Comparison of general purpose ECG analyzers in patients with an implanted CRT device

Many QRS boundaries detectors have been previously proposed and have showed good performance with conventional ECG databases. In this study, two of such algorithms, one based on empirical mode decomposition and one based on the curve-length transformation were tested against a database obtained from patients with an implanted CRT device. This is a case of interest since increased QRS duration is a crucial clinical criteria for CRT implantation and also because the alterations on QRS morphology and stimulation artefacts present in these cases may mislead general purpose algorithms. Results showed that both methods were incapable of accurately measure QRS duration, showing significant differences (p<;0,05) respect to measures obtained from expert annotation. Average errors of 17±9 and -22±19ms in the estimation of the duration of the QRS complex were obtained with EMD and CLT methods respectively. These results highlight the necessity of validating general purpose algorithms before using them in more concrete scenarios.