Performance of a QRS detector on self-collected database using a handheld two-electrode ECG

This paper presents a development of a handheld device for single-channel ambulatory ECG measurements. Our ECG device is designed for easy-to-use. ECG signal will be recorded for 1 minute by placing both thumbs on two dry electrodes and we have used this device to collect ECG signals from 50 volunteers. An ECG database is created from the collected ECG signals with self-annotated locations of QRS complex. This process is indispensable for developing the algorithm and for testing the performance of ECG analysis. The self-collected database has been used to evaluate the performance of a simple slope-based QRS detection algorithm, which had a relatively high accuracy on MIT-BIH Arrhythmia Database from our previous work without preprocessing filtering. As a result, the performance of the same QRS detector on our 50 short-term ECG records has an excellent sensitivity of 99.74%, but a moderate positive predictivity of 94.25% because of high level of EMG and movement artefacts from thumbs. After applying preprocessing filter (band pass 9-20 Hz), the positive predictivity is significantly improved by about 4%. This work showed that for the development of QRS detector for a handheld two-electrode ECG device in our case, the detection algorithm should be optimized by using a device-specific database because of some particular noise characteristics of ECG signals recorded from the handheld ECG device.