ArticlePDF Available

Measuring blood pressure from Korotkoff sounds as the brachial cuff inflates on average provides higher values than when the cuff deflates

Authors:

Abstract and Figures

Objectives. In this study, we test the hypothesis that if, as demonstrated in a previous study, brachial arteries exhibit hysteresis as the occluding cuff is deflated and fail to open until cuff pressure (CP) is well below true intra-arterial blood pressure (IAPB), estimating systolic (SBP) and diastolic blood pressure (DBP) from the presence of Korotkoff sounds as CP increases may eliminate these errors and give more accurate estimates of SBP and DBP relative to IABP readings. Approach. In 62 subjects of varying ages (45.1±19.8, range 20-6 - 75.8 years), including 44 men (45.3±19.4, range 20.6 – 75.8 years) and 18 women (44.4±21.4, range 20.9 - 75.3 years), we sequentially recorded SBP and DBP both during cuff inflation and cuff deflation using Korotkoff sounds. Results. There was a significant (p<0.0001) increase in SBP from 122.8±13.2 to 127.6±13.0 mmHg and a significant (p=0.0001) increase in DBP from 70.0±9.0 to 77.5±9.7 mmHg. Of the 62 subjects, 51 showed a positive increase in SBP (0 to 14 mmHg) and 11 subjects showed a reduction (-0.3 to -7 mmHg). The average differences for SBP and DBP estimates derived as the cuff inflates and those derived as the cuff deflates were 4.8±4.6 mmHg and 2.5±4.6mmHg, not dissimilar to the differences reported between IABP and NIBP measurements. Although we could not develop multiparameter linear or non-linear models to explain this phenomenon we have clearly demonstrated through ANOVA tests that both body mass index (BMI) and pulse wave velocity (PWV) are implicated, supporting the hypothesis that the phenomenon is associated with age, higher BMI and stiffer arteries. Significance. The implications of this study are that brachial sphygmomanometry carried out during cuff inflation could be more accurate than measurements carried out as the cuff deflates. Further research is required to validate these results with intra-arterial blood pressure measurements.
Content may be subject to copyright.
Physiological Measurement
ACCEPTED MANUSCRIPT • OPEN ACCESS
Measuring blood pressure from Korotkoff sounds as the brachial cuff
inflates on average provides higher values than when the cuff deflates
To cite this article before publication: Branko G Celler
et al
2024
Physiol. Meas.
in press https://doi.org/10.1088/1361-6579/ad39a2
Manuscript version: Accepted Manuscript
Accepted Manuscript is “the version of the article accepted for publication including all changes made as a result of the peer review process,
and which may also include the addition to the article by IOP Publishing of a header, an article ID, a cover sheet and/or an ‘Accepted
Manuscript’ watermark, but excluding any other editing, typesetting or other changes made by IOP Publishing and/or its licensors”
This Accepted Manuscript is © 2024 The Author(s). Published on behalf of Institute of Physics and Engineering in Medicine by IOP
Publishing Ltd.
As the Version of Record of this article is going to be / has been published on a gold open access basis under a CC BY 4.0 licence, this Accepted
Manuscript is available for reuse under a CC BY 4.0 licence immediately.
Everyone is permitted to use all or part of the original content in this article, provided that they adhere to all the terms of the licence
https://creativecommons.org/licences/by/4.0
Although reasonable endeavours have been taken to obtain all necessary permissions from third parties to include their copyrighted content
within this article, their full citation and copyright line may not be present in this Accepted Manuscript version. Before using any content from this
article, please refer to the Version of Record on IOPscience once published for full citation and copyright details, as permissions may be required.
All third party content is fully copyright protected and is not published on a gold open access basis under a CC BY licence, unless that is
specifically stated in the figure caption in the Version of Record.
View the article online for updates and enhancements.
This content was downloaded from IP address 119.13.208.20 on 03/04/2024 at 14:23
1
Measuring blood pressure from Korotkoff sounds as the
brachial cuff inflates on average provides higher values than
when the cuff deflates
Branko G. Celler
Biomedical Systems Research Laboratory in the School of Electrical Engineering and
Telecommunications, the University of New South Wales, Sydney, NSW 2052, Australia.
E-mail: b.celler@unsw.edu.au
Ahmadreza Argha
Graduate School of Biomedical Engineering, the University of New South Wales, Sydney,
NSW 2052, Australia.
E-mail: a.argha@unsw.edu.au
Abstract.
Objectives.
In this study, we test the hypothesis that if, as demonstrated in a previous study, brachial
arteries exhibit hysteresis as the occluding cuff is deflated and fail to open until cuff
pressure (CP) is well below true intra-arterial blood pressure (IAPB), estimating systolic
(SBP) and diastolic blood pressure (DBP) from the presence of Korotkoff sounds as CP
increases may eliminate these errors and give more accurate estimates of SBP and DBP
relative to IABP readings.
Approach. In 62 subjects of varying ages (45.1±19.8, range 20-6 - 75.8 years), including 44
men (45.3±19.4, range 20.6 75.8 years) and 18 women (44.4±21.4, range 20.9 - 75.3
years), we sequentially recorded SBP and DBP both during cuff inflation and cuff deflation
using Korotkoff sounds.
Results. There was a significant (p<0.0001) increase in SBP from 122.8±13.2 to
127.6±13.0 mmHg and a significant (p=0.0001) increase in DBP from 70.0±9.0 to
77.5±9.7 mmHg. Of the 62 subjects, 51 showed a positive increase in SBP (0 to 14 mmHg)
and 11 subjects showed a reduction (-0.3 to -7 mmHg). The average differences for SBP and
DBP estimates derived as the cuff inflates and those derived as the cuff deflates were
4.4.6 mmHg and 2.5±4.6mmHg, not dissimilar to the differences reported between IABP
and NIBP measurements. Although we could not develop multiparameter linear or non-linear
models to explain this phenomenon we have clearly demonstrated through ANOVA tests that
both body mass index (BMI) and pulse wave velocity (PWV) are implicated, supporting the
hypothesis that the phenomenon is associated with age, higher BMI and stiffer arteries.
Significance. The implications of this study are that brachial sphygmomanometry carried out
during cuff inflation could be more accurate than measurements carried out as the cuff
deflates. Further research is required to validate these results with intra-arterial blood
pressure measurements.
Page 1 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
2
1.
Introduction
Brachial cuff sphygmomanometry involves the placement of an appropriately sized cuff on
the upper arm which is inflated until the underlying brachial artery is fully occluded and blood
ceases to flow (Korotkoff 1905). As the cuff pressure (CP) is slowly reduced, conventional
wisdom states that the CP at which the first Korotkoff sounds (KS) are heard by a trained
operator using a sensitive stethoscope corresponds closely to systolic blood pressure (SBP)
and the termination of the KS aligns closely to diastolic blood pressure (DBP). This is the
basis for the calibration of all non-invasive blood pressure (NIBP) monitoring devices
according to the recently released universal standard (ISO 81060-2:2018) (Stergiou et al.
2018) for
the validation of NIBP monitors. This standard states that “A device will be
considered acceptable if its estimated probability of a tolerable error ( 10 mmHg) is
85%”. Given that public health consequences (Psaty et al. 1997) of even errors as small ±5
mmHg can be significant, the relatively large errors of 10 mmHg considered acceptable
suggest some inherent problems with the method of classical sphygmomanometry. A
related explanation may be that these broad limits reflect the long known and well
documented (Dankel et
al. 2019; Kallioinen et al. 2017; Picone et al. 2017; Seidlerova´ et al. 2019) reality that
NIBP measurements generally underestimate SBP and overestimate DBP when compared to
simultaneous intra-arterial measurement. In the study conducted by Kallioinen et al. (2017),
significant effects of individual sources of error ranging from -23.6 to +33 mmHg for SBP
and -14 to +23 mmHg for DBP were recorded.
KS are fundamental to the calibration of all NIBP monitors. Blank et al. (1988) used
wideband (0.1–2000 Hz) external pulse recording during cuff deflation to identify three
distinct phases of the KS (K1, K2, K3). K1 is a low-amplitude, low frequency signal <
20 Hz, that is present with cuff pressures above SBP, K2 is a triphasic signal appearing at
SBP and disappearing at DBP, which approximately corresponds to the audible KS and has
frequency spectra in the range of 20-80 Hz and K3 appears with cuff pressure between SBP
and DBP and continues to be present below DBP. Using Millar catheters for greater accuracy,
Blanks also noted that he onset and disappearance of K2 was closely correlated to SBP and
DBP derived from auscultation, with mean values of SBP generally higher than SBP and
lower than DBP derived from auscultation.
Blank et al. (1988) showed also that the visual technique detects the onset
(disappearance) of K2 a few beats before (after) the Korotkoff sound becomes audible
(inaudible) at SBP (DBP) and as a result gives closer values to intra-arterial BP determinations
than the conventional auscultatory technique. Furthermore, numerous studies including ours
(Celler, Le, Basilakis, et al. 2017) have demonstrated that there are significant inter-operator
differences in estimating BP using sphygmomanometry, especially with the determination of
DBP.
Our previous studies (Celler, Basilakis, et al. 2015; Celler, Le, Basilakis, et al. 2017)
suggest that the accuracy of auscultatory sphygmomanometry is dependent on (i) the sensitivity
of the stethoscope, (ii) the hearing acuity of the operator, and (iii) the amplitude and particular
waveform morphometry of the Korotkoff sounds. It was also shown that complete silence
Page 2 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
3
occurs on average in less than 50% of cases, making the determination of DBP particularly
unreliable when Korotkoff sounds are listened to with stethoscopes. Other studies in the
literature of automatic or semi-automatic auscultatory NIBP estimation methods (B. S. Alpert
2018; Chu et al. 2017; H. Wu et al. 2016; Zhang et al. 2019) also showed that visual
auscultatory technique is significantly correlated with the manual auscultation.
In a recently published paper (Celler, Butlin, et al. 2021) reporting on invasive
experiments where the CP, the KS and the intra-arterial blood pressure (IABP) were
simultaneously recorded, the authors demonstrated conclusively that the K2 KS did indeed
correlate accurately to the very first blood flow from the occluded artery and ceased when
the CP was approximately at DBP. The most important result of this study however was the
observation that as the CP was reduced, first blood flow and hence K2 KS, were in many
cases delayed until well below true SBP, up to 24 mmHg. We concluded that following
occlusion of the brachial artery, there was a variable delay in the re-opening of the artery
probably associates with low arterial compliance. We noted in one subject with particularly
low BP, and in subjects following the infusion of glyceryl trinitrate (GTN) (100-200 mg), that
the brachial artery could indeed also re-open prematurely. These data confirm and to large
measure explain the individual subject to subject variability in NIBP measurements reported
in many studies (Picone et al. 2017; Saherwala et al. 2018).
If indeed there is a delay in the reopening of the brachial artery following occlusion by
an inflated cuff a logical hypothesis follows, that if the cuff is inflated whilst recording K2 KS,
the first sounds heard should be associated with DBP and the very last sounds heard should
be heard just before occlusion of the artery as the CP exceeds SBP, thus obviating the delayed
re-opening observed when the CP falls post occlusion of the artery.
Aims and Objectives
The aims of this study are to investigate whether SBP and DBP estimates based on the
presence of K2 KS are different when the cuff pressure is linearly deflated as per conventional
methods, to when the cuff pressure is linearly inflated.
Based on our hypothesis we would expect that on average the SBP on inflation will be
higher than SBP measured on deflation and that DBP should be relatively similar. We will
further explore whether non-invasive estimates of arterial compliance can explain some of the
expected variability in SBP estimation.
2.
Methods
We have developed a battery powered instrument to achieve in the non-invasive setting, a
similar study to that reported in Celler, Butlin, et al. (2021). The instrument was designed to
record CP, high fidelity broadband (< 1-500 Hz) KS using a piezo-resistive transducer, a
single lead electrocardiogram (ECG) and a photoplethysmogram (PPG) signal. The data
acquisition and the servo-control of the CP was achieved using a National Instruments USB6002
multifunction I/O module connected to a battery powered laptop computer via a USB cable. We
developed a proportional-integral (PI) controller using LabView to control the Oken Seiko air
pump (P54A02R) and a release valve to permit the CP to be increased rapidly 30-40 mmHg past
estimated SBP, and then to fall linearly at approximately 3.0 mmHg/sec to approximately 20
Page 3 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
4
mmHg, before beginning a linear inflation to the same peak CP.
Figure 1 Typical recording of IABP (Blue trace) and Cuff Pressure (Black trace), during a complete rapid inflation and slow
deflation cycle. Red lines mark SBP (134 mmHg) and dotted red lines mark DBP (67 mmHg). The blue dotted line marks the
beginning of blood flow through the brachial artery as evidenced by an initial increase in IABP.
2.1.
Experimental protocol and data acquisition
In traditional brachial sphygmomanometry the CP is inflated to approximately 30 mmHg
above the estimated SBP to ensure complete occlusion of both the brachial artery and vein.
As the cuff pressure is reduced, blood begins to flow below the cuff and a reactive hyperemia
(RH) is observed as an overshoot in the SBP. RH arises to ensure that enough oxygen flows
to the occluded area and that dead cells and metabolites are rapidly flushed from the area to
reduce possible tissue damage. In Figure 1 we note that reactive hyperemia is completed by
the time the cuff pressure is reduced to below DBP.
Blood flow restriction (BFR) via the application of external pressure to occlude venous
return and restrict arterial inflow, has been shown to increase muscular size and strength when
combined with low-load resistance exercise (Mouser et al. 2018). Following BFR blood flow
decreased in a nonlinear, stepped fashion. Blood flow decreased at 10% of occlusion and
remained constant until decreasing again at 40%, where it remained until 90% of occlusion.
From roughly 30% to 60% arterial occlusion pressure (AOP), tissue pressure is increased
underneath the cuff but remains insufficient to have a direct impact on occluding the arteries
of the arm. Finally, from roughly 60% to 100% AOP, the pressure applied by the cuffs is
impacting arterial flow directly leading at 100% to the complete occlusion of the brachial
artery.
Superficial venous pressure is low in the arms (Ochsner JR et al. 1951; Todini et al. 2012)
Page 4 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
5
typically 10–15 mmHg, and the pressures applied in this study at 10% AOP ranged from 12
to 25 mmHg. This would be sufficient in most cases to block all superficial venous return. As
pressure increases, the deep veins of the arm would also become occluded, as pressure transfer
from the cuff to the tissue is to a small degree attenuated with increased tissue depth (Graham
et al. 1993). Even though this attenuation only amounts to about 20 mmHg, that appears
sufficient to require increased pressure in order to completely occlude all venous return from
the deep veins.
The brachial veins are deep veins which share the same name of the arteries they
accompany. Pressure within the named veins is usually between 8 and 10 mmHg. Veins
have thinner walls and larger diameters than arteries with less muscle and elastic tissue. This
means that they have high vascular compliance so that the rate of change in volume with
changing pressure is high and, therefore, changes in venous blood volume produce relatively
small changes in venous distending pressure. In fact, veins have a compliance that is 30 times
that of arteries and can expand easily to accommodate large volumes of blood.
These data clearly suggest blood flow in the deep veins continues to flow until cuff
pressures increase from roughly 60% to 100% AOP. Venous engorgement is therefore unlikely
to be a significant factor given the short duration ( 20 seconds) of complete arterial occlusion
and more than 45 seconds of only partial or no occlusion as shown in Figure 1. An alternative
method of ensuring rapid venous drainage would be to elevate the subject’s arm for a few
seconds, but this procedure interfered with the operation of the servo-control system and was
therefore not adopted.
2.2.
Data acquisition
Following informed consent and the recording of basic demographics including age, gender,
height, weight, arm circumference and the distance between the mid sternum and the tip of the
index finger, subjects were required to relax comfortably seated with both feet on the ground
for a minimum of five minutes. A brachial cuff of appropriate dimensions for the patient’s arm
diameter was placed on the right upper arm and a sensitive piezoelectric transducer with flat
frequency response from < 1-500 Hz, was placed at the border of the cuff over the brachial
artery to record the KS. ECG leads were placed at the LA, LL and RL position for a Lead
III configuration. A Nellcor Compatible SpO2 Sensor (DS-100A) was placed on the right
index finger. The cuff was rapidly inflated to 30-40 mmHg above the patient’s SBP, and then
deflated linearly at a rate of 3 mmHg/sec to approximately 20 mmHg. The cuff was then
linearly inflated at the same rate under servo-control to the selected peak CP. During this
time, CP, KS, PPG and ECG signals were recorded at 1000 samples/sec. This procedure was
repeated three times, with a rest interval of five minutes between recordings.
It was later observed that a physical filter which was added after the differential pressure
transducer to reduce pump noise during inflation was causing a pressure drop during rapid
inflation or deflation. However, because of the low rate of inflation the bias introduced was
measured as being very low, in the order of a few mmHg. Nonetheless a digital filter was
designed to compensate for this pressure drop and was applied to all data recorded.
Page 5 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
6
The monitoring instrument and the research protocol were approved by the University of
New South Wales Human Research Ethics Committee (HC200066) on the 27th July 2020.
This research was conducted in accordance with the principles embodied in the Declaration
of Helsinki and in accordance with local statutory requirements. All participants gave
written informed consent to participate in the study.
2.3.
Data Analysis
All data processing and analysis were carried out using MATLAB (2020b). The broadband
Korotkoff signal was high pass filtered at 20 Hz according to the K2 algorithm proposed by
Blank et al. (1988) to accentuate the high frequency (HF) components of the KS (HF KS) that
are in the audible range and are related to blood flow in the artery. To improve the signal to
noise ratio and facilitate the automated detection of peaks, the root mean square (rms) energy
of the K2 Korotkoff signal was calculated and a moving average zero phase digital filter using
a Hamming window of 80 msec width, was applied by processing the input data in both the
forward and reverse directions.
Peak detection of the Korotkoff energy was carried out using the findpeaks Matlab
command with a threshold of detection set at 10% of the peak energy signal. A similar
command was used to detect the QRS peaks of the ECG and the peaks of the PPG signal.
From the R peak of the ECG, the S peak was identified using gradient methods. The PPG data
were filtered using the bandpass command (0.5-10 Hz), which performs zero phase filtering
on the input using a bandpass filter with a stopband attenuation of 60 dB. The foot of the PPG
was determined as the maximum value of the second derivative of the PPG wave.
Pulse transit time (PTT) is the sum of the pre-ejection period (PEP) and the vascular
transit time (VTT) and is often incorrectly used to calculate the pulse wave velocity (PWV).
In our study the brachial PWV study was calculated according to the method of Kortekaas et
al. (2018) where the VTT was estimated as the interval from the S wave of the ECG to the foot
of the PPG signal. The S wave closely coincides with the opening of the aortic valve and the
ejection of blood as evidenced using Doppler mode echocardiography. Dividing the distance
from mid-sternum to the tip of the index finger in meters, by the VTT in seconds provides an
estimate of PWV in meters/sec. The PWV was calculated and averaged from between three
and five consecutive cardiac beats.
2.4.
Determination of SBP and DBP
The determination of SBP and DBP points both during cuff deflation and cuff deflation were
semi-automated, by detecting all peaks of the Korotkoff energy that were larger than 10% of
the highest energy recorded. Noise and artefact could on occasion generate energy signals
greater than the 10% threshold, but these could be easily ignored if not part of a consistent
ensemble of signals one cardiac period apart. The SBP and DBP points were then manually
selected using some simple rules. These include no adjacent peaks within two adjacent cardiac
beats and participation in an ensemble of signals separated by one cardiac period, generally
increasing and then decreasing in amplitude. In due course this process can be fully automated
using deep learning algorithms (Argha and Celler 2019; Argha, Celler, and Lovell 2020a;
Page 6 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
7
Argha, Celler, and Lovell 2020b; Argha, J. Wu, et al. 2019; Celler, Le, Argha, et al. 2019).
2.5.
Statistical analysis
All data were tested for normality using the single sample Kolmogorov-Smirnov goodness
of fit hypothesis test and the Lilliefors’ composite goodness-of-fit test. Continuous variables
such as % changes were tested with a one sample t-test with a null hypothesis of “mean
is zero”. For data that was not normally distributed the Wilcoxon rank sum test for equal
medians was used. Before and after analysis of the same variables was carried out using
the paired t-test. One-way analysis of variance (ANOVA) was carried out using the anova1
command when comparing the means of more than two groups of data. Multi-way (n-
way) ANOVA (anovan) was used to determine whether particular categorical variables could
explain the variance in an output variable.
Figure 2 Example of one experimental record showing results when CP is first reduced linearly and is then increased to the
same peak. Peak detection of Korotkoff energy is shown by black circles. The Horizontal dotted line shows the 10% threshold
below which local peaks are ignored. SBP and DBP points on both falling and rising phases of CP are identified by vertical lines
and dotted lines respectively. For this subject BP recorded was 105/67 mmHg as cuff was deflated and 124/70 mmHg as cuff
was inflated, a difference in SBP of 19 mmHg.
3.
Results
All subjects were volunteers from staff and students at the University of New South Wales
or family and friends of the authors. Sixty two subjects with an average age of 45.1±19.8
years (range 20.6-75.8 years) were tested, including 44 men (45.3±19.4, range 20.6-75.8
years) and 18 women (44.4±21.4, range 20.9-75.3 years). Fifty-eight subjects had three
recordings and four had only two recordings. No recordings were rejected for any reason
other than technical problems such as excessive noise or artefacts. Each recording was
analysed individually. However as there were no significant differences between recordings
(p»0.05), they were averaged and analysed as a single record.
Page 7 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
8
Figure 2 is an example of a single experimental record and Table 1 summarises all significant
variables for all subjects, male subjects, and female subjects respectively. All data were tested
for normality. In 62 subjects the mean difference between SBP recorded as CP decreases,
and SBP as CP increases was 4.8±46 mmHg, of which fifty one were positive (range 0 -
14 mmHg) and eleven were negative (range -0.3 - to 70 mmHg). These data effectively
confirm the core hypothesis motivating this study. Of those who were negative four were in
the age bracket < 30, two were between 30 and 40 years, and five were older than 60.
It is instructive to look at Figure 2 in some detail. The slope of the CP traces down and
up is servo-controlled at approximately 3.5 mmHg/sec and were not significantly different
(p=0.5977) across all subjects. The first Korotkoff energy pulse detected as the cuff deflates
is always significantly (p<0.0001) larger as a fraction of the peak (0.22±0.08), than for the
rising CP phase (0.19±0.12). This suggests that as the cuff deflates the break-through pulse
of blood flow is more vigorous than for the rising phase where the occlusion of the artery is
progressively more complete and blood pressure is pumping against an increasingly higher
pressure as the brachial circulation is increasingly occluded by the increasing cuff pressure.
We note in Table 1 that despite SBP increasing as age progresses, none of the SBPdiff
values are significantly different (p<0.05) for the three age group 30, 30-60 and > 50.
However when we consider age and BMI, DBPdiff of the subgroup of subjects of age 50
and BMI 25 relative to the older age group, aged > 50 and BMI > 25 is significantly
different (p=0.0067) increasing from 1.4±3.5 to 5.2±4.2 mmHg. Both BMI (p<0.0001) and
PWV (p=0.0029) also increase significantly between these two groups. PWV increases
significantly across all five sub-categories. These results are summarised as box plots in
Figure 3 and show that subjects aged > 50 and with a BMI > 25 consistently have SBPdiff
that are positive. The younger cohort and those with low BMI, in contrast do on occasion
demonstrate negative values of SBPdiff.
Page 8 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
9
Table 1Summary data for all subjects (N=62) and sub-groups (1-8) by Age, BMI and PWV values (SBPdiff = SBP Up-SBP
Down, (DBPdiff=DBP Up-DBP Down)
# Group N Age
(years)
SBP
Down
(mmHg)
DBP
Down
(mmHg)
SBP
Up
(mmHg)
DBP
Up
(mmHg)
SBP
diff
(mmHg)
DBP
diff
(mmHg)
BMI
(Kg/m2)
PWV
(m/sec)
1 All Ages 62
122.8
1
70.0
2
127.6
1
77.5
2
4.8
2.5
24.8
7.9
(13.2)
(9.0)
(13.0)
(9.7)
(4.6)
(4.6)
(3.6)
(1.7)
2 All Males 44
125.83
76.64
130.93
79.54
5.1
2.9
25.3
8.1
(13.2)
(8.8)
(12.6)
(10.1)
(4.8)
(4.5)
(3.5)
(1.8)
3 All Females 18
115.4
5
72.8
119.5
5
72.8
4.1
1.6
23.4
7.2
(10.1)
(6.6)
(10.3)
(6.6)
(3.9)
(4.8)
(3.8)
(1.4)
4 All LE 30 21
119.0
6
70.4
123.5
6
70.8
4.5
0.3
23.6
7.1
(13.3)
(7.6)
(13.3)
(7.6)
(4.2)
(4.4)
(3.8)
(1.0)
5
All GT 30
LT 6 0
22
119.9
7
78.6
8
125.0
7
81.9
8
5.1
3.3
25.3
7.7
(11.4)
(8.5)
(10.8)
(8.1)
(3.3)
(3.1)
(3.4)
(0.9)
6 All GE 60 19
130.3
9
76.0
10
135.1
9
80.0
10
4.8
4.0
25.5
8.8
(12.5)
(9.3)
(12.3)
(9.7)
(6.2)
(5.4)
(3.6)
(2.4)
7
All LE50
BMI LE25
22
115.1
11
71.3
119.4
11
72.6
4.2
1.4
22.0
6.8
(10.8)
(8.1)
(10.7)
(8.4)
(4.0)
(3.5)
(2.0)
(1.0)
8
All GT50
BMI GT25
14
124.2
12
73.7
13
130.7
12
76.7
13
6.5
5.2
28.0
8.8
(11.3)
(6.4)
(11.3)
(9.3)
(5.5)
(4.2)
(2.0)
(2.6)
p-Values: differences in SBP and DBP within groups, paired t-test
1<0.0001, 20.0001, 3<0.0001, 40.0001, 50.0003, 60.0001, 7<0.0001, 80.0001, 9<0.0033,
100.0046, 110.0001, 120.0001, 130.008
p-Values: differences in SBP Down across groups
1-30.0311, 1-70.0167, 2-30.0039, 2-70.0017, 3-60.0007, 3-80.0307, 4-60.0171, 5-60.0199, 6-70.0003,
7-80.0250
p-Values: differences in DBP Down across groups
1-40.0394, 2-30.0301, 2-40.0075, 2-70.0197, 3-50.0095, 3-60.0377, 4-50.0019, 4-60.0109, 5-70.0054,
6-70.0257
p-Values: differences in SBP Up across groups
1-30.0178, 1-60.0495, 1-70.0092, 2-30.0013, 2-40.0338, 2-70.0005, 3-60.0002, 3-80.0078, 4-60.0090,
5-60.0134, 6-70.0001, 7-80.0057
p-Values: differences in DBP Up across groups
1-70.0377, 2-30.0125, 2-40.0009, 2-70.0081, 3-50.0005, 3-60.0025, 3-80.0281, 4-5<0.0001,
4-60.0002, 4-80.0056, 5-70.0006, 6-70.0021, 7-80.0379
p-Values: Differences in BMI across groups (superscripts), t-test
1-70.0011, 1-80.0028, 2-70.0001, 2-80.0102, 3-80.0005, 4-80.0005, 5-70.0003, 5-80.0143, 6-70.0004,
6-80.0261, 7-8<0.0001
p-Values: Differences in PWV across groups (superscripts), t-test
1-70.0103, 2-40.0230, 2-70.0037, 3-60.0417, 4-60.0039, 4-80.0115, 5-70.0085, 6-70.0008, 7-80.0029
p-Values: Differences in DBPdiff across groups (superscripts), t-test
2-40.0368, 3-80.0401, 4-50.0155, 4-60.0104, 4-80.0034, 7-80.0067
Page 9 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
10
Figure 3: Boxplot of eight categories based on gender, age and
BMI levels.
For DBP, Bland-Altman plots (not shown) give a mean difference of 2.5mmHg, an r2 value
of 0.7788 and a sum of squares error (SSE) of 5.3 mmHg
3.1.
Multiparameter linear and non-linear modelling
Multiparameter linear regression (MLR) is used to model the linear relationship between
the explanatory (independent) variables and response (dependent) variable. In this study we
used the fitlm function in MATLAB to explore which independent variables could predict the
SBPdiff. Attempts were made to develop MLR models to describe SBPdiff as a function of
numerous variables including BMI, PWV and heart rate (HR) but none gave p values < 0.05
and acceptable root mean squared error. Unsupervised learning analysis based on k-means,
kmedoids (Arora, Varshney, et al. 2016) and Gaussian mixture (Maugis et al. 2009) clustering
Figure 3 Boxplot of eight categories based on gender, age and BMI levels.
Figure 4 Bland-Altman plots showing differences between SBP recorded as the cuff inflates (SBP UP) and as the cuff
deflates (SBP DOWN)
Page 10 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
11
was similarly unsuccessful.
3.2.
Three-way analysis of variance using anovan
Three-way ANOVA (anovan) is used to determine if there is an interaction effect between
three independent variables (Age, PWV and BMI) on a continuous dependent variable
SBPdiff. That is, we test if a three-way interaction exists. We use all the available data (N=62)
and segment each independent variable into two factors, being < or a selected break point.
We then test the interaction for a number of breakpoints. Table 2 shows the results for Age,
BMI and PWV with breakpoints set at 50 years of age, a BMI of 25 and mean PWV (7.86
m/sec) respectively.
Table 2 Three-way analysis of variance with Age, BMI and PWV as independent categorical variables.
Source
Sum Sq.
d.f.
Mean Sq.
F
p Va l u e
AgeCat45
17.1
1
17.095
0.91
0.3446
BMICat25
52.65
1
52.645
2.8
0.1000
PWVCatMean
8.72
1
8.715
0.46
0.4989
AgeCat45*BMICat25
0.91
1
0.907
0.05
0.8270
AgeCat45*PWVCatMean
17.32
1
17.319
0.92
0.3415
BMICat25*PWVCatMean
141.29
1
141.291
7.51
0.0083
Error
1034.61
55
18.811
Total
1272.41
61
The low p values for BMICat25*PWVCatMean suggest that SBPdiff, the dependent variable
is significantly influenced or interacts primarily with the only two independent variables
BMI and PWV at the selected breakpoints.
4.
Discussion
In this study we propose a novel new method for brachial sphygmomanometry whereby we
record the K2 KS as the brachial cuff is inflated linearly under servo-control. An extensive
literature review revealed only a small number of publications where consideration was given
to using either the rising or falling phase of CP inflation or deflation for estimating NIBP.
B. S. Alpert (2007) undertook a clinical evaluation of the Welch Allyn SureBP algorithm
for automated blood pressure measurement. Nukita et al. (2020) reported that the repeated
estimation of BP on the inflation phase can significantly reduce the risk of subcutaneous
hemorrhage. Yamashita and Irikoma (2018) conclude that NIBP detection on cuff inflation
detected hypotension faster than conventional NIBP without compromising the reliability of
measurement, thus leading to early treatment of maternal hypotension and the prevention
of adverse events related to the mother and the fetus. The efficacy of a new BP monitor
(Takahashi et al. 2020), based on detection of BP during inflation was evaluated in a
randomised controlled study.
However, in all the quoted studies (B. S. Alpert 2007; Nukita et al. 2020; Takahashi et al.
2020; Yamashita and Irikoma 2018), the algorithm for detecting SBP and DBP is based on
the oscillometric method and the rationale for choosing the rising phase of cuff pressure was
Page 11 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
12
in large part, to shorten the time required to take a measurement, to reduce the maximal
inflation pressure and improve patient comfort and outcomes. None discuss the fundamental
questions raised in this study. A technical paper (Vazquez et al. 2021) describes a sensor
fused BP measuring device capable of recording KS in inflationary curves, with the authors
noting that “The pump noise makes it difficult to get a reliable automatic auscultatory
reading”. Zheng, Di Marco, et al. (2012) and Zheng, Pan, et al. (2013) derived the
oscillometric waveform envelope (OWE) (Forouzanfar et al. 2015) for both the rising phase
of CP and then the falling phase. They found that during the rising phase the OWE was
shifted upwards and resulted in a higher estimation of mean arterial pressure (MAP). These
results are indirectly consistent with the conclusions of this study.
It is well recognized that BP is variable over time and changes subject to emotional
and environmental factors (Adams and Leverland 1985; Soueidan et al. 2010; Soueidan et
al. 2012). Intrinsic physiological oscillations in BP can also cause shifts of up to 20 mmHg
within a few heartbeats (Hansen and Staber 2006). In this study we took great care to maintain
all environmental variables comfortable and constant and followed the recommendations of
the international standard body (Stergiou et al. 2018) by waiting for five minutes between
sequential experiments. Although differences between the three measurements were not
significant, we did observe in some cases significant changes in SBP and DBP between the
first and the last measurement, as the subject relaxed and became less mentally stressed.
We minimize this effect by averaging the results of the three experiments, but the issue of
maintaining a constant state of mental alertness during the recording of NIBP warrants further
consideration.
5.
Conclusion
In this study we conclusively demonstrate that SBP estimated from K2 KS and brachial cuff
sphygmomanometry as the CP is increased is generally higher by -7-14 mmHg than SBP
estimated when CP is decreasing. In a previous study (Celler, Butlin, et al. 2021) where
we simultaneously recorded both invasive and NIBP in sedated subjects, we discovered that
following occlusion of the brachial artery at CPs higher than SBP, as the CP is reduced, the
artery fails to re-open until CP is well below SBP. Typical errors were found to be from 4
to 24 mmHg, not dissimilar to the range (7 to 14 mmHg) observed in this study, and in
broad agreement with the conclusion of 62 separate studies (Dankel et al. 2019) that indirect
measures of SBP underestimated true SBP by an average of 4.55 mmHg and overestimated
DBP by 6.20 mmHg (95% CI = 5.09 to 7.31).
Although we could not develop multiparameter linear or non-linear models to explain
this phenomenon we have clearly demonstrated through ANOVA that both BMI and PWV
are implicated, supporting the hypothesis that the phenomenon is associated with age, higher
BMI and stiffer arteries, as evidenced by higher PWV.
Whilst further work needs to be done duplicating this study in an invasive experiment
with sequential deflation and then inflation of the cuff, the data presented provide a strong
suggestion that by recording K2 KS as the CP is inflated may allow more accurate estimates
of true intra-arterial systolic blood pressure independent of age, gender, BMI or arterial
Page 12 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
13
compliance.
The implications of this study are potentially quite profound as current international standards
(ISO 81060-2:2018) (Stergiou et al. 2018) for the calibration of all NIBP monitor require two
expert operators to listen to the KS as the CP deflates. Clearly this new method for brachial
cuff sphygmomanometry would require a radical amendment to the international standards and
a change in the operating modality of every NIBP monitor in the market.
Acknowledgement
The authors wish to acknowledge the contribution made to this study by sixty-two subjects
of varying age and gender who willingly contributed their time.
References
Adams, C. E., & Leverland, M. B. (1985). Environmental and behavioral factors that can
affect blood pressure. The Nurse Practitioner, 10(11), 39–40.
Alpert, B. S. (2007). Clinical evaluation of the Welch Allyn SureBP algorithm for automated
blood pressure measurement. Blood pressure monitoring, 12(4), 215–218.
Alpert, B. S. (2018). The Accutension Stetho, an automated auscultatory device to validate
automated sphygmomanometer readings in individual patients. Journal of Human
Hypertension, 32(6), 455–459.
Argha, A., & Celler, B. G. (2019). Blood pressure estimation from time-domain features of
oscillometric waveforms using long short-term memory recurrent neural networks.
IEEE Transactions on Instrumentation and Measurement, 69(6), 3614–3622.
Argha, A., Celler, B. G., & Lovell, N. H. (2020a). Artificial Intelligence Based Blood Pressure
Estimation From Auscultatory and Oscillometric Waveforms: A Methodological
Review. IEEE Reviews in Biomedical Engineering.
Argha, A., Celler, B. G., & Lovell, N. H. (2020b). A Novel Automated Blood Pressure
Estimation Algorithm Using Sequences of Korotkoff Sounds. IEEE Journal of
Biomedical and Health Informatics.
Argha, A., Wu, J., Su, S. W., & Celler, B. G. (2019). Blood Pressure Estimation From Beat-
by-Beat Time-Domain Features of Oscillometric Waveforms Using Deep-Neural-
Network Classification Models. IEEE Access, 7, 113427–113439.
Arora, P., Varshney, S. et al. (2016). Analysis of k-means and k-medoids algorithm for big
data. Procedia Computer Science, 78, 507–512.
Blank, S. G., West, J. E., Mu¨ ller, F., Cody, R. J., Harshfield, G. A., Pecker, M. S., Laragh, J.,
& Pickering, T. G. (1988). Wideband external pulse recording during cuff deflation: a
new technique for evaluation of the arterial pressure pulse and measurement of blood
pressure. Circulation, 77(6), 1297–1305.
Celler, B. G., Basilakis, J., Goozee, K., & Ambikairajah, E. (2015). Non-Invasive
measurement of blood pressure-Why we should look at BP traces rather than
listen to Korotkoff sounds. 2015 37th Annual International Conference of the IEEE
Engineering in Medicine and Biology Society (EMBC), 5964–5967.
Page 13 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
14
Celler, B. G., Butlin, M., Argha, A., Tan, I., Yong, A. S. C., & Avolio, A. (2021). Are
Korotkoff sounds reliable markers for accurate estimation of systolic and diastolic
pressure using brachial cuff sphygmomanometry. IEEE Transactions on Biomedical
Engineering.
Celler, B. G., Le, P. N., Argha, A., & Ambikairajah, E. (2019). GMM-HMM-Based
Blood Pressure Estimation Using Time-Domain Features. IEEE Transactions on
Instrumentation and Measurement, 69(6), 3631–3641.
Celler, B. G., Le, P. N., Basilakis, J., & Ambikairajah, E. (2017). Improving the quality
and accuracy of non-invasive blood pressure measurement by visual inspection and
automated signal processing of the Korotkoff sounds. Physiological measurement,
38(6), 1006.
Chu, G., Zhang, Z., Xu, M., Huang, D., & Dai, Q. (2017). Validation of a smartphone
auscultatory blood pressure kit Accutension XYZ-110 in adults according to the
ANSI/AAMI/ISO 81060-2: 2013 standard. Blood Pressure Monitoring, 22(5), 290
294.
Dankel, S. J., Kang, M., Abe, T., & Loenneke, J. P. (2019). A meta-analysis to determine the
validity of taking blood pressure using the indirect Cuff method. Current hypertension
reports, 21(1), 11.
Forouzanfar, M., Dajani, H. R., Groza, V. Z., Bolic, M., Rajan, S., & Batkin, I. (2015).
Oscillometric blood pressure estimation: past, present, and future. IEEE reviews in
biomedical engineering, 8, 44–63.
Graham, B., Breault, M. J., Mcewen, J. A., & Mcgraw, R. W. (1993). Occlusion of arterial
flow in the extremities at subsystolic pressures through the use of wide tourniquet
cuffs. Clinical Orthopaedics and Related Research®, 286, 257–261.
Hansen, S., & Staber, M. (2006). Oscillometric blood pressure measurement used for
calibration of the arterial tonometry method contributes significantly to error.
European Journal of Anaesthesiology— EJA, 23(9), 781–787.
Kallioinen, N., Hill, A., Horswill, M. S., Ward, H. E., & Watson, M. O. (2017). Sources of
inaccuracy in the measurement of adult patients’ resting blood pressure in clinical
settings: a systematic review. Journal of hypertension, 35(3), 421.
Korotkoff, N. (1905). A contribution to the problem of methods for the determination of the
blood pressure. Rep Imperial Milit-Med Acad St Petersburg, 11, 365.
Kortekaas, M. C., van Velzen, M. H., Gru¨ ne, F., Niehof, S. P., Stolker, R. J., & Huygen, F. J.
(2018). Small intra-individual variability of the pre-ejection period justifies the use of
pulse transit time as approximation of the vascular transit. PloS one, 13(10), e0204105.
Maugis, C., Celeux, G., & Martin-Magniette, M.-L. (2009). Variable selection for clustering
with Gaussian mixture models. Biometrics, 65(3), 701–709.
Mouser, J. G., Ade, C. J., Black, C. D., Bemben, D. A., & Bemben, M. G. (2018). Brachial
blood flow under relative levels of blood flow restriction is decreased in a nonlinear
fashion. Clinical physiology and functional imaging, 38(3), 425–430.
Page 14 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
15
Nukita, S., Nakano, S., Morimoto, Y., & Shiramoto, H. (2020). Inflationary noninvasive blood
pressure measurement reduces the incidence of subcutaneous hemorrhage. Blood
pressure monitoring, 25(1), 39–41.
Ochsner JR, A., Colp JR, R., & Burch, G. (1951). Normal blood pressure in the superficial
venous system of man at rest in the supine position. Circulation, 3(5), 674–680.
Picone, D. S., Schultz, M. G., Otahal, P., Aakhus, S., Al-Jumaily, A. M., Black, J. A.,
Bos, W. J., Chambers, J. B., Chen, C.-H., Cheng, H.-M. et al. (2017). Accuracy of
cuff-measured blood pressure: systematic reviews and meta-analyses. Journal of the
American College of Cardiology, 70(5), 572–586.
Psaty, B. M., Smith, N. L., Siscovick, D. S., Koepsell, T. D., Weiss, N. S., Heckbert, S. R.,
Lemaitre, R. N., Wagner, E. H., & Furberg, C. D. (1997). Health outcomes associated
with antihypertensive therapies used as first-line agents: a systematic review and meta-
analysis. Jama, 277(9), 739–745.
Saherwala, A. A., Stutzman, S. E., Osman, M., Kalia, J., Figueroa, S. A., Olson, D. M., &
Aiyagari, V. (2018). Correlation of noninvasive blood pressure and invasive intra-
arterial blood pressure in patients treated with vasoactive medications in a neurocritical
care unit. Neurocritical care, 28(3), 265–272.
Seidlerova´, J., T u˘m o v a´ , P., Rokyta, R., & Hromadka, M. (2019). Factors influencing the
accuracy of non-invasive blood pressure measurements in patients admitted for
cardiogenic shock. BMC cardiovascular disorders, 19(1), 1–10.
Soueidan, K., Chen, S., Dajani, H. R., Bolic, M., & Groza, V. (2010). The effect of blood
pressure variability on the estimation of the systolic and diastolic pressures. 2010 IEEE
International Workshop on Medical Measurements and Applications, 14–18.
Soueidan, K., Chen, S., Dajani, H. R., Bolic, M., & Groza, V. (2012). Augmented blood
pressure measurement through the noninvasive estimation of physiological arterial
pressure variability. Physiological Measurement, 33(6), 881.
Stergiou, G. S., Alpert, B., Mieke, S., Asmar, R., Atkins, N., Eckert, S., Frick, G., Friedman,
B., Graßl, T., Ichikawa, T. et al. (2018). A universal standard for the validation
of blood pressure measuring devices: Association for the Advancement of Medical
Instrumentation/European Society of Hypertension/International Organization for
Standardization (AAMI/ESH/ISO) Collaboration Statement. Hypertension, 71(3),
368374.
Takahashi, K., Asai, T., & Okuda, Y. (2020). Efficacy of a new blood pressure monitor
(inflationary non-invasive blood pressure, iNIBPTM): a randomised controlled study.
Anaesthesia, 75(1), 37–44.
Todini, A., Pergolini, D., Morucci, M., D’Andrea, V., Paiella, M., Cantisani, V., & Biancari, F.
(2012). Brachial venous pressure measurement for diagnosis of pulmonary embolism.
La Clinica Terapeutica, 163(5), e287–97.
Vazquez, E. A., Ewert, D., Jorgenson, D., & Sand, M. (2021). Sensor Fused Blood Pressure
Measuring Device Capable of Recording Korotkoff Sounds in Inflationary Curves.
Frontiers in Biomedical Devices, 84812, V001T02A001.
Page 15 of 16 AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
16
Wu, H., Wang, B., Zhu, X., Chu, G., & Zhang, Z. (2016). A new automatic blood pressure
kit auscultates for accurate reading with a smartphone: a diagnostic accuracy study.
Medicine, 95(32).
Yamashita, A., & Irikoma, S. (2018). Comparison of inflationary non-invasive blood pressure
(iNIBP) monitoring technology and conventional deflationary non-invasive blood
pressure (dNIBP) measurement in detecting hypotension during cesarean section. JA
clinical reports, 4(1), 1–3.
Zhang, Z., Xi, W., Wang, B., Chu, G., & Wang, F. (2019). A convenient method to verify
the accuracy of oscillometric blood pressure monitors by the auscultatory method: A
smartphone-based app. The Journal of Clinical Hypertension, 21(2), 173–180.
Zheng, D., Di Marco, L. Y., & Murray, A. (2012). Blood pressure difference between
the measurements taken during cuff inflation and deflation. 2012 Computing in
Cardiology, 729–732.
Zheng, D., Pan, F., & Murray, A. (2013). Effect of mechanical behaviour of the brachial
artery on blood pressure measurement during both cuff inflation and cuff deflation.
Blood pressure monitoring, 18(5), 265–271.
Page 16 of 16AUTHOR SUBMITTED MANUSCRIPT - PMEA-105515.R1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Accepted Manuscript
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
In general, existing machine learning based approaches, developed for systolic and diastolic blood pressure (SBP and DBP) estimation from oscillometric waveforms (OWs), employ features extracted from the OW envelope (OWE) alone and ignore important beat-by-beat (BBB) features which represent fundamental physical properties of the entire non-invasive blood pressure (NIBP) measurement system. Unlike the existing literature, this paper proposes a novel deep-learning based method for BP estimation trained with BBB time-domain features extracted from OWs. First, we extract six time-domain features from each beat of the OW, relative to the preceding beat. Second, using the extracted BBB features along with the corresponding cuff pressures, we form a feature vector for each OW beat and locate it in one of three different classes, namely pre-systolic (PS), between systolic and diastolic (BSD) and after diastolic (AD). We then devise a deep-belief network (DBN)-deep neural network (DNN) classification model as well as a novel artificial feature extraction method for estimating SBP and DBP from feature vectors extracted from OWs and their corresponding deflation curves. The proposed DBN-DNN classification approach can effectively learn the complex nonlinear relationship between the artificial feature vectors and target classes. The SBP and DBP points are then obtained by mapping the beats at which the network output sequence switches from PS phase to BSD phase and from BSD phase to AD phase, respectively, to the deflation curve. Adopting a 5-fold cross-validation scheme and using a data base of 350 NIBP recordings gave an average mean absolute error of 1.1±2.9 mmHg for SBP and 3.0±5.6 mmHg for DBP relative to reference values. We experimentally show that the proposed DBN-DNN-based classification algorithm trained with BBB timedomain features can outperform traditional deep-learning based methods for BP estimation trained with features extracted only from OWEs.
Article
Full-text available
Background: Although invasively measured blood pressure (invBP) is regarded as a "gold standard" in critically ill cardiac patients, the non-invasive BP is still widely used, at least at the initiation of medical care. The erroneous interpretation of BP can lead to clinical errors. We therefore investigated the agreement of both methods with respect to some common clinical situation. Methods: We included 85 patients hospitalized for cardiogenic shock. We measured BP every 6 h for the first 72 h of hospitalization, in all patients. Each set of BP measurements included two invasive (invBP), two auscultatory (auscBP), and two oscillometric (oscBP) BP measurements. InvBP was considered as a gold standard. Mean non-invasive arterial pressure (MAP) was calculated as (diastolic pressure + (pulse pressure ÷ 3)). We used Bland-Altman analysis and we calculated concordance correlation coefficients to assess agreement between different BP methods. Results: We obtained 967 sets of BP measurements. AuscMAP and oscMAP were on average only 0.4 ± 8.2 and 1.8 ± 8.5 mmHg higher than invMAP, respectively. On the other hand, auscSBP and oscSBP were on average - 6.1 ± 11.4 and - 4.1 ± 9.8 mmHg lower than invSBP, respectively. However, the mean differences and variability for systolic and diastolic BP variability were large; the 2 standard deviation differences were ± 24 and 18 mmHg. In hypotension, non-invasive BP tended to be higher than invBP while the opposite was true for high BP values. Clinical conditions associated with hypotension generally worsened the accuracy of non-invasive MAP. Conclusions: Mean arterial pressure measured non-invasively appears to be in good agreement with invasive MAP in patients admitted for cardiogenic shock. Several clinical associated with hypotension can affect accuracy of non-invasive measurement. Auscultatory and oscillometric measurements had similar accuracy even in patients with arrhythmia.
Article
Full-text available
Purpose of Review The purpose of this meta-analysis was to compare the magnitude of systematic bias (mean difference) and random error (standard deviation of mean difference) between the cuff method of indirect blood pressure and directly measured intra-arterial pressure. Recent Findings Blood pressure is almost exclusively assessed using the indirect cuff method; however, numerous individual studies have questioned the validity relative to directly measured intra-arterial blood pressure. Summary PubMed, SportsDiscus, and Scopus were searched through February 2018. Data were analyzed using a random effects model. A total of 62 studies met the inclusion criteria for quantitative analysis including 103 effect sizes for systolic and 114 effect sizes for diastolic blood pressure. Indirect measures of systolic blood pressure were underestimated (− 4.55 (95% CI = − 5.58 to − 3.53) mmHg), while diastolic blood pressure was overestimated (6.20 (95% CI = 5.09 to 7.31) mmHg). The random error (SD units) was 10.32 (95% CI = 9.29 to 11.36) for systolic and 7.92 (95% CI = 7.35 to 8.50) for diastolic blood pressure which corresponds to an estimation accuracy (95% confidence) of ± 20.2 mmHg for systolic blood pressure and ± 15.5 mmHg for diastolic blood pressure. These data indicate that it may be difficult to accurately estimate intra-arterial blood pressure using the cuff method. These results not only have implications for clinicians in diagnosing hypertension, but also may detail a potential underestimation of the association between blood pressure and numerous other health outcomes found in epidemiological studies.
Article
It is well known that non-invasive blood pressure measurements significantly underestimate true systolic blood pressure (SBP), and overestimate diastolic blood pressure (DBP). The aetiology for these errors has not yet been fully established. This study aimed to investigate the accuracy of Korotkoff sounds for detection of SBP and DBP points as used in brachial cuff sphygmomanometry. Brachial cuff pressure and Korotkoff sounds were obtained in 11 patients (6 males: 69.0 ± 6.2 years, 5 females: 71.8 ± 5.5 years) undergoing diagnostic coronary angiography. K2 Korotkoff sounds were obtained by high-pass filtering (>20 Hz) the microphone-recorded signal to eliminate low frequency components. Analysis of the timing of K2 Korotkoff sounds relative to cuff pressure and intra-arterial pressure shows that the onset of K2 Korotkoff sounds reliably detect the start of blood flow under the brachial cuff and their termination, marks the cuff pressure closely coincident with DBP. We have made the critical observation that blood flow under the cuff does not begin when cuff pressure falls just below SBP as is conventionally assumed, and that the delay in the opening of the artery following occlusion, and the consequent delay in the generation of K2 Korotkoff sounds, may lead to significant errors in the determination of SBP of up to 24 mmHg. Our data suggest a potential role of arterial stiffness as a major component of the errors recorded, with underestimation of SBP much more significant for subjects with stiff arteries than for subjects with more compliant arteries.
Article
Cardiovascular disease is the number one cause of death globally, with elevated blood pressure (BP) being the single largest risk factor. Hence, BP is an important physiological parameter used as an indicator of cardiovascular health. The use of automated non-invasive blood pressure (NIBP) measurement devices is growing, as measurements can be taken by patients at home. While the oscillometric technique is most common, some automated NIBP measurement methods have been developed based on the auscultatory technique. By utilizing (relatively) large BP data annotated by experts, models can be trained using machine learning and statistical concepts to develop novel NIBP estimation algorithms. Amongst artificial intelligence (AI) techniques, deep learning has received increasing attention in different fields due to its strength in data classification and feature extraction problems. This paper reviews AI-based BP estimation methods with a focus on recent advances in deep learning-based approaches within the field. Various architectures and methodologies proposed todate are discussed to clarify their strengths and weaknesses. Based on the literature reviewed, deep learning brings plausible benefits to the field of BP estimation. We also discuss some limitations which can hinder the widespread adoption of deep learning in the field and suggest frameworks to overcome these challenges.
Article
The use of automated non-invasive blood pressure (NIBP) measurement devices is growing, as they can be used without expertise, and BP measurement can be performed by patients at home. Non-invasive cuff-based monitoring is the dominant method for BP measurement. While the oscillometric technique is most common, a few automated NIBP measurement methods have been developed based on the auscultatory technique. Amongst artificial intelligence (AI) techniques, deep learning has received increasing attention in different fields due to its strength in data classification, and feature extraction problems. This paper proposes a novel automated AI-based technique for NIBP estimation from auscultatory waveforms (AWs) based on converting the NIBP estimation problem to a sequence-to-sequence classification problem. To do this, a sequence of segments was first formed by segmenting the AWs, and their corresponding decomposed detail, and approximation parts obtained by wavelet packet decomposition method, and extracting features from each segment. Then, a label was assigned to each segment, i.e. (i) between systolic, and diastolic segments, and (ii) otherwise, and a bidirectional long short term memory recurrent neural network (BiLSTM-RNN) was devised to solve the resulting sequence-to-sequence classification problem. Adopting a 5-fold cross-validation scheme, and using a data base of 350 NIBP recordings gave an average mean absolute error of 1.7±3.71.7\pm 3.7 mmHg for systolic BP (SBP), and 3.4±5.03.4 \pm 5.0 mmHg for diastolic BP (DBP) relative to reference values. Based on the results achieved, and comparisons made with the existing literature, it is concluded that the proposed automated BP estimation algorithm based on deep learning methods, and auscultatory waveform brings plausible benefits to the field of BP estimation.
Article
Objective: We verified the hypothesis that in noninvasive blood pressure (NIBP) measurement, inflationary NIBP measurement using the new type of cuff (YP-71xT series, Nihon Koden, Tokyo, Japan) might be associated with a reduced risk of subcutaneous hemorrhage. Methods: The study involved 30 healthy volunteers (15 males and 15 females). The blood pressure was measured by deflationary NIBP measurement + conventional cuff (control group), deflationary NIBP measurement + cuff (YP-71xT series) (deflationary measurement group), or inflationary NIBP measurement + cuff (YP-71xT series) (inflationary measurement group). NIBP measurement was performed five times in a row, then the presence or of subcutaneous hemorrhage was evaluated. The three different methods were used as cross-over design at 1-week interval for each subject so that all three methods were used for all the subjects. Results: The measurement time was significantly shorter in the inflationary measurement group than other groups. The incidence of subcutaneous hemorrhage significantly was lower in the inflationary measurement group (3%) than in control group (53%) (P < 0.001) and the deflationary measurement group (37%) (P = 0.002). Conclusion: This study revealed that inflationary NIBP measurement was associated with a dramatically reduced incidence of subcutaneous hemorrhage. Synergistic effect of the newly designed cuff, short measurement time, and low inflation pressure may allow the risk of subcutaneous hemorrhage.
Article
The inflationary non‐invasive blood pressure monitor (iNIBP™) uses a new measurement method, whereby the cuff is slowly inflated whilst simultaneously sensing oscillations, to determine the diastolic blood pressure first and then the systolic pressure. It may measure blood pressure more quickly than the conventional non‐invasive blood pressure monitor. We studied 66 patients undergoing general anaesthesia, comparing the time taken to measure the blood pressure between the two monitors at times when there were marked changes (increases or decreases by 30 mmHg or greater) in the systolic blood pressure. The median (IQR) [range]) time was significantly longer for the non‐invasive blood pressure monitor (38.8 (31.5–44.7) [18.0–130.0] s) than for the iNIBP (14.6 (13.7–16.4) [11.5–35.5] s), p = 0.001, 95%CI for difference 22–25 s). We also studied 30 volunteers to evaluate the accuracy of the iNIBP, comparing it with the mercury sphygmomanometer. There was good agreement between the two monitors, with a mean difference of 0 (95% limit of agreement −12 to 11) mmHg for the systolic blood pressure. We also compared the degree of pain during cuff inflation between the automated non‐invasive blood pressure and iNIBP monitors. Pain was significantly more for the non‐invasive blood pressure monitor (22 of 30 volunteers had less pain with the iNIBP). We have shown that the iNIBP measured the blood pressure quicker than the conventional non‐invasive blood pressure monitor and the speed of measurement was not significantly affected by marked changes in the blood pressure.
Article
This paper presents a novel method for estimating systolic blood pressure (SBP) and diastolic blood pressure (DBP) from time domain features extracted from oscillometric waveforms (OWs) using a long short term memory (LSTM) recurrent neural network (RNN) method. First, we extract seven time domain features from each cycle of OW including the cuff pressure, the cardiac period, the trough to peak amplitude of OW, the time between the trough and the peak of the OW, the slopes of the OWE and the maximum up-slope of individual OWs. Second, we locate each feature vector in an non-invasive blood pressure (NIBP) record in one of three different phases (classes), namely pre-systolic (PS), between systolic and diastolic (BSD) and after diastolic (AD), and form a target sequence. Then, we propose an LSTM-RNN approach to effectively learn the complex nonlinear relationship between the feature vector sequences and target sequence. The SBP and DBP points are then obtained by mapping the beats at which the network output sequence switches from PS phase to BSD phase and from BSD phase to AD phase, respectively, to the deflation curve. Adopting a 10- fold cross-validation scheme and using a data base of 350 NIBP recordings gave an average mean error of -1.2 ± 5.9 mmHg for SBP and 1.8 ± 8.8 mmHg for DBP relative to reference values derived from a visual method of determining SBP and DBP. The proposed RNN-based approach uses all time domain features available from each NIBP recording and can outperform traditional methods in BP estimation.
Article
This paper presents a novel method of estimating systolic blood pressure (SBP) and diastolic blood pressure (DBP) from time domain features extracted from auscultatory and oscillometric waveforms and using Gaussian Mixture Models and Hidden Markov Model (GMM-HMM). The nine time domain features selected include the cuff pressure (CP), the cardiac period (T), the energy of the Korotkoff pulses (KE), the oscillometric waveform envelope (OWE), the lag between the trough of the oscillometric waveforms (OWs) and the peak of the Korotkoff energy (Lag), the time between the trough and the peak of the OW (OWD), the slopes of the KE and OWE (SKE, SOWE) and the maximum up-slope of individual OWs (MSOW). Adopting a 5-fold cross-validation scheme and using a data base of 350 non-invasive blood pressure (NIBP) recordings gave an average mean error (± standard deviation of error) of -0.3±4.2 mmHg for SBP and 2.9±8.1 mmHg for DBP relative to reference values derived from a visual method of determining systolic and diastolic blood pressure. The significantly larger spread of DBP estimates relative to SBP, suggests that the criteria for determining DBP are poorly defined and would benefit from further experimental studies involving simultaneous invasive and non-invasive methods of measuring arterial pressure. We conclude that the proposed GMM-HMM BP estimation method outperforms previously reported methods in the literature and is a very promising method improving the accuracy of automated non-invasive measurement of blood pressure.