RESUMO
BACKGROUND: Electrocardiogram variations (ECG) due to body position changes and electrode placements are common problems of continuous ST-T monitoring. Body position changes may cause QRS and ST-T changes and trigger false alarms. Placement of arm and leg electrodes in a coronary care unit environment is usually near the thorax instead of standard position at the wrists and ankles. This may affect the limb leads and complicate diagnostic interpretation. The purpose of this study was to assess the effects of these sources of ECG variation and to correct for them. Continuous 12-lead ECG recordings were obtained from 160 patients admitted to the coronary care unit. Each patient underwent a body position test (supine, left-lateral, and upright position). Scalar and spatial approaches were investigated for reconstruction of the ECG in supine position. The scalar approach uses linear regression. The spatial approach transforms the ECG into a derived vestorcardiogram. The spatial QRS-loop is then rotated and scaled to match the vector loop in supine position and transformed back to a 12-lead ECG. MATERIALS AND METHODS: To assess the effect of electrode placement, monitoring and standard limb leads were simultaneously recorded in a group of 80 patients. To map the monitoring leads to standard leads, general and patient-specific reconstruction coefficients were derived by linear regression from half of the patients and tested on the other half. Similarity between the reference and reconstructed ECGs was measured by correlation, similarity coefficient [(SC=1-RMS(residual error)/RMS(signal)], and difference in frontal QRS-Axis. RESULTS AND CONCLUSION: Only 14% (23 of 160) of the patients showed marked ECG changes (ST elevations, QRS-axis shifts, T-wave inversions). The scalar method (median correlation > 0.994, SC > 0.902, QRS axis difference 0 degrees) performed better than spatial (median correlation 0.946, SC > 0.792, QRS axis difference 0 degrees). Monitoring leads can be mapped to standard limb leads in good to excellent approximaiton. General reconstruction (median correlation 0.993 and SC 0.764) performed slightly worse than patient-specific reconstruction (median correlation 0.997 and SC 0.908).