Deriving respiration from the electrocardiogram by serial comparison with statistical mean shape

A new algorithm to derive a surrogate respiration signal from a single-channel electrocardiogram (ECG) was developed. The generalized mean shape of QRS-complexes was estimated and then used as a template to measure scale variation interpreted as the respiration. Compared to four other ECG-derived respiration (EDR) algorithms, median normalized cross-correlations and median magnitude squared coherences between the new EDR and actual spirometer or chest belt respiration signals showed consistent top-of-the-line performance across different breathing rates and postures. The relative performance of the new EDR algorithm was especially favourable in controlled breathing at the 0.33 Hz respiratory rate.