Empirical mode decomposition based real-time blood pressure delineation and quality assessment

Noise and artifacts compromise the fidelity of arterial blood pressure (ABP) waveform, resulting in unreliable parameter estimation and high false alarm rate. This study proposes empirical mode decomposition (EMD) based algorithm that integrates real-time ABP delineation and quality assessment. The data-driven decomposition ensures insensitivity to inter-subject variability. Adaptive EMD detrending procedure, combined with short-term memory parameter tracking, solves the problem of waveform morphology changes. Delineation algorithm was tested on a publicly available expert annotated database consisting of 13079 beats. Algorithm showed sensitivity of 99.85% and positive predictivity of 99.76% for systolic blood pressure, and sensitivity of 99.89% and positive predictivity of 99.82% for onset detection. Six artifact types were simulated on available dataset and successfully detected using the algorithm. The frequency band separation of intrinsic mode functions (IMFs) proved as relevant feature for selective artifacts detection on the most affected IMFs.