Comparative study of signal decomposition methods for enhancement of the accuracy of T-wave end localisation

The goal of this study was to compare several ECG signal decomposition methods in order to enhance the accuracy of T wave end localisation. PTB Diagnostic ECG Database comprising 549 recordings was used. The idea was to combine the 8 independent leads (I, II, V1,..., V6) of the standard 12-leads ECG into a single lead.The signal decomposition methods were applied to reconstruct the combined T wave in such a way as to obtain maximal accuracy of the automatic T-end localisation. Four signal decomposition methods were used: Dominant T-wave (DTW), Principal Component analysis (PCA), and two Spatial Vectors (SVI & SV2) synthesized from the orthogonal x, y, and z vectors. The T-ends were localized using the same previously published software program. The results were compared to a published 'gold standard' dataset of manually determined T ends. Mean and standard deviation of the difference between the automatic and manual T-end locations were calculated in [ms]: DTW=8.72+14.19; PCA=10.30+12.69; SVI= -8.14+14.53; SV2=8. 59+17.93. The best results (i.e. smallest standard deviation) were obtained by PCA followed by DTW. Compared to manual measurement, all signal decomposition methods except SVI moved the mean Tend location slightly to the right.