Automated detection of the onset and systolic peak in the pulse wave using Hilbert transform. (July 2015)

• Record Type: Journal Article
• Authors: Ricardo Ferro, B.T.; Ramírez Aguilera, A.; Fernández de la Vara Prieto, R.R.
• Is Part Of: Biomedical signal processing and control. Volume 20(2015)
• Page Start: 78
• Page End: 84
• Publication Date: 2015-07
• Subjects: Onset -- Systolic peak -- Photoplethysmography -- Hilbert transform; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2015.04.009
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Hilbert transform method can be used for detecting both the onset and systolic peak in pulse wave signals. It achieves good performance and precision when compared to expert annotation. It is robust when noise and interference are present in pulse wave and also. Shows a low detection error rate. It is computationally simple. Abstract: Pulse transit time (PTT) and pulse wave velocity (PWV) are the markers most widely used to evaluate the vascular effects of aging, hypertension, arterial stiffness and atherosclerosis. To calculate these markers it is necessary to determine the location of the onset and systolic peak of the arterial pulse wave (APW). In this paper, a method employed for electrocardiography (ECG) R peak detection, with a slight modification, is applied for both the onset and systolic peak detections in APW. The method employs Shannon energy envelope (SEE) estimator, Hilbert transform (HT) and moving average (MA) filter. The minimum value and the positive zero-crossing points of the odd-symmetry function of the HT correspond to the locations of the onset and systolic peak respectively. The algorithm was evaluated using expert's annotations, with 10 records of 5 min length and different signal-to-noise ratios (15, 12 and 9 dB) and achieved a good performance and precision. When compared to, expert's annotation, the algorithm detected these fiducial points with average sensitivity, positive predictivity and accuracy of 100% and presented errors less than 10 ms. In APW signals contaminated with noise in both cases the relative error is less than 2% respect to pulse wave periods of 800 ms. The performance of algorithm was compared with both foot approximation and adaptive threshold methods and the results show that the algorithm outperforms theses reported methods with respect to manuals annotation. The results are promising, suggesting that the method provides a simple but accurate onset and systolic peak detection and can be used in the measurement of pulse transit time, pulse wave velocity and pulse rate variability.