Benchmarking open-source algorithms for QRS detection and RRI editing in textile electrocardiography

Area of Science:

• Biomedical Engineering
• Cardiovascular Monitoring
• Wearable Technology

Background:

• Textile electrocardiography (ECG) sensors in T-shirts offer potential for posthospital care monitoring.
• Heart rate variability (HRV) analysis, crucial for clinical parameters, relies on accurate detection of heartbeats (QRS complex) and their intervals (RR intervals - RRI).
• A lack of comprehensive benchmarking for QRS detectors and RRI editing algorithms in textile ECG hinders clinical application.

Purpose of the Study:

• To evaluate the performance of open-source QRS detectors using textile ECG data.
• To assess the robustness of these detectors against noise and subject activity.
• To test the efficacy of an RRI editing algorithm on detected heartbeats.

Main Methods:

• Evaluated 12 open-source QRS detectors on a public dataset from 13 healthy volunteers wearing a T-shirt.
• Introduced additional noise using the MIT-BIH Noise Stress Test Database.
• Applied a state-of-the-art RRI editing method to the output of the best-performing QRS detector.

Main Results:

• The xqrs, kalidas2017, and Rodrigues2021 algorithms demonstrated superior performance (sensitivity/precision) on noisy textile ECG data.
• Detector performance remained largely unaffected by additional noise or diverse subject activities.
• Other QRS detectors exhibited insufficient sensitivity and/or precision for reliable textile ECG analysis.
• The RRI editing method successfully rejected false positive detections in 25% of subjects when applied to xqrs results.

Conclusions:

• Specific QRS detectors (xqrs, kalidas2017, Rodrigues2021) are suitable for HRV analysis using textile ECG.
• These detectors show resilience to common challenges in wearable monitoring, including noise and movement.
• RRI editing can further improve the reliability of HRV analysis from textile ECG data.