Automatic detection of QRS complexes in ECG signals collected from patients after cardiac surgery

A novel automatic QRS detection algorithm that is based on a wavelet pre-filter and an adaptive threshold technique is presented. The algorithm utilizes a bi-orthogonal wavelet filter to de-noise the ECG signal. The QRS complexes are then identified by computing the first derivative of the signal and applying a set of adaptive thresholds that are not limited to a strict range. QRS complexes are identified in multiple ECG channels of a 5-lead configuration and an inter-channel comparison is performed to verify QRS locations. The algorithm was initially developed using ECG signals from Physionet website, but was later refined using ECG data collected from post cardiac surgery patients in the intensive care units. The proposed algorithm was able to detect QRS complexes with high sensitivity (>99%) and specificity (>99%) when compared to the algorithm used in Physionet ECG database. Additionally, the new algorithm can be implemented in real-time and can successfully detect QRS complexes for a wide variety of ECG shapes and characteristics often encountered in cardiac patients.