Limitations on the Re-use of patient specific coefficients for 12-lead ECG reconstruction

Abstract

Patient specific coefficients for reconstructing missing precordial leads (patient-specific single-use or PSS) show good performance but require a 12-lead ECG to start monitoring. A more convenient approach is either the use of population based coefficients (POP) or patient specific coefficients from an old 12-lead ECG (patient-specific multi-use or PSM). We used a data set of 1493 resting 12-lead ECGs from 224 patients. Waveform comparisons were made between recorded 12-lead and reconstructed cases using RMS difference. Three cases were compared, PSS, PSM and POP. Median RMS reconstruction error in the ST-T region was 16, 46 and 40 muV for lead configuration V1/V4 in the PSS, PSM and POP cases respectively. For the V2/V5 configuration, median ST-T RMS error was 8, 40 and 41 muV. The RMS error for the PSS case was lower and significantly better by paired T-test. The difference between the two more convenient use-models, PSM and POP, was not significant. Population based coefficients are preferred over patient-specific coefficients if the single-use use-model cannot be followed.