Seismo

Blood Pressure Monitoring using Built-in Smartphone Accelerometer and Camera

 
Seismo_concept.jpg

CHI'18 (Best Paper Honorable Mention Top 5% of Submitted Papers) : Edward Jay Wang, Junyi Zhu, Mohit Jain, Tien-Jui Lee, Elliot Saba, Lama Nachman, Shwetak Patel


Seismo is a BP measurement technique using the existing sensors on the smartphone. Seismo uses pulse transit time (PTT) -- the time taken by the heart's pulse to propagate between two arterial sites -- which is inversely related to BP. In particular, Seismo tracks the time when the blood is ejected from the heart as the aortic valve opens and when the pulse arrives at the fingertip. To perform this, Seismo relies on Seismocardiography (SCG), which uses the vibration caused by the movement of the blood and valve activities as the heart beats, allowing for accurate measurement of aortic valve opening time. The SCG is captured using the phone's accelerometer pressed against the chest. In this position, the user holds the phone with their finger covering the camera, which then captures the photoplethysmogram (PPG) at the finger, thus measuring the pulse as it arrives. This technique conveniently captures both the proximal (close to the heart) and distal (away from the heart) timing all from one device, without the need for any supplemental hardware. Additionally, PTT-based techniques can measure beat-to-beat BP, thus it can more reliably measure short-term BP changes (such as post-exercise), which are difficult to measure using cuff-based devices. One major distinction between Seismo and previous solutions that enable smartphone blood pressure tracking without additional hardware is the use of accelerometer to capture SCG as a proximal timing. Other work has mainly focused on using the sound created by the heart, otherwise known as phonocardiography (PCG). However, the fundamental limitation of using PCG as a proximal timing is that the sound being captured is actually created by the closing of the heart valves rather than the opening, thus not an ideal reference for when the blood is actually ejected. 

seismo_ptt.jpg
seismo_sync.png

Here we show a sample SCG and PPG signal as measured by our Seismo system on a Google Pixel phone. The ECG is captured using an Alivecor ECG monitor for reference. In order to properly synchronize our system, we used a work around by using a triple audio beep sequence to vibrate the phone after each measurement in order for the camera feed and the accelerometer feed to time synchronize.  After the data feeds are synchronized, we capture the timing information from the PPG and SCG signals to measure the PTT. The maximal point of acceleration of the PPG (APG) is compared with the aortic valve opening (AO)point of the SCG. To convert PTT to BP, an individualized calibration of PTT to BP is generated based on reference recording with a blood pressure cuff.

seismo_algorithm.jpg

System Validation

To validate the performance of Seismo, we recruited seven participants to perform a series of stationary bike tasks. All participants were recruited from within the author's institution, with an age range from 26 to 66 years. Participants are asked to ride a stationary bike five times, for 2-minutes each. The bike riding intensity varies from 50%, 75%, 100%, 75%, and 50% of the participants' maximum speed. The participant's blood pressure is first taken at rest. Then the participant performs the first 2-minute bike ride at 50% of the maximum speed.
Just after the bike ride, their blood pressure is taken again. The participant repeats this until they finish all five rounds of biking. At the end, the participant is asked to rest for two minutes, and their blood pressure is measured again after the rest. In total, the blood pressure and PTT are measured 7 times (once before biking, 5 times after biking at each intensity, and once after two minutes of rest). 

For each participant, the first days' PTT measurement by Seismo and the blood pressure reference measured by the cuff are used for calibration. With this calibration, the subject's blood pressures for the rest of the sessions are estimated from the PTTs measured and compared against the reference cuff-based BP measurements. The calibration performed in our work is limited to the mechanism of BP perturbation caused by exercise, and our work does not include other vasomotor tone modulation induced by cold pressor or stress.

seismo_result_table.PNG
seismo_result_regression.jpg
seismo_result_cdf.jpg

Our results demonstrates comparable performance to other PTT based measurement systems that uses blood pressure cuffs as a validation. A potential improvement to our correlation could be possible with the use of more continuous blood pressure monitoring techniques such as the Finapres digital artery measurement system. Our performance suggests that using just the built-in sensors and adhoc synchronization method, it is possible to capture PTT across multiple days for consistent blood pressure estimation. We believe that our work contributes to the field of mobile health sensing and pulse transit time monitoring by introducing a simple and low cost sensing solution to capture reliable PTT. 

Incorporating Smartphone Blood Pressure Monitoring to Personal Health Tracking Applications

By only using the built-in sensors of the smartphone, Seismo can monitor blood pressure without any modifications to a smartphone with a camera and accelerometer. We envision such a smartphone based PTT measurement can be useful in a variety of scenarios depending on the availability of resources. 

Supplement to At-Home Blood Pressure Cuff

A smartphone solution to measure blood pressure can be a good supplement for patients who are already monitoring their blood pressure with the cuff. Although a person can easily keep a cuff around at home and monitor once or twice a day to gain knowledge of their general blood pressure trends, it is hard to carry a cuff around as they go out of their house or when they travel. Having a smartphone based method can certainly help in such a scenario. Moreover, Seismo can be use for blood pressure monitoring for pre- and post-exercise. Here blood pressure change pre- and post-workout can even act as an additional measure of workout quality beyond heart-rate tracking. Using Seismo, the user can also track their blood pressure recovery time after a cardio workout. One of the advantage to a pulse transit time based blood pressure monitoring is the ability to track blood pressure continuously, something that can not be tracked even if the user had brought a cuff with them to the gym. 

Cardiovascular Risk Screening

In order for PTT-based blood pressure monitor to measure blood pressure accurately, the user needs to calibrate it. As such, for users who do not already have access to a cuff, Seismo would not be able to readily translate the measured PTT to blood pressure. However, in this case, Seismo can still offer insight into the person's cardiovascular health. The relationship between PTT and blood pressure also depends on arterial stiffness. Arterial stiffening occurs naturally as people age, however excessive stiffening usually suggests some form of health risks, such as plaque build-up. PTT has been shown to be a potent index for monitoring arterial stiffness by taking into account arm length, gender, and age. In well-equipped hospitals, an ECG and PPG system is often used to monitor PTT to screen for arterial stiffening. However, this is not available in low resource areas. Using Seismo, reliable PTT measurements can be made available through phones that are already available in these areas to enable arterial stiffness screening.