On the development of an efficient, low-complexity and highly reproducible method for systolic peak detection. (July 2021)

• Record Type: Journal Article
• Authors: Argüello Prada, Erick Javier; Bravo Gallego, Carlos Alberto; Castillo García, Javier Ferney
• Is Part Of: Biomedical signal processing and control. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: Photoplethysmography -- Systolic peak detection -- Computational cost -- Wearable devices -- Real-time monitoring -- Artifact corruption; 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.2021.102606
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An efficient, computationally expedient, and highly reproducible method for systolic peak detection was developed. Given its low computational requirements, the proposed method is suitable for real-time applications. Some time-domain strategies could be useful to improve the performance of very simple systolic peak detection methods. Highly reproducible peak detection methods are more likely to be implemented and improved. Abstract: Background and Objectives: Numerous methods for systolic peak detection on PPG signals have been reported in the literature. However, such approaches can hardly be replicated and implemented in wearable applications because of their complicated methodologies and high computational requirements. In an attempt to address this issue, an efficient, low-complexity, and highly reproducible method for real-time systolic peak detection is proposed. Methods: The method calculates the difference between the value of the current PPG sample and the maximum computed over a certain window that includes the current value. If such difference is negative, then the previous PPG value is labeled as a peak provided that the number of times that the difference is equal to zero reaches or exceeds the window size. To overcome some of the disturbances usually accompanying PPG signals, some well-documented, time-domain strategies were included. The performance of the method was assessed off- and online by using free-available and locally-acquired data sets. Results: The proposed method can perform faster than several other peak detection methods previously reported in the literature, including some that could not be implemented due to their detection parameters are not available. However, the method performance may be affected by motion artifact corruption, especially when the sampling rate increases. Conclusion: Given its low computational requirements, novel techniques for artifact detection and removal could be added to our method to improve its robustness. More accurate comparisons between results yielded by this and several other studies could be performed as long as detection parameters were properly reported.