An investigation of the impact of artifact detection on heart rate determination from unsupervised electrocardiogram recordings

A variation of an existing technique for movement artifact detection in single lead ECG signals acquired in the unsupervised telehealth environment is examined. The impact on heart rate (HR) estimation is investigated using this artifact detection technique to remove noisy sections of signal. The estimated artifact masking and HR values are compared to a gold standard scoring, performed by consensus of an expert panel. The employment of the proposed artifact detection scheme shows an improvement in the estimated values of HR; the error in the estimated HR, from 126 of 192 signals, was less than ??0.5 BPM; compared to only 67 of 212 signals using no artifact detection; the estimation bias was reduced from an underestimation of -1.33 BPM to -0.63 BPM; the standard deviation of the error was reduced from 4.81 BPM to 3.58 BPM. The results indicate that the automated interpretation of inherently noisy ECG recordings, from the telehealth environment, becomes a feasible proposition when ECG signal quality indicators are leveraged.