This paper proposes a smartphone-based method for measuring blood pressure (BP) using the oscillometric method. To measure BP, it is necessary to measure (1) the pressure applied to the artery and (2) the local blood volume change. This is accomplished by performing an oscillometric measurement at the finger’s digital artery, whereby a user presses down on the phone’s camera with steadily increasing force. The camera is used to capture the blood volume change using photoplethysmography. We devised a novel method for measuring the force applied by the finger without the use of specialized smartphone hardware with a technique called Vibrometric Force Estimation (VFE). The fundamental concept of VFE relies on a phenomenon where a vibrating object is dampened when an external force is applied on to it. This phenomenon can be recreated using the phone’s own vibration motor and with powerthe resulting damped vibration measured using the Inertial Measurement Unit (IMU). A cross device reliability study with three smartphones of different manufacturers, shape, and prices results in similar force estimation performance across all smartphone models. In an N = 24 validation study of the BP measurement, the smartphone technique achieves a MAE of 9·35 mmHg and 7·94 mmHg of systolic and diastolic BP, respectively, compared to an FDA approved BP cuff. The vision for this technology is not necessarily to replace existing BP monitoring solutions, but rather to introduce a downloadable smartphone software application that could serve as a low-barrier hypertension screening measurement fit for widespread adoption.
Article
Oscillometric Blood Pressure Measurements on Smartphones using Vibrometric Force Estimation
https://doi.org/10.21203/rs.3.rs-4033076/v1
This work is licensed under a CC BY 4.0 License
Journal Publication
published 31 Oct, 2024
You are reading this latest preprint version
This paper proposes a novel oscillometric smartphone BP measurement method that requires no hardware modifications, no physical attachments, nor any need for individual calibration with cuff-based BP measurements. The proposed measurement can be enabled solely with software by downloading a smartphone application on nearly any smartphone. It is NOT an “optical” BP measurement relying on pulse transit time (PTT), pulse arrival time (PAT), pulse wave analysis (PWA), or any other technique estimating the relative change in BP. The application performs absolute BP measurements based on the oscillometric method similar to that employed by automated BP cuffs, in accordance with the recommendation of the American Medical Association (AMA) and the American Heart Association (AHA)1.
The oscillometric method requires measuring two key metrics: applied force and local blood volume of an artery. Local blood volume of an artery can be accurately estimated via photoplethysmography (PPG) with the smartphone camera. Measuring applied force with no attachments serves as the key insight presented in the paper to enable a novel BP measurement. Figure 1 depicts these key concepts of smartphone oscillometric BP measurements proposed in this paper.
To measure applied force, we propose a force estimation technique utilizing only the vibration motor and inertial measurement unit (IMU). Nearly every phone contains an IMU with an accelerometer and gyroscope to measure the linear acceleration and angular velocity that enables phone position inferences used for rotating the screen, playing games, counting steps, gesture recognition, and much more. Similarly, nearly all smartphones contain a vibration motor to provide haptic feedback for messages, phone calls, and screen interactions. Taking advantage of the preexisting vibration motor and IMU allows us to create a model to measure force. The phone is set to maximally vibrate while the IMU measures the motion as the phone oscillates (from the vibration). When force is applied during vibration, the IMU signal changes as the oscillations are altered or dampened. The alterations in the vibration signal allows us to train a model to estimate the applied force. We refer to this method of estimating applied force using the vibration motor and IMU signal as Vibrometric Force Estimation (VFE).
This work represents the first absolute BP measurement technique on a smartphone without special components or attachments. Most prior research into smartphone and smartwatch BP measurements perform relative BP measurements using PTT or PWA, but these techniques are fundamentally different. PTT and PWA only provide relative change in BP. These measures of relative change assume an individual has access to a BP cuff to calibrate their individual measurement due to the model's reliance on arterial stiffness and arm length, otherwise, the relative measurements cannot be meaningfully interpreted2–9. Prior works on absolute smartphone BP measurement require custom built attachments to measure force10, 11. One prior work functions only on a specific smartphone model with a built in force sensitive screen, which is not a common feature12. Our proposed method is the first smartphone BP measurement that requires no attachment and functions on nearly any smartphone model, only requiring a vibration motor, camera, and IMU.
Blood Pressure Study Participants
The full BP validation study recruitment included N = 30 participants, each with three resting smartphone BP measurements. From the 30 participants, 23 participants also participated in the exercise portion of the study for a total of 133 recorded smartphone BP samples. The participant population was 20% White, 67% Asian, and 13% Hispanic consisting of 50% males and 50% females. The mean age of the participants is 24·7 years with a standard deviation of 4·3 The average resting systolic and diastolic BP is 103mmHg and 70mmHg, respectively. Further details about participant statistics are included in Table 1. During the study, N = 29 participants were able to correctly apply force in at least 1 of their measurements as guided by the smartphone application. N = 24 participants contained at least 1 usable measurement (valid force and ppg signal combination) for BP prediction.
The 133 samples consist of 90 normal samples and 43 exercise samples. Of the 133 samples, 65 were excluded based on the criteria detailed in the final paragraph of the methods section. For first time participants and participants performing exercise during the measurement, user error is more common so the high exclusion rate is expected. For the exercise samples, which involved participants performing a wall sit while taking a measurement, 28 of the 43 measurements are excluded. This resulted in usable exercise measurements from only 11 of the 23 exercise participants. For the non-exercise measurements, 37 of the 90 measurements are excluded. The data included in the BP model training and testing consisted of 53 non-exercise measurements and 15 exercise measurements. For detailed information on exclusion criteria for participant measurements, please refer to the last paragraph of the methods section.
Table 1. Summary of Participant Demographics. This table summarizes participant demographic information including BP measurement statistics. Detailed information is available in the supplemental section.
Blood Pressure Proof-of-Concept Study
The BP proof-of-concept study comparing the smartphone to the FDA-approved BP cuff with a hold one participant out validation achieved a mean absolute error of 9·35 mmHg and 7·94 mmHg for systolic and diastolic, respectively. The Pearson correlation coefficient was 0·69 and 0·43 for systolic and diastolic measurements, respectively. The BP cuff measurements ranged from 89 to 147 mmHg systolic and 60 to 107 mmHg for diastolic. These results are depicted in Fig. 2.
In evaluating the proof of concept more specifically around hypertension screening, we categorize BP values based on the AMA/ACC standards13, which involves determining if an individual is below 120mmHg (normal), between 120–130 mmHg (elevated), and above 130 mmHg (hypertensive). In categorizing the model results as normal (below 130mmHg systolic BP) or hypertensive (above 130mmHg systolic BP), the ROC analysis reveals an area under the curve (AUC) score of 0·92 for detecting high BP individuals, characterized by systolic BP greater than 130 mmHg. Under the same conditions, the model achieved a sensitivity and specificity of 0·91 and 0·69, respectively.
As shown in Fig. 3 part B, the oscillograms of each BP group contain characteristics supporting the clinical understanding of the oscillometric method. The most obvious characteristic is the peak of the oscillogram, often noted to represent the mean arterial pressure (MAP). As the MAP increases, the peak shifts right towards higher applied pressures..
Skin Tone and Ethnicity.
An ANOVA test is conducted to determine the impact of skin tone and ethnicity on performance. For both skin tone and ethnicity, the group variations are not statistically significant with p-values of 0·86 and 0·22, respectively. The skin tone and ethnicity of the participant population is listed in Table 1.
Cross Device Validity
We ran a benchtop experiment demonstrating the force estimation technique across multiple models of phones. The accuracy of measuring applied finger force using VFE is evaluated across three smartphones: Google Pixel 4, Samsung Galaxy A53, and Motorola Moto G Power. The mean absolute error, standard deviation, bias, and correlation coefficient for the Google Pixel 4 is 0·77N, 0·90N, -0·46N, and 0·92, respectively. The mean absolute error, standard deviation, bias, and correlation coefficient for the Samsung Galaxy A53 is 0·55N, 0·64N, -0·18N, and 0·95 respectively. The mean absolute error, standard deviation, bias, and correlation coefficient for the Motorola Moto G Power 2022 is 0·39N, 0·53N, 0·01N, and 0·96, respectively. The average correlation coefficient across all phones is 0·92. These results are visualized in Fig. 4.
The study results demonstrate the feasibility of oscillometric smartphone BP measurements using VFE and PPG measurements. The systolic correlation coefficient of 0·69 and an MAE of 9·35 mmHg are comparable to state-of-the-art smartphone-based BP measurements performed with attachments or custom equipment10, 11, 12. Additionally, the results suggest the smartphone measurement functions equally for groups of different ethnicities or skin tones. The central improvement offered in the work is the method for performing absolute BP measurements that can be applied across smartphone models without modifications or attachments.
Despite the promise, it is important to recognize the limitations of the proposed technology. The performance of the diastolic BP prediction is limited with a correlation coefficient of 0·42. This suggests that isolated diastolic hypertension could not be well detected; however, systolic BP alone can still accurately screen for hypertension in all but 3·2% of the population because the prevalence of isolated diastolic hypertension is extremely low compared to other types of hypertension14. Future exploration of data collection at lower force levels and features more specific to diastolic BP will likely improve the diastolic BP performance.
The VFE technique is smartphone dependent. However, the calibration can be performed relatively quickly and once the profiling is completed for a phone model, VFE can work on all phones of that model. It would be expected that the application developer would perform the calibrations so the end user can simply download a ready-to-use application.
While the overall goal of this paper is to present and investigate the novel smartphone BP measurement methodology using vibrometric force estimation, there is significant potential for future work. The proof-of-concept studies performed here demonstrate feasibility, but future studies involving larger scale data collections of participants with varied hypertensive conditions will be performed to further develop vibrometric force estimation model parameters, train larger models, and better evaluate accuracy. Additionally, the studies performed thus far have all been performed under supervision in a controlled lab environment. There is a need for future work to explore the usability of the system for different demographics, especially focusing on the potential for individuals to perform measurements in at-home environments. Also, future research should include investigations of how conditions like diabetic neuropathy, Parkinson’s disease, or thick skin calluses might affect the measurement.
Blood Pressure Validation Study
In the BP validation study, the smartphone device is compared against an FDA-approved cuff-based device (Omron BP7350). The participant first records a trial measurement using the smartphone to learn how to use the device and practice applying force. The trial measurement is not used for analysis. The cuff-based device is used to measure the participants BP following standard protocols. Immediately afterwards, the participant performs three BP measurements using the smartphone. After completing three BP measurements on the smartphone, the cuff-based device is used to measure the BP again. The full measurement sequence is as follows: cuff, smartphone, smartphone, smartphone, cuff. If the two cuff BP readings were successful, the two BP readings are averaged to provide a ground truth for all three smartphone measurements. If one of the cuff readings failed, the single successful cuff measurement serves as the ground truth.
The participants are also asked to optionally complete a second phase of the data collection involving exercise. The purpose of this exercise phase is to obtain high BP data. The N = 12 consenting participants are asked to perform a wall sit for approximately one minute. During the wall sit, the participants simultaneously measure their BP with the BP cuff device and the smartphone device using contralateral arms. Both BP measurements are initiated approximately ten seconds after the start of the wall sit.
The methods and experimental protocols described here are approved by the UC San Diego Internal Review Board (IRB) under protocol number 804668. The experiments were carried out in accordance with the relevant guidelines and regulations. Informed consent was obtained from all participants.
Blood Pressure Study Participants
A total of N = 30 participants were recruited for the BP proof of concept study. Participant demographic information is available in Table 1. Of the participants recruited, N = 6 participants were excluded from the data analysis. N = 1 participant was unable to perform any valid force measurements; N = 1 participant had low prominence in all measurements; and N = 1 participant had poor gaussian skew fit in all measurements. The other N = 3 excluded participants had a mixture of criteria below.
Participant measurements are excluded based on the following exclusion criteria:
-
Cuff BP exceeds 155mmHg: If the reference measurement exceeded 160mmHg, the measurement is excluded based on prior research demonstrating that the fingertip and arm BP values differ at extremely high BP values induced from exercise24.
-
Inconsistent/Undetected Pulse or Excessive Noise: The Fast Fourier Transform (FFT) of the PPG signal between zero and two Hz (frequencies relevant for human pulse) must have a narrow, identifiable peak indicating the pulse rate.
-
Insufficient Applied Force: The applied force measurement must have a correlation coefficient greater than 0·88. The first 0·5 seconds of the force signal must have a correlation coefficient less than 0·21N. Also, the minimum applied force must be less than six N and the maximum applied force value must be greater than seven N.
-
Poor Skew Gaussian Fit: The mean absolute error for the normalized Skew Gaussian of best fit must be less than 0·18.
-
Saturation in PPG signal: If the PPG signal rises above 254 pixel intensity units, the measurement is discarded. This is indicative of the finger being lifted off the smartphone camera during a measurement.
-
Unchanging in PPG Prominence: The standard deviation of the normalized PPG prominence must be greater than 0·17.
Smartphone Application
For smartphone force estimation and BP studies, we develop a custom application called “VibroBP” to guide users and collect data. The smartphone application, shown in Fig. 1, provides the user with two key visual signifiers. First, the image of the fingertip near the smartphone camera provides an intuitive understanding of how to position the finger. Second, during the force measurement, the smartphone app plots the applied force in real time overlaid with a force guideline. The force guide aids the user to apply the correct amount of force during the measurement.
The application additionally provides the research staff with helpful information during data collection. The PPG signal is plotted in real time before and during the measurement so the research staff can ensure data quality. If the PPG signal is not within the desired range, the research staff can press a button to calibrate the PPG signal before the measurement to ensure the PPG signal is not over- or under- saturated. At the end of the study, the full force and PPG signals are plotted within the application to provide an overview of the data.
Force Dampening Effect Profiling
The proposed smartphone BP measurement method largely depends on accurately estimating force without an attachment. Each smartphone model has its own force dampening effect profile. To characterize the profile, a small-scale data collection is performed on each phone model to capture the vibration modulation at different known forces measured by a force sensitive resistor. This calibration procedure and all studies in the paper are performed on a flat wooden desktop surface.
We conducted this profiling for three different smartphones. With N = 5 participants of various hand sizes, a user applies pressure with their index finger onto a 0·3mm thick force sensor positioned on top of the smartphone front camera. The user applies force with their index finger in the same vertical positioning as the BP measurement. The force sensor, a calibrated, linear force sensor (SingleTact 15mm 4·5N), is placed over the smartphone front camera. The user presses their finger on the force sensor (and smartphone camera) to measure the applied force of the finger and collect smartphone data.
During a measurement, real-time readings from the ground truth force sensor are displayed on a computer screen in front of the user with a force guide. The user is instructed to use real time feedback to follow the force guide in applying a range of pressures. Like the BP measurement, the force guide instructs users to continuously increase the applied force. The study includes three repetitions of a 40 second session of applying a range of forces. These three repetitions are performed by each participant on all three smartphones for a resulting 10 minutes of raw applied force data on each phone (2 minutes per participant).
Vibrometric Force Estimation
For each smartphone, data from the force dampening effect profiling study provides IMU data during a range of applied force values. To estimate force, the IMU profile is used to train a machine learning model for each phone model. The IMU data contains three axis accelerometer data and three axis gyroscope data. Each axis of the IMU data contains high frequency signals from the vibration motor as well as noise at lower frequencies. The applied force from the finger most significantly affects the amplitude of the signal oscillation. To leverage this understanding, the rolling standard deviation of each axis of the accelerometer and gyroscope data serve as features. The features most significantly correlated to the force data are included as model inputs, while other features are excluded. With these IMU data features as inputs, a regression model (multivariate linear regression model or gradient boosting) is trained using a ten percent holdout for validation. This process is repeated for each smartphone such that a distinct force estimation model exists for each smartphone model.
Smartphone Camera PPG
For the oscillometric BP measurement, we leverage the smartphone camera PPG to sense volumetric changes in blood flow at the transverse palmar arch branch of the digital artery. As the volume of blood flowing through the artery near the fingernail bed changes, the reflective properties of the tissue in the region changes. As such, the smartphone camera can record changes in blood volume by measuring the changes in the reflective properties.
In the study, the phone screen is set to a pure white background with maximal brightness and the user is instructed to place their index finger over the camera (as aided by a visual signifier). The bright white screen illuminates the finger and the camera records pixel intensity changes in the red channel as a proxy for blood volume changes. To account for skin tone and lighting variations, the smartphone performs the calibration outlined in Xuan and Barry et. al.21 with the user’s finger over the camera. This calibration protects against under- or over- saturated PPG measurements.
Oscillometry and Blood Pressure Estimation
During the smartphone BP measurement, the smartphone camera records the PPG signal while the user exerts force with the index finger on the camera. Applying force causes changes in the PPG signal to create an oscillogram shape. When the applied force is greater than the diastolic BP, the blood volume during the systolic pulse phase increases and the blood volume in the diastolic phase decreases. The total volume of blood flowing through the artery is approximately the same, but pressure on the artery causes more blood to flow through at the systolic pulse phase and less at the diastolic pulse phase. This increases the prominence of the PPG signal. When the applied force surpasses the mean arterial pressure, the blood volume begins to be limited even during the systolic phase of the pulse. As such, the systolic phase PPG prominence begins to decrease and the blood volume during the diastolic phase approaches zero. The PPG prominence continues to decrease even after the applied force surpasses the systolic BP. As the pressure increases, the signal will decrease to zero (or noise) as the blood flows through the artery during both the systolic and diastolic phase approach zero.
A plot of the pulse prominence through the range of applied forces (from less than diastolic BP to greater than systolic BP) creates a Gaussian or skew Guassian-like shape that we refer to as the oscillogram. The oscillometric method utilized by most FDA approved BP cuffs and numerous prior academic works utilize the oscillogram shape to estimate BP7. Our method of estimating BP relies on the features and findings identified in prior work.
We generate eleven features by focusing on features and concepts previously identified in prior work22 for oscillometric BP measurements. The following features are included in the BP estimation model: the applied force at the maximum prominence value, maximum gradient, minimum gradient, applied force at the maximum gradient, applied force at the minimum gradient, the maximum value of the oscillogram, prominence at the maximum gradient, prominence at the minimum gradient, applied force at the maximal low pass filtered PPG values, and applied force at the maximum gradient of low pass filtered PPG values. The concepts for these key features are visualized in Fig. 3 part A.
These key features of the oscillometric method are the only inputs into our BP prediction model. Note that the attributes related to exercise including heart rate and pulse shape are not included in the features provided to the model. Additionally, indirect measurements of heart rate or pulse are further avoided by interpolating all oscillograms to have the same number of points with the same force values. With the key features, we trained a least absolute shrinkage and selection operator (LASSO) regression model to predict systolic and diastolic measures of BP23. The model is trained and tested using a hold one participant out validation. In this training scheme, the model is trained on all data excluding the one or more BP measurements from a single participant. The model is then used to predict the BP of the holdout participant. This process is repeated for each participant.
Funding:
This project is not funded.
Author contributions:
C.B. and E.W. contributed to project conceptualization, reviewing, and editing. C.B. is responsible for prototype development, data collection, data processing, and writing the manuscript. A.M. provided expert medical insight and reviewed the paper. Y.X. contributed to code for PPG signal acquisition and insights for project conceptualization. E.W. supervised the project. A.F. reviewed the paper and provided insight into figure formation.
Competing interests:
Colin Barry and Edward Jay Wang are cofounders of Billion Labs Inc. which intends to pursue FDA clearance of the proposed technology.
Colin Barry and Edward Jay Wang are inventors on a provisional patent pertaining to this manuscript with the attorney docket number 09062-8487 US00
At the time of submission, Yinan Xuan holds an internship position at Meta.
Alison Moore is affiliated with Wake Forest University, holds a leadership position in the American Geriatrics Society, and participates on an advisory board for Cedars Sinai.
Ava Fascetti declares no competing interests.
Data and materials availability: Additional information, data, code, or materials relevant to the research are available upon request to the corresponding author, Colin Barry, at [email protected].
- D. Shimbo, N. T. Artinian, J. N. Basile, L. R. Krakoff, K. L. Margolis, M. K. Rakotz, G. Wozniak, On behalf of the American Heart Association and the American Medical Association, Self-Measured Blood Pressure Monitoring at Home: A Joint Policy Statement From the American Heart Association and American Medical Association, Circulation 142 (4) (Jul. 2020). doi:10.1161/CIR.0000000000000803. URL https://www.ahajournals.org/doi/10.1161/CIR.0000000000000803
- A. Dias Junior, S. Murali, F. Rincon, D. Atienza, Methods for reliable estimation of pulse transit time and blood pressure variations using smartphone sensors, Microprocessors and Microsystems 46 (2016) 84–95. doi:10.1016/j.micpro.2016.06.001. URL https://linkinghub.elsevier.com/retrieve/pii/S0141933116300667
- L. Frey, C. Menon, M. Elgendi, Blood pressure measurement using only a smartphone, npj Digital Medicine 5 (1) (2022) 86. doi:10.1038/s41746- 022-00629-2. URL https://www.nature.com/articles/s41746-022-00629-2
- A. Joosten, A. Jacobs, O. Desebbe, J.-L. Vincent, S. Sarah, J. Rinehart, L. Van Obbergh, A. Hapfelmeier, B. Saugel, Monitoring of pulse pres- sure variation using a new smartphone application (Capstesia) versus stroke volume variation using an uncalibrated pulse wave analysis monitor: a clinical decision making study during major abdominal surgery, Journal of Clinical Monitoring and Computing 33 (5) (2019) 787–793. doi:10.1007/s10877-018-00241-4. URL https://link.springer.com/10.1007/s10877-018-00241-4
- A. D. Junior, S. Murali, F. Rincon, D. Atienza, Estimation of Blood Pressure and Pulse Transit Time Using Your Smartphone, in: 2015 Euromicro Conference on Digital System Design, IEEE, Madeira, Portugal, 2015, pp. 173–180. doi:10.1109/DSD.2015.90. URL http://ieeexplore.ieee.org/document/7302267/
- H. Liu, K. Ivanov, Y. Wang, L. Wang, Toward a Smartphone Application for Estimation of Pulse Transit Time, Sensors 15 (10) (2015) 27303– 27321. doi:10.3390/s151027303. URL http://www.mdpi.com/1424-8220/15/10/27303
- R. Mukkamala, J.-O. Hahn, O. T. Inan, L. K. Mestha, C.- S. Kim, H. Toreyin, S. Kyal, Toward Ubiquitous Blood Pressure Monitoring via Pulse Transit Time: Theory and Practice, IEEE Transactions on Biomedical Engineering 62 (8) (2015) 1879–1901. doi:10.1109/TBME.2015.2441951. URL http://ieeexplore.ieee.org/document/7118672/
- C. Poon, Y. Zhang, Cuff-less and Noninvasive Measurements of Arterial Blood Pressure by Pulse Transit Time, in: 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, IEEE, Shanghai, China, 2005, pp. 5877–5880. doi:10.1109/IEMBS.2005.1615827. URL http://ieeexplore.ieee.org/document/1615827/
- P. Schoettker, J. Degott, G. Hofmann, M. Proen ̧ca, G. Bonnier, A. Lemkaddem, M. Lemay, R. Schorer, U. Christen, J.-F. Knebel, A. Wuerzner, M. Burnier, G. Wuerzner, Blood pressure measurements with the OptiBP smartphone app validated against reference auscultatory measurements, Scientific Reports 10 (1) (2020) 17827. doi:10.1038/s41598-020-74955-4. URL https://www.nature.com/articles/s41598-020-74955-4
- A. Chandrasekhar, C.-S. Kim, M. Naji, K. Natarajan, J.-O. Hahn, R. Mukkamala, Smartphone-based blood pressure monitoring via the oscillometric finger-pressing method, Science Translational Medicine 10 (431) (2018) eaap8674. doi:10.1126/scitranslmed.aap8674. URL https://www.science.org/doi/10.1126/scitranslmed.aap8674
- Y. Xuan, C. Barry, J. De Souza, J. H. Wen, N. Antipa, A. A. Moore, E. J. Wang, Ultra-low-cost mechanical smartphone attachment for no- calibration blood pressure measurement, Scientific Reports 13 (1) (2023) 8105. doi:10.1038/s41598-023-34431-1. URL https://www.nature.com/articles/s41598-023-34431-1
- A. Chandrasekhar, K. Natarajan, M. Yavarimanesh, R. Mukkamala, An iPhone Application for Blood Pressure Monitoring via the Oscillo- 17 metric Finger Pressing Method, Scientific Reports 8 (1) (2018) 13136. doi:10.1038/s41598-018-31632-x. URL http://www.nature.com/articles/s41598-018-31632-x
- P. K. Whelton, R. M. Carey, W. S. Aronow, D. E. Casey, K. J. Collins, C. Dennison Himmelfarb, S. M. DePalma, S. Gidding, K. A. Jamerson, D. W. Jones, E. J. MacLaughlin, P. Muntner, B. Ovbiagele, S. C. Smith, C. C. Spencer, R. S. Stafford, S. J. Taler, R. J. Thomas, K. A. Williams, J. D. Williamson, J. T. Wright, 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: A report of the american college of cardiology/american heart association task force on clinical practice guidelines, Hypertension 71 (6) (2017). doi:10.1161/HYP.0000000000000065. URL https://www.ahajournals.org/doi/10.1161/HYP.0000000000000065
- C. K. Chow, Prevalence, Awareness, Treatment, and Control of Hypertension in Rural and Urban Communities in High-, Middle-, and Low-Income Countries, JAMA 310 (9) (2013) 959. doi:10.1001/jama.2013.184182. URL http://jama.jamanetwork.com/article.aspx?doi=10.1001/jama.2013.184182
- R. Darbar, D. Samanta, SurfaceSense: Smartphone Can Recognize Where It Is Kept, in: Proceedings of the 7th International Conference on HCI, IndiaHCI 2015, ACM, Guwahati India, 2015, pp. 39–46. doi:10.1145/2835966.2835971. URL https://dl.acm.org/doi/10.1145/2835966.2835971
- J. Cho, I. Hwang, S. Oh, VibePhone: efficient surface recognition for smartphones using vibration, Pattern Analysis and Applications 19 (1) (2016) 251–265. doi:10.1007/s10044-015-0460-8. URL http://link.springer.com/10.1007/s10044-015-0460-8
- J. Cho, I. Hwang, S. Oh, Vibration-Based Surface Recognition for Smartphones, in: 2012 IEEE International Conference on Embedded and Real-Time Computing Systems and Applications, IEEE, Seoul, Ko- rea (South), 2012, pp. 459–464. doi:10.1109/RTCSA.2012.42. URL http://ieeexplore.ieee.org/document/6301467/
- M. Feng, Y. Fukuda, M. Mizuta, E. Ozer, Citizen sensors for SHM: use of accelerometer data from smartphones, Sensors (Basel, Switzerland) 15 (2) (2015) 2980–2998. doi:10.3390/s150202980.
- A. Feldbusch, H. Sadegh-Azar, P. Agne, Vibration analysis using mobile devices (smartphones or tablets), Procedia Engineering 199 (2017) 2790–2795. doi:10.1016/j.proeng.2017.09.543. URL https://linkinghub.elsevier.com/retrieve/pii/S1877705817340614
- K. Ali, A. X. Liu, Fine-grained Vibration Based Sensing Using a Smart- phone, arXiv:2007.03874 [cs, eess] (Aug. 2020). URL http://arxiv.org/abs/2007.03874
- Y. Xuan, C. Barry, N. Antipa, E. J. Wang, A calibration method for smartphone camera photophlethysmography, Frontiers in Digital Health 5 (2023). doi:10.3389/fdgth.2023.1301019. URL https://www.frontiersin.org/articles/10.3389/fdgth.2023.1301019
- A. Chandrasekhar, M. Yavarimanesh, J.-O. Hahn, S.-H. Sung, C.-H. Chen, H.-M. Cheng, R. Mukkamala, Formulas to Explain Popular Oscillometric Blood Pressure Estimation Algorithms, Frontiers in Physiology 10 (2019) 1415. doi:10.3389/fphys.2019.01415. URL https://www.frontiersin.org/article/10.3389/fphys.2019.01415/full
- R. Tibshirani, Regression shrinkage and selection via the lasso: a retrospective: Regression Shrinkage and Selection via the Lasso, Journal of the Royal Statistical Society: Series B (Statistical Methodology) 73 (3) (2011) 273–282. doi:10.1111/j.1467-9868.2011.00771.x. URL https://onlinelibrary.wiley.com/doi/10.1111/j.1467-9868.2011.00771.x
- J. A. Nijboer, J. C. Dorlas, J. Lubbers, The difference in blood pressure between upper arm and finger during physical exercise, Clinical Physi- ology 8 (5) (1988) 501–510. doi:10.1111/j.1475-097X.1988.tb00215.x. URL https://onlinelibrary.wiley.com/doi/10.1111/j.1475-097X.1988.tb00215.x
Competing interest reported. Colin Barry and Edward Jay Wang are cofounders of Billion Labs Inc. which intends to pursue FDA clearance of the proposed technology.
Colin Barry and Edward Jay Wang are inventors on a provisional patent pertaining to this manuscript with the attorney docket number 09062-8487 US00
At the time of submission, Yinan Xuan holds an internship position at Meta.
Alison Moore is affiliated with Wake Forest University, holds a leadership position in the American Geriatrics Society, and participates on an advisory board for Cedars Sinai.
Ava Fascetti declares no competing interests.