Preliminary Results for Estimating Pulse Transit Time Using Seismocardiogram

Pulse transit time (PTT) is the time it takes a pulse wave to travel between two arterial sites. The speed of this pulse is proportional to blood pressure and has been proposed to be used as an unobtrusive method of blood pressure measurement [1]. In such a measurement, one needs to take into account proximal and also distal measurements of the PTT. Pulseplethysmogram (PPG) has been proposed to measure the distal pulse, and electrocardiogram's (ECG) QRS wave has been used as an indicator of the proximal pulse. However, research has shown that using ECG as the proximal pulse indicator will cause errors in estimating the blood pressure, as it also includes the isovolumic contraction period [2].In this work and for the first time, seismocardiogram (SCG) is introduced as the proximal pulse for measuring PTT [3]. SCG is the low frequency chest accelerations recorded by accelerometers. There is a certain point on SCG signal that corresponds to aortic valve opening (AO). We have suggested using this point as the reference for calculation of PTT. However, locating AO on SCG is not a trivial task and ECG has always been used for this purpose [4]. In this work, we have also proposed a novel method that uses the distal pulse in order to find the AO on the SCG signal. The rise time of the same pulse is used to calculate the end of PTT. Thus, all that would be needed for estimation of PTT is an accelerometer on the chest and a pulse measuring device on the finger.The data were recorded from ten subjects undergoing lower body negative pressure (LBNP) at Simon Fraser University, Aerospace Physiology Lab. The study was conducted under an ethics approval from Simon Fraser University, and informed consent forms were signed by all participants.Subjects were lying supine inside the LBNP box and 5 min of data were recorded from them when they were relaxed and at rest. The pressure inside the box was then gradually reduced to −30 mm Hg. The SCG was recorded using a B&K accelerometer, and the distal pulse was measured using a Portapress device both at sampling rate of 1000 Hz. The SCG data were hand annotated as described in Ref. [4].The idea behind the algorithm can be seen in Fig. 1. The first step is processing of the finger pulse signal and detecting the rise time of the pulse.After detecting the rise time of the finger pulse, the AO point of the same cardiac cycle was detected. PTT was calculated as the time difference between the above timings. All processing steps were implemented in matlab. The overall block diagram of step by step processing of this algorithm is shown in Fig. 2.The signal detrending step was performed using an infinite impulse response (IIR) high pass filter with following transfer function:H(z)=1+a21-z-11-az-1where α = 0.99.After signal detrending, variable thresholding steps were performed to remove the parts of signal that cannot be a candidate for peaks. Apart from varying amplitude based thresholding, a distance based thresholding was also performed, to make sure we do not detect two peaks that are too close to each other since having two peaks in very close proximity could be due to noise. After thresholding, the location of peaks in a pulse signal was obtained by computing the zero crossing of derivatives of the signal. Details of thresholding and peak detection scheme are discussed in Ref. [5].To preprocess a SCG signal, a combination of detrending and smoothing filter was used. For detrending above mentioned IIR filter was used while for smoothing one-dimensional median filter was used.After processing all the steps of the algorithm as discussed in Fig. 2, the peak locations in the pulse signal and AO locations in a SCG signal were obtained. Figure 3 shows a part of SCG signal and AO locations are marked with red circles.After determining peak locations in the pulse signal and AO locations in the SCG signal, PTT was computed as a time difference between the rise time of pulse signal and the AO time. For the ten subjects mentioned previously, the PTT was measured at rest and at the LBNP of −30 mm Hg. The results can be seen in Table 1.From Table 1, it is noticed that PTT has increased from rest to the LBNP pressure of −30 mm Hg. This confirms our expectations, as we do know that LBNP decreases blood pressure. A decrease in blood pressure naturally increases PTT as observed in Table 1.The preliminary results presented in this paper suggest the possibility of extracting PTT using a system including only two sensors, one pulse sensor at the finger and one accelerometer on the chest. This simple system once calibrated can provide a portable blood pressure monitoring device.In this study, for the finger pulse, we used the pressure signals derived from a Portapress device. The rise time of the pulse detected using this device is very close to the rise time detected from signal recorded using PPG with a delay of less than 20 ms. We believe that the same algorithm will work equally well with a standard PPG transducer. Our next study will use the standard PPG to determine pulse rise time.We would like to thank Dr. Andrew P. Blaber for his assistance in data acquisition.