Blood Pressure Monitoring Using a Smartphone Camera: Performance of the OBPM Technology

Abstract

While hypertension globally affects two out of five adults worldwide, there exists no easily-scalable technology to measure blood pressure out of the clinics. The idea of using smartphone sensors to estimate blood pressure was already tackled in the past, but failed because of low accuracy. Based on data from a previous study (NCT02651558), we recently developed a library of algorithms that predicts blood pressure from optical signals: the Optical Blood Pressure Monitoring (OBPM) technology. In the current work, we studied the performances of this technology when applied to video sequences acquired by a commercial smartphone camera. We implemented a measurement campaign on 35 healthy volunteers that performed physical exercises. The volunteers were requested to apply their right forefinger on top of the camera of a commercial smartphone while video sequences were acquired. The video sequences were then processed by the OBPM algorithms, and the predicted blood pressure values were compared to reference oscillometric blood pressure readings. After a calibration procedure, the predicted diastolic blood pressure values showed to comply with the ISO81060-2 performance requirements with a mean error of 0.32 mmHg, and a standard deviation of the error of 7.02 mmHg. This study provides first experimental evidence to support that a commercial smartphone can be transformed into a blood pressure monitor by means of the OBPM technology.

Full text

Abstract— While hypertension globally affects two out of five
adults worldwide, there exists no easily-scalable technology to
measure blood pressure out of the clinics. The idea of using
smartphone sensors to estimate blood pressure was already
tackled in the past, but failed because of low accuracy.
Based on data from a previous study (NCT02651558), we
recently developed a library of algorithms that predicts blood
pressure from optical signals: the Optical Blood Pressure
Monitoring (OBPM) technology. In the current work, we
studied the performances of this technology when applied to
video sequences acquired by a commercial smartphone camera.
We implemented a measurement campaign on 35 healthy
volunteers that performed physical exercises. The volunteers
were requested to apply their right forefinger on top of the
camera of a commercial smartphone while video sequences were
acquired. The video sequences were then processed by the
OBPM algorithms, and the predicted blood pressure values were
compared to reference oscillometric blood pressure readings.
After a calibration procedure, the predicted diastolic blood
pressure values showed to comply with the ISO81060-2
performance requirements with a mean error of 0.32 mmHg,
and a standard deviation of the error of 7.02 mmHg.
This study provides first experimental evidence to support
that a commercial smartphone can be transformed into a blood
pressure monitor by means of the OBPM technology.
I. BLOOD PRESSURE MEASUREMENTS MADE EASY
Attempts to use off-the-shelf smartphone sensors to
estimate blood pressure have been done in the past, but they
failed because of too low accuracy [1]. This study aims at
demonstrating that – using clinical-grounded algorithms [2] –
video sequences acquired by smartphone cameras do contain
reliable blood pressure-related informations (see Figure 1).
MEASUREMENT
OPTICAL PULSES
OBPM ALGORITHM BP = 86 mmHg
Figure 1. Transforming a smartphone into a blood pressure monitor: from
video sequences towards blood pressure measurements.
J. Solà, M. Proença, A. Lemkaddem, F. Braun, C. Verjus, and M. Bertschi
are with the CSEM S.A. - Centre Suisse d’Electronique et de
Microtechnique, Neuchâtel, Switzerland (e-mail: Josep.Sola@csem.ch).
T. Kunz, E. Jones, and P. Schoettker are with Biospectal Inc., Lausanne,
Switzerland (e-mail: patrick@biospectal.com)
II. MATERIALS AND METHODS
The right forefinger of 35 healthy volunteers was placed in
contact with a embedded smartphone camera (Samsung
Galaxy S7) while the smartphone flash light was on. A
custom-made dedicated Android application (Biospectal Inc.,
Switzerland) allowed to record high speed video sequences in
MP4 format (sampling rate = 120 frames/s). Reference
diastolic blood pressure measurements were recorded by a
commercial inflation-mode oscillometric device (Braun
ExactFit 3 BP6100) placed on the left arm. During the
measurement campaign, volunteers increased their blood
pressure by performing three isovolumetric leg extension
exercises. Diastolic blood pressure was increased on average
by 30 ± 6 mmHg during the protocol. Our
previously-published algorithms to estimate blood pressure
from optical signals (OBPM algorithms [3]) were applied to
pulsatile time series obtained from the acquired video
sequences of each volunteer.
III. RESULTS
The overall performance over the entire cohort of 35 healthy
volunteers is provided in Table I. Two calibration strategies
were tested. For the initial calibration, the output of the
OBPM algorithm was shifted in offset using the first reference
oscillometric diastolic blood pressure measurement. For the
full calibration, the output of the OBPM algorithm was shifted
and scaled to best match the reference measurements.
TABLE I. PERFORMANCE OF OBPM ALGORITHMS WHEN APPLIED TO
SIGNALS RECORDED BY A SMARTPHONE CAMERA.
OBPMTM
Algorithm Initial Cal. Full Cal.
ISO81060
-2
performances
Mean Error 4.0 mmHg 0.32 mmHg
Stdev Error 8.22 mmHg 7.02 mmHg
REFERENCES
[1] T. B. Plante, “Validation of the Instant Blood Pressure Smartphone
App”, JAMA Intern Med. 176:700-702, 2016,
doi: 10.1001/jamainternmed.2016.0157
[2] J. Solà, “Continuous non-invasive monitoring of blood pressure in the
operating room: a cuffless optical technology at the fingertip”,
Proceedings BMT2016, Basel, Switzerland, 2016,
doi: 10.1515/cdbme-2016-0060
[3] M. Proença, "Method, apparatus and computer program for
determining a blood pressure value”, WO Patent WO/2016/138965
Blood Pressure Monitoring Using a Smartphone Camera:
Performance of the OBPM Technology
J. Solà, M. Proença, P. Schoettker, A. Lemkaddem, F. Braun,
C. Verjus, M. Bertschi, E. Jones, and T. Kunz