A multilead wavelet-based ECG delineator based on the RMS signal

In this work a multilead (ML) algorithm for ECG delineation was validated against the CSE and the PTB databases (DB). This algorithm is based on a previously validated single-lead (SL) algorithm, which is also used as reference in this work. For both DB a set of manual reference annotations made by experts is available. The algorithm performance was assessed by the measurement of QRS onset (QRS_on) and T wave offset (T_off) in both DB. The differences (meanplusmnsd in ms) between automatic and experts annotations for CSEDB was for QRS_on 6.7 plusmn 7.2 and for T_off 0.1 plusmn 10.2. For the PTBDB, results were for QRS_on -5.7 plusmn 6.9 and for T_off 11.3 plusmn 24.4. Also standard deviation of QT interval in 10 consecutive beats was calculated as a measurement of stability in the PTBDB, resulting in 6.7 plusmn 14.3. In conclusion this ML approach is more recommendable than SL in applications where measurement stability is mandatory and when absolute value of QT is required as a lead-independent measure.