Background: A vectorcardiography approach to electrocardiology contributes to the non-invasive assessment of electrical heterogeneity in the ventricles of the heart and to risk stratification for cardiac events including sudden cardiac death. The aim of this study was to develop an automatic method that identifies a representative QRST complex (QRSonset to Tend) from a Frank vectorcardiogram (VCG). This method should provide reliable measurements of morphological VCG parameters and signal when such measurements required manual scrutiny.
Methods: Frank VCG was recorded in a population-based sample of 1094 participants (550 women) 50-65 years old as part of the Swedish CArdioPulmonary bioImage Study (SCAPIS) pilot. Standardized supine rest allowing heart rate stabilization and adaptation of ventricular repolarization preceded a recording period lasting ≥5 minutes. In the Frank VCG a recording segment during steady-state conditions and with good signal quality was selected based on QRST variability. In this segment a representative signal-averaged QRST complex from cardiac cycles during 10s was selected. Twenty-eight morphological parameters were calculated including both conventional conduction intervals and VCG-derived parameters. The reliability and reproducibility of these parameters were evaluated when using completely automatic and automatic but manually edited annotation points.
Results: In 1080 participants (98.7%) our automatic method reliably selected a representative QRST complex where its instability measure effectively identified signal variability due to both external disturbances ("noise") and physiologic and pathophysiologic variability, such as e.g. sinus arrhythmia and atrial fibrillation. There were significant sex-related differences in 24 of 28 VCG parameters. Some VCG parameters were insensitive to the instability value, while others were moderately sensitive.
Conclusion: We developed an automatic process for identification of a signal-averaged QRST complex suitable for morphologic measurements which worked reliably in 99% of participants. This process is applicable for all non-invasive analyses of cardiac electrophysiology including risk stratification for cardiac death based on such measurements.