removing the deflation from the recorded nibp data fig. 2: a) filtered pulses (right scale) obtained...
TRANSCRIPT
![Page 1: Removing the deflation from the recorded NIBP data Fig. 2: a) Filtered pulses (right scale) obtained by band pass (0.3--20 Hz) filtering and b) Oscillometric](https://reader038.vdocuments.us/reader038/viewer/2022103004/56649ccb5503460f94994365/html5/thumbnails/1.jpg)
Removing the deflation from the recorded NIBP data
Fig. 2: a) Filtered pulses (right scale) obtained by band pass (0.3--20 Hz) filtering and
b) Oscillometric pulses (right scale) obtained by segmentation of data into heart beats.
Measured NIBP data (left scale) are plotted with red colour. Segmentation borders
(vertical dashed lines) coincide with time points that determine the negative envelope
of filtered pulses. The deflation signal, calculated by the interpolation of data between
subsequent segment borders, is subtracted from the measured data to obtain only
pulses with positive deflections (oscillometric pulses).
Periodic fist closing artefact
Fig. 5: Steps in the fist artefact extraction.
The person was closing her/his fist every 5 seconds during the NIBP measurement.
We performed the following steps during the analysis displayed in Fig. 5:
a) heart beat segmentation using filtered data (see also Fig. 2)
b) artefact segmentation - 2 s interval around the artefact peak
c) calculation of pulses using heart beat segmentation from (a)
d) estimation of oscillometric waveform from the normalised reference waveform,
obtained from artefact-free recordings (see, Fig. 3)
e) subtraction of estimated oscillometric waveform (d) from (c) to obtain only artefacts
f) extraction of artefacts from (e) using artefact segmentation in (b), and averaging
these artefacts. Averaged shape is shown on the top and plotted with a thicker line.
Results of forward and backward transformations
When comparing waveforms obtained from different recordings, shifts of ±N (usually 5)
beats are allowed to reach the best match between them. We called this procedure
waveform optimisation (WO). The average waveform is then calculated by summing
optimally shifted waveforms, see Fig. 3c. Criteria for the WO include minimum value of
relative difference (RD), maximum value of correlation coefficient (CC) and minimum
value of maximum difference (MD) between the compared waveforms.
To get the reference signal for a given measured oscillation waveform in a real
(measured) time scale, one has to transform the normalised reference oscillation
waveform back to the real time scale of the given measured data. However, since
oscillometric pulses slightly differ from measurement to measurement, we have used
beside the above WO also a single beat optimisation (SBO) during the transformation.
In both types, we first transform a given waveform into the heart beat view. Then we
compare it with the normalised reference waveform. Finally, we transfer it back to the
real time view. In the case of SBO, a single oscillometric pulse at the given cuff
pressure level is compared with pulses from the normalised reference. Amplitude
corrections and time shifts of these pulses are allowed within some reasonable
constraints and the best fitting reference pulse is determined for each pulse.
Averaging oscillometric non-invasive blood pressure recordings: Transformation into the normalised view
Vojko Jazbinšek*, Janko Lužnik, Zvonko Trontelj
Institute of Mathematics, Physics and Mechanics, University of Ljubljana, Ljubljana, Slovenia
*Email: [email protected]
Data acquisition.
For NIBP measurements, we have used the device
(see, Fig.1) that was designed by LODE (Groningen,
NL) for the EU-project ”Simulator for NIBP” [3].
We performed measurements on the upper arm of
healthy volunteers. Altogether, we have recorded data
on 23 persons (11 females and 12 males) between 20
and 66 years old. Most of attention was paid to the
external artefacts that could be repeated for every
volunteer and arose from known effects: beating the
cuff, external sound in the environment, moving of the
arm support, tremor, coughing, speaking, walking,
muscle contraction in the upper arm induced by
moving the arm, hand, finger or fist, etc. Fig. 5 show
results for the periodic fist closing artefact.
Fig. 1: NIBP device built in a personal computer.
Acknowledgment
We thank [3] for technical and financial support.
References[1] NG, K-G. and SMALL, C.F. Survey of automated non-invasive blood pressure monitors. Journal of Clinical Engineering, 19:452–475, 1994.
[2] JAZBINSEK, V., LUZNIK, J., and TRONTELJ, Z. Non-invasive blood pressure measurements: separation of the arterial pressure oscillometric waveform from the deflation using digital filtering. IFBME proceedings of EMBEC’05, 2005.
[3] European 5th framework programme. ”Simulator for NIBP”, Grant No. G6DR-CT-2002-00706.
[4] JACKSON, L. B. Digital Filters and Signal Processing. Kluwer Academic Publishers, Boston, 1986.
Fig. 3: Oscillometric pulses from the 1st (a) and the 2nd recording (b), and the normalised reference (c) obtained by averaging pulses from (a) and (b).
Introduction
Many oscillometric non-invasive blood pressure
(NIBP) measuring devices are based on recording the
arterial pressure pulsation in an inflated cuff wrapped
around a limb during the cuff pressure deflation [1].
The deflation can be removed from the recorded NIBP
data by a digital band pass filtering [2] (see, Fig. 2).
The obtained arterial pressure pulses are known as
the oscillometric pulses.
However, some NIBP measurements are
contaminated with the external artefacts arising mainly
from person’s movements. In most cases, these
artefacts generate pressure changes in the cuff, which
have similar frequency response as the oscillometric
pulses and it is therefore difficult to separate them. In
order to get oscillometric pulses independent on
variations of each heart beat duration, we have
introduced a transformation of data into the normalised
heart beat view. The transformed pulses can be
averaged to obtain the normalised reference pulses.
We have used such reference to extract artefacts from
measured data.
Transformation into the normalised view
The variable duration ∆tvar (sampled with νm into Nvar
points) of each measured heart beat is re-scaled to a
fixed value ∆tfix (νfix, Nvar). The core of the forward
transformation (∆tvar → ∆tfix) is to determine a
resampling frequency (νre) to represent each heart
beat (∆tvar) with the Nfix points:
For the backward transformation (∆tfix → ∆tvar) one has
to resample the "normalised" data on the ∆tfix with a
frequency νb to obtain the original points again
Fig. 3 demonstrates how two recordings (Figs. 3a,b)
can be transformed by Eq. 1 and then averaged to
obtain a reference waveform in the normalised view.
Fig. 4 shows how such reference can be subtracted
from the third recording (Fig. 4a) and then transformed
back to the real time view by using Eq. 2.
var var var
. (1)fix fix fixre fix m
N t N
t t N
var var . (2)varb m fix
fix fix fix
N t N
t t N
Fig. 4: Oscillometric pulses from the a) 3rd recording and residuals after reference subtracting obtained by using b) WO and c) SBO.