Computer package generating non-invasive atrial electrograms: Detection and subtraction of QRS and T waves

Computerized systems to localize specific patterns appearing in ECG signal, such as the QRS complex and T wave usually utilize predefined templates. Where the signal is not clean, such as during atrial fibrillation (AF), analysis based on self-similarities of the ECG signal appears to be more stable. The authors describe a software package which uses self-similarity techniques to allow the subtraction of ventricular signals during AF. The dataset used in this study was Holter ECG recorded in 23 patients (11 male, age 61.1±8.7 years) with chronic AF using an Altair 6500 digital Holter (Burdick Inc., 3 channels, 1 kHz sampling). Self-similarities between corresponding QRS and corresponding T windows were assessed using Spearman correlation coefficients, utilising the commercial software derived fiducial point but with shifted windows to overcome possible minor imprecision thereof. A linear order algorithm successfully subtracted 87% of QRS's and 67% of T waves.