A Deep Learning Architecture Using 3D Vectorcardiogram to Detect R-Peaks in ECG with Enhanced Precision

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Reliable QRS complex detection is crucial for automated electrocardiogram (ECG) analysis and medical device synchronization.
• Deep learning (DL) offers superior accuracy and generalization for ECG analysis compared to classical algorithms.
• 3D vectorcardiograms (VCG) provide a lead-independent framework for R-peak detection.

Purpose of the Study:

• To develop and validate a DL architecture for precise R-peak detection using 3D VCG.
• To evaluate the model's performance across multiple public ECG databases without pre- or post-processing.
• To demonstrate the potential for real-time inference on edge devices.

Main Methods:

• A deep learning architecture was trained and applied on 3D vectorcardiograms (VCG).
• No pre-processing or post-processing steps were required for the model.
• Experiments utilized four diverse public ECG databases for validation.

Main Results:

• Achieved high F1-scores of 99.80% (cross-validation) and 99.64% (cross-database).
• Demonstrated superior precision (≥99.88%) significantly reducing false detections.
• Maintained high recall (≥99.39%), indicating minimal missed R-peaks.

Conclusions:

• The proposed DL approach on 3D VCG offers enhanced precision for R-peak detection.
• This method significantly reduces false detections without compromising accuracy.
• Potential applications include real-time medical devices, particularly where high precision is critical, such as cardiac MRI.