P wave delineation using spatially projected leads from wavelet transform loops

A previously proposed automatic delineation strategy for multilead (ML) location of wave boundaries is now extended to P wave boundaries. The method obtains a transformed lead by projecting the wavelet transform spatial loop into a direction that optimizes the SNR. The performance was compared with single lead delineation (SL) and with the global marks obtained by post-processing rules (SLR) calculating the Sensitivity, mean and standard deviation [S(%), m ± s (ms)] respectively for onset| end location errors. Validation is performed over CSE database files obtaining, using 3D loops [100, -10.2 ± 15.0] | [100, 7.8 ± 14.2], and using 2D loops [96, -2.4 ± 7.4] | [96, 5.9 ± 7.0]. SLR achieved [85, 2.1 ± 5.6] | [89, 1.8 ± 6.7], with SL producing always higher s values in one or both boundaries. We conclude that ML strategy is appropriate for P wave delineation, with higher S that SLR, and 2D loops are sufficient, allowing a more efficient processing when compared to SLR.