Dynamic ECG Signal Quality Evaluation Based on the Generalized bSQI Index

Abstract

With the fast development of wearable electrocardiogram (ECG) monitoring, real-time and dynamic signal quality assessment (SQA) become an imperious demand. Thus, many signal quality indices (SQIs) have been developed in the past several years. bSQI is a typical SQI defined from two common QRS complex detectors (`ep_limited' and `wqrs'). However, in actual application, bSQI heavily relies on the QRS complex detectors used. Therefore, if using different combination of QRS detectors can improve the performance of SQA needs to be explored. In this paper, we utilized up to ten QRS detectors to re-define bSQI from the combination of any two QRS detectors to test which combination outputs the highest performance. Then, we generalized the two QRS detector-based bSQI to multiple QRS detector-based bSQI (i.e., GbSQI), to systematically test the effects of type and number of QRS detectors on SQA performances. The results showed that for the single GbSQI feature-based classifier, the combination of six QRS detectors reported the highest classification accuracy with a mAcc of 94.03%. For the multiple GbSQI feature-based classifier, the combination of four QRS detectors showed the best classification accuracy with a mAcc of 94.76%. As a conclusion, we recommended using U3, UNSW, DOM, and OKB detectors for calculating GbSQI for the wearable ECG monitoring application.

Keywords

• Electrocardiogram (ECG)
• signal quality assessment (SQA)
• signal quality index (SQI)
• QRS detection