Robust derivative-based method to determine filtered QRS limits in high resolution electrocardiography

The accuracy of high resolution electrocardiogram (HRECG) as a risk stratification tool relies on QRS offset detection, which is subject to variation due to signal residual noise level (RNL). This study proposed a technique for improving the accuracy of QRS offset detection, based on successive signal derivation (DER) and compared to the standard method recommended by the American College of Cardiology (ACC). The control group was composed by 18 healthy volunteers with no cardiac disease and SMVT Group by 18 subjects with sustained monomorphic ventricular tachycardia. Mean QRS duration was compared by paired Student t-test between groups. Sensitivity, specificity and predictive values (PV) were compared by χ² test. For ACC and DER methods, the QRS offset was determined at a range of different RNLs simulated in the HRECG signal (from 0.2 to 0.8 μV), and the occurrence of linear correlation was tested (α <; 0.05). Mean QRS duration calculated by DER showed no differences when compared to blind visual detection by a specialist. Positive PV was higher in DER than ACC (p <; 0.05). Linear correlation between QRS offset and RNL was found in ACC method for both groups. The DER method showed higher accuracy and lower sensitivity to RNL than standard ACC for QRS offset detection.