Valencell was scheduled to present the results of a clinical study on its groundbreaking blood pressure monitoring technology at the American College of Cardiology conference in March, but unfortunately that conference was cancelled. So we’ve decided to share that presentation and research here in a webinar format with an open Q&A session. You can find more information on Valencell's blood pressure technology here: https://valencell.com/bloodpressure/
7. A biometric audio earbud
can accurately monitor
BP during free-living
conditions without the
need for a cuff
Presenting Author: Dr. Steven F. LeBoeuf
Additional authors: Dr. Chris Eschbach, Tushar Tank, Dr. Ryan
Hodges, Seth Long, and Dr. Daniel Moore
Valencell, Inc., Raleigh, NC
BACKGROUND
RESULTS
A training dataset (Fig. 2) was collected from BP measurements on
>3000 subjects in 3 sequential readings: a manual auscultatory
reading, an automated oscillometric cuff reading, and a PPG reading
from a commercial PPG sensor embedded within a wireless audio
earbud. With the manual readings serving as “ground truth”, a machine
learning model for ear-PPG BP (EPBP) was trained to predict both
systolic and diastolic BP based solely on ear-PPG sensor data and
subject meta data (age, height, weight, and gender).This machine
learning model was then integrated into Valencell’s PerformTek®
biometric software to provide an EPBP solution comprising 2 key
software components: 1) an automated data qualifier and 2) a BP
estimator. The data qualifier was configured to automatically determine
whether the PPG measurement was likely to yield cuff-like accuracy,
and the BP estimator was configured to estimate BP based on the
qualified data.
An unbiased accuracy assessment of the EPBP solution was
generated by statistical analysis of the model’s BP predictions for a
qualified test dataset of >100 total participants (Fig. 3). This test
dataset was collected following the ISO 81060-2(2018) standard for
noninvasive sphygmomanometers and was never used to train the
EPBP model.
Daily self-monitoring of blood pressure (BP) is widely recognized as a
cost-effective method to identify and manage hypertension in
conjunction with physician-prescribed interventions. However, few
American households own a BP monitor and even less comply with
daily monitoring, in part due to annoyances associated with porting and
properly donning a BP cuff. The purpose of this study was to
demonstrate the accuracy of a cuff-less, calibration-free BP monitoring
technology, based upon photoplethysmography (PPG), embedded
within a popular consumer audio earbud form-factor (Fig. 1).
The resulting qualified test dataset comprised 654 measurements from
147 subjects (Fig. 3). For the qualified test dataset, the accuracy and
tracking characteristics of the EPBP model predictions were found to be
comparable to that of the automated cuff , having a mean ±
estimation error within 5 ± 8 mmHg, with respect to the manual reading
(Fig. 5).
DISCUSSION
For more information, scan the
QR code, go to
www.valencell.com, or email
info@valencell.com
METHODS
Fig. 1: Valencell developed and validated a novel machine learning model, based on
PPG datasets collected from thousands of unique participants, to estimate BP from PPG.
FIG 1 - METHODS
FIG 3 – QUALIFIED TEST DATASET
BIOMETRIC AUDIO
EARBUDS PRODUCE CUFF-
LIKE ACCURACY IN
MONITORING BLOOD
PRESSURE
The ultimate purpose of this R&D program was to develop and validate the
world’s first commercially viable cuff-less, calibration-free, all-PPG BP monitoring
solution that could ultimately: 1) provide demonstrable public health value,
2) provide cuff-like accuracy in a general population, and 3) be integrated with
existing popular consumer form-factors. Both objectives were achieved. Moreover,
for the qualified test dataset, it was found that the biometric earbud solution was
able to categorize hypertension status with cuff-like accuracy (Fig. 4).
CONCLUSION
A PPG-enabled audio earbud can accurately monitor BP during free-
living conditions without the need for a cuff.
DISCLOSURE INFORMATION
*All authors are employed by Valencell, Inc.: valencell.com/patents
FIG 5 – RESULTS
PPG Auto-cuff
Accuracy 89% 88%
Precision 73% 69%
Sensitivity 77% 75%
Specificity 93% 92%
FIG 2 – TRAINING DATASET
FIG 4 – HYPERTENSION STATUS
Fig. 4: The PPG-based BP model was
shown to identify hypertension status as
effectively as an auto-cuff.
17. “Your numbers look great. I’ve updated your
diet and exercise plan. Don’t forget you have
an 8AM meeting today.”
“I’ve recommended the route that keeps
your BP lowest for your commute today.”
“I noticed you’re
having an after-
lunch food coma
- you might want
to take a quick
walk and get a
cup of coffee.”
“I’ve customized this
exercise session based
on your activity today to
get you in Heart Rate
Zone 2 which helps lower
blood pressure.
Enjoy!”
“Would you like me to play
your relaxing track for your
commute today?”
“Your BP is elevated, and you appear to
be stressed. Would you like me to start a
relaxation session?”
“It seems like you had a stressful day. Would
you like to do a quick meditation to relax?
Overall your weekly BP trend look good –
keep up the exercise plan.”
“I’ve prepared a
summary of how
well you did on
your exercise
goals on your
phone. I’ll play
your relaxation
music.”
General Wellness:
Monitoring BP using a
biometric audio earbud