Introduction The detection of the QRS complex is an essential step in the automated electrocardiogram (ECG) analysis algorithm. Considering the multi-lead analysis of QRS complex would be resistant to presence of noise, in this work we aimed to develop a QRS complex detection method on the 12-lead ECG signals.
Method The proposed method consists of two steps: detection of QRS complex on every single-lead, and multi-lead fusion of all detected QRS complexes. In the QRS detection, after filtering the ECG signals, the auxiliary signal was produced by calcualting the derivation, square and integration of the filtered signal successively. Finally, the QRS complex was determined by thresholding the amplitude of the auxiliary signal. If no QRS complex was detected during a specific interval, the maximal peak between the given thresholds was treated as the QRS complex. In the stage of multi-lead fusion, the fusion window of 150ms was used to analyze the QRS complexes. Firstly, the detected QRS complexes on all leads within the same fusion window were clustered a QRS complex candidate. Secondly, the QRS complex candidates were identified and the false positives were removed.
Result The proposed method was evaluated on the St.Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database. This database contains 75 12-lead ECG recordings. These records are 30 minutes long, sampled at 257Hz. There are over 175000 beat annotation in the reference annotation files. The mean sensitivity and specificity of the QRS detection on the 12 single-lead were 97.6% and 98.4% respectively. After the multi-lead fusion, the sensitivity and precision were increased to 99.92% and 99.94% respectively.
Conclusion Experimental result shows the proposed method can produce an accurate QRS complex detection on the 12-lead ECG signals, which would be advantageous to implement in the clinic practices.