Session P36.3

Reconstructed Precordial Lead Performance in a Realistic Clinical Scenario

RE Gregg*, S Babaeizadeh, DQ Feild, ED Helfenbein,
DJ Kelly, JM Lindauer, SH Zhou

Philips Healthcare
Andover, MA, USA

Background: Deriving 12-lead ECG from a reduced subset of leads improves patient comfort in arrhythmia and ST monitoring. Patient specific coefficients for reconstructing missing precordial leads show good performance but require a 12-lead ECG before monitoring. A more practical clinical scenario is either the use of population based coefficients or patient specific coefficients from a previous 12-lead ECG. In this study we analyze two configurations, (1) precordial leads V2, V3, V5 and V6 derived from V1/V4 and (2) precordial leads V1, V3, V4 and V6 derived from V2/V5.
Methods: We used a data set of 1493 resting 12-lead ECGs from 224 patients. Representative beats for each ECG were generated from the beat selection, QRS-onset and T-end fiducials as calculated by the Philips 12-lead algorithm. Waveform comparisons were made between recorded 12-lead and reconstructed cases using RMS difference averaged across the reconstructed leads. For the ideal scenario, patient specific derivation was performed for all ECGs. Population based coefficients were used for the population scenario. In the realistic scenario, the patient’s first ECG was used to generate patient specific coefficients for all subsequent ECGs. To facilitate paired T-test comparison between scenarios, ECGs used in the realistic scenario for patient specific coefficient derivation were excluded.
Results: Median RMS reconstruction error in the QRS region was 50, 199 and 206µV for configuration V1/V4 in the ideal, population and realistic scenarios respectively. For the V2/V5 configuration, median QRS RMS error was 40, 169 and 178µV. Of the two realistic scenarios, the population scenario showed lower error. The RMS error for the ideal case was the lowest and significantly better by paired T-test as expected. No other RMS error difference between the three scenarios was significant whether QRS or ST-T region.
Conclusion: The reconstruction error for the two realistic clinical scenarios are equivalent, population based coefficients or patient specific coefficients calculated from a previous ECG.

(Abstract Control Number: 15)