Departamento de Tecnologías de la Información y las Comunicaciones, Facultad de Ingeniería, Universidad Santiago de Cali, Santiago de Cali, Colombia.
Physiol Meas. 2020 Apr 16;41(3):035001. doi: 10.1088/1361-6579/ab7878.
One of the biggest obstacles to reliable pulse rate variability (PRV) analysis is the erroneous detection of photoplethysmographic (PPG) pulses. Among all the disturbances that may hinder pulse detection, the ripples appearing at the smooth segments of the PPG signal can become a serious problem when the amplitude of the signal decreases considerably.
To present a low-complexity PPG pulse detection method for reliable PRV estimation under conditions in which a sudden decrease in the amplitude of the PPG signal can be expected.
2-min ECG and PPG data (sampling rate at 500 Hz) were obtained from thirty healthy subjects, who were asked to take a deep inspiration to provoke a sudden amplitude decrease (SAD) of the PPG signal. After introducing a new parameter denoted as C, through which it is possible to jump over the ripples hindering the accurate detection of the systolic peaks, 500 Hz-sampled PPG recordings were down-sampled (400, 300, 200 and 100 Hz) to investigate the effect of the sampling rate on pulse detection. For ECG recordings, automatic R-peak detection was performed by the Pan and Tompkins (PT) algorithm, whereas PPG pulse detection was performed by the well-known maximum of the first derivative (M1D) and the proposed method, once the C-value for best detection results on 500 Hz-sampled PPG recordings was found. The agreement between heart rate variability (HRV) and PRVs estimated from each pulse detection method was assessed and the correlation between HRV and PRV-derived indexes was computed for comparison.
The proposed method can perform well on PPG-SAD segments, provided that the proper value of the parameter C is used. Moreover, a good agreement between HRV and PRV series, as well as lower relative errors and higher correlation coefficients between HRV and PRV indexes, were achieved by the proposed pulse detection method during SADs.
Results show that the proposed method can dynamically adapt to circumstances in which a decrease in the amplitude of the PPG signal can be expected, providing continuous systolic peak detection and reliable PRV estimation under those conditions. However, more extensive testing under a wide range of conditions is needed to perform a more rigorous validation.
可靠的脉搏率变异（PRV）分析面临的最大障碍之一是光体积描记图（PPG）脉搏的错误检测。在所有可能阻碍脉搏检测的干扰中，当 PPG 信号的幅度显著下降时，出现在 PPG 信号平滑段的波纹可能会成为一个严重的问题。
提出一种低复杂度的 PPG 脉搏检测方法，用于在预期 PPG 信号幅度突然下降的情况下进行可靠的 PRV 估计。
从 30 位健康受试者中获得 2 分钟的心电图和 PPG 数据（采样率为 500Hz），要求他们进行深呼吸以引起 PPG 信号的幅度突然下降（SAD）。引入一个新参数 C 后，可以跳过阻碍收缩峰准确检测的波纹，对 500Hz 采样的 PPG 记录进行下采样（400、300、200 和 100Hz），以研究采样率对脉搏检测的影响。对于心电图记录，采用 Pan 和 Tompkins（PT）算法进行自动 R 波检测，而 PPG 脉搏检测则采用著名的一阶导数最大值（M1D）和所提出的方法进行，一旦在 500Hz 采样的 PPG 记录上找到最佳检测结果的 C 值。评估了从每种脉搏检测方法估计的心率变异性（HRV）和 PRV 之间的一致性，并计算了 HRV 和 PRV 衍生指标之间的相关性进行比较。
在所提出的方法中，可以在 PPG-SAD 段中很好地工作，只要使用适当的 C 值。此外，在 SAD 期间，所提出的脉搏检测方法可实现 HRV 和 PRV 系列之间的良好一致性，以及 HRV 和 PRV 指标之间的较低相对误差和更高的相关系数。
结果表明，所提出的方法可以动态适应预期 PPG 信号幅度下降的情况，在这些情况下提供连续的收缩峰检测和可靠的 PRV 估计。然而，需要在更广泛的条件下进行更多的测试，以进行更严格的验证。
