Accurate ECG R-peak detection for telemedicine

Abstract

Electrocardiograms (ECGs) are usually recorded in a clinical setting by medical professionals using twelve leads attached to the patient. Our industry partner has developed a single-lead ECG machine for use by patients at home. Patients can then send these readings to remote doctors. The goal of the machines is to make medical expertise more accessible, affordable, and convenient. The ECGs recorded by patients with a single-lead suffer greatly from baseline wandering and high frequency noises, as compared to ECGs recorded with twelve-leads in a clinical setting. Accurate R-peak detection is an important step in ECG analysis. A variety of methods have been proposed in the past against standard clinical twelve-lead ECG recordings. In this study, we propose a new R-peak detection algorithm for singlelead mobile ECG recordings. Our area-based approach is built on the understanding that QRS complexes are typically narrow and tall, resulting in large areas over the curve around these locations. Our algorithm is simple to implement, computationally efficient, and does not require any signal pre-processing. This conceptual simplicity is a quality that distinguishes our approach from existing solutions. We evaluated our algorithm against data collected by patients from single-lead portable devices, and yielded 99.4% precision and 99.4% recall. The MIT/BIT Arrhythmia Database of twelvelead clinical ECG recordings was also used to verify our algorithm. On this dataset we obtained a precision of 99.3% and recall of 98.6%.