneck in diagnosis.

Purpose of the Study:

• To develop and validate an effective automatic cardiac episode detection technique to reduce the manual burden in analyzing long-term ECG signals.
• To present a signal processing framework for detecting ventricular ectopic beat (VEB) episodes across different databases.

Main Methods:

• ECG signals were preprocessed, and features like RR intervals and QRS complex parameters were extracted.
• Four classification models, including Support Vector Machine, k-means nearest neighbor, nearest mean classifier, and Nearest RMS (NRMS) classifier, were employed.
• Models were trained and tested on both open-source and proprietary ECG databases for VEB detection.

Main Results:

• The Nearest RMS (NRMS) classifier demonstrated superior performance compared to other models.
• NRMS achieved high accuracy (98.68% on open-source, 99.97% on proprietary) and F1-scores (94.12% on open-source, 94.54% on proprietary) for VEB detection.
• Excellent recall rates (100% and 98.62%) and specificity (98.53% and 99.98%) were observed across both databases.

Conclusions:

• The proposed signal processing framework effectively detects ventricular ectopic beat (VEB) episodes in long-term ECG signals.
• The developed automated system shows significant potential for integration into clinical practice for efficient cardiac diagnosis.
• The NRMS classifier offers a robust and accurate solution for automated VEB detection, reducing physician workload.