Left ventricular PV loop estimation using 3D echo A feasibility study to generate PV loops from 3D echo inter- intraobserver variability analysis on end systolic and end diastolic volumes and on calculated Emax BMTE 07.42 F.J. van Slochteren, P.F. Grundeman, A.J. Teske, P.H.M. Bovendeerd
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Left ventricular PV loop estimation using
3D echo
A feasibility study to generate PV loops from 3D echo
inter- intraobserver variability analysis on end systolic and
end diastolic volumes and on calculated Emax
BMTE 07.42
F.J. van Slochteren, P.F. Grundeman, A.J. Teske, P.H.M. Bovendeerd
2
ABSTRACT
Objective: A non invasive method to determine myocardial contractility can be of great use in daily
clinical practice. PV loops contain contractility information. Volume and pressure measurements can be
used to generate PV loops. 3D echo can be used to determine the left ventricular volume non invasively.
During this research project invasively measured ventricular pressures are combined with non invasively
measured volumes to construct PV loops.
Methods: First a method to generate PV loops from volumes that are measured by 3D echo is developed.
The accuracy of the PV loops, and the accuracy of the contractility information that is subtracted from the
PV loops, are investigated. The results of these investigations in case of a porcine cardiac failure model
are discussed in this document.
Results: The intra observer variability of the measured volumes shows a significant reliability. The inter
observer variability of the measured volumes shows poor reliability. The range of the calculated Emax
values is too wide to represent a physiological situation.
Conclusion: PV loop generation from invasively measured pressure and 3D echo measured volumes is
possible. The quality of the obtained echo data and the volumes that result from the observer analysis
influence the outcomes significantly.
Keywords: Cardiac dysfunction, PV loop, 3D echo, ESPVR, Contractility, Inter observer variability, Intra
observer variability
3
CONTENTS
Topic page
INTRODUCTION 4
METHOD 5
Animal model 5
Volume measurement 5
Pressure measurement 7
Combining the pressure and volume data sets 7
Emax calculation 7
Statistics 8
RESULTS 9
Volume results 9
Pressure results 9
Combining the pressure and the volume data sets results 9
Emax calculation results 10
Statistical results 11
Volume 11
Emax 12
DISCUSSION 13
Method 13
Animal model 13
Volume 13
Pressure 13
Combining the pressure and the volume data sets 13
Emax 14
Statistics 14
Results 14
Volume 14
Pressure 14
Combining the pressure and the volume data sets results 14
Emax 15
Volume statistics 15
Emax statistics 16
Considerations 17
CONCLUSION 17
RECOMMENDATIONS 17
REFERENCES 18
4
INTRODUCTION
Cardiac dysfunction is a disease that decreases the intrinsic force that can be delivered by a cardiac muscle
to eject blood. Cardiac dysfunction is induced by several pathologies of the heart. Cardiac ischemia and
dilation of the ventricle are a few diseases that affect the ability of the myocardium to generate force.
Inotropy (contractility) of the myocardium is a load independent measure that expresses the ability of the
cardiac muscle to generate force. It can be used to quantify cardiac (dys)function. Quantification of cardiac
dysfunction is necessary to be able to determine the most suitable treatment for a patient. The slope of the
end systolic pressure volume relation (ESPVR) of the left ventricle can be used as an estimation of the
inotropy of the left ventricle. The slope of the ESPVR is hereafter referred to as Emax. The inotropy of the
myocardium alters during exercise through neural and hormonal regulation mechanisms. During a stable
period of the patient the inotropic state of the myocardium can be assumed stable. Assessment of the Emax
during a stable phase qualifies the momentary inotropic state of the myocardium. When this is done in
different inotropic states (exercises) insight in the cardiac regulatory mechanisms can be generated. An
easy way to assess the inotropy can be of great use in daily clinical practice.
To generate insight into the inotropic state of the myocardium a few methods are clinically available. A
short overview of these methods with their pro’s and cons is given in the next section.
Emax is a parameter that represents the relation between the left ventricular pressure and volume at end
systole. When the pressure and the volume measurements are combined, a pressure volume loop (PV loop)
is generated. From a PV loop Emax can be derived. In Figure 1 a PV loop is drawn. The different methods
to assess a PV loop differ in the way LV volume is measured. Currently the left ventricular pressure is
measured invasively because there is no other method available at this time. Three different methods to
determine the left ventricular volume are discussed below.
1. LV volume can be acquired through a Baan catheter. This Catheter is placed in the left ventricle
and uses the conductance of the blood in the ventricle to determine the amount of blood in the
ventricle. Due to its invasiveness the procedure of the Baan catheter is highly demanding for the
patient.
2. MRI or CT can be used to acquire an image of the ventricle. The shape and size of the ventricle
can be calculated from the images. Due to the positioning in the scanner and the radiation during a
CT scan, these techniques are demanding for the patient.
3. Echo cardiography is a non invasive technique that uses ultrasound to examine the heart.
Information about flow, size (volume), and shape can be extracted from echo data. The most
recent echo equipment can generate moving 3D images. These images contain volume and time
information of one heart cycle.
The main subject of this study is to investigate the ability to construct a PV loop from volumes that are
derived from 3D echo data, combined with invasively measured pressures.
5
Figure 1: Theoretical PV loop
METHOD
Animal model
The 3D echo measurements are performed in a porcine model. The porcine model is subjected to a non-
ischemic surgical radius enlargement on the beating heart. In the pigs, the shape and radius of the LV are
altered by inserting a pericardial patch supported by a woven graft on the beating heart.
Due to the shape of a porcine thorax, a thorough trans thoracic scan of the left ventricle can not be made. A
3D oesophageal probe is not available yet. For this reason the echoes need to be acquired during the
termination phase of the pig, three months after the enlargement operation. In this phase the sternum is
almost completely removed to have full access to the heart. The echo probe can then be placed directly on
the heart. A lot of different probe positions can be tried during acquisition. A drawback of the probe
positioning directly on the heart is that it is difficult to catch the complete region of interest in the echo
window due to the small size of the echo window in the region that is near the echoprobe.
Volume measurement
The echocardiographic equipment that is used during this study is the Philips iE33. The 3D probe (X3-1)
is used. This echo probe contains 2400 elements that can send and receive ultrasound signals with a
frequency of 1.3 to 4 MHz. The frame rate is 22 Hz. Due to the fixed frame rate the volume curve of one
heart cycle at a high heart rate is represented in fewer frames than a volume curve of a ventricle beating at a
lower heart rate. A heart rate of maximal 80 beats per minute is necessary to generate a sufficient temporal
resolution. It is necessary to be able to have access to the whole heart when images are acquired directly
from the heart during the open thorax surgery. Therefore the size of the triangle shaped window where the
echo data is acquired is set to maximal. The spatial resolution decreases due to this.
A complete dataset of one heartbeat is acquired during 4 consecutive heartbeats. In each step (RR
interval) the echo beam measures ¼ of the total volume of the eventual dataset. The stepwise measuring
process is triggered on the ECG signal. To avoid interference lines it is important that the measurement
setup (probe positioning) is not changed during the measurement period. The 4 segments are added
together to make up the complete dataset of the volume that is measured. Volume information of a
complete cardiac cycle is reconstructed. The dataset is saved in a Dicom format.
The accuracy of the constructed PV
loops and the factors that influence
the accuracy will be investigated as
well. In this pilot study the intra-
and interobserver variability of the
measured end systolic volume
(ESV) and end diastolic volume
(EDV) are investigated.
How these volumes, and the
coupling of the pressure to the
volume curves, affect the Emax of
the left ventricle is also investigated.
When the left ventricular pressure is
measured or estimated by a non
invasive method as well, a complete
non invasive method to have insight
into the inotropic state of the
myocardium is generated.
6
The 3D dataset that is acquired is analyzed in the quantification software Qlab 3D Advanced (version
4.2.1.). The user interface of this software is shown in figure 2. This package contains an image analysis
function through which volumes of the heart compartments can be calculated.
Figure 2: echo data representation in the Qlab 3DQ Advanced quantification software. Perpendicular long axis views (top left and right) and short axis view (bottom left) The color of the lines that surround the images indicate the planes in the other images. The result of the volume calculation is shown in a 3D representation (bottom right). The ECG of the four subsequent heartbeats used for this analysis are shown at the right bottom of this image. In the column on the right the values of the EDV, ESV, Stroke volume and Ejection fraction are presented.
Figure 3: Volume curve as calculated by the Qlab 3DQ quantification software
In the quantification software the Dicom file that contains the echodata is opened. The data can be viewed
from different viewpoints. The data can be represented in three perpendicular planes, as depicted in figure
2. In this view traces can be drawn over the endocardial borders in the perpendicular planes. For correct
processing it is important that the position of the mitral valve annulus and the apex are marked carefully.
This needs to be done for both the end systolic and end diastolic frames. The end systolic and end diastolic
frames can be recognized from analysis of the ECG signal and from analysis of the shape of the ventricle.
When both the end systolic and end diastolic trace are determined, the software automatically draws the
endocardial border in all the frames that are in between, and calculates the volumes of these frames. During
the process of drawing and optimization of the trace of the endocardial border, the quantification software
uses a knowledge database to determine the ‘most likely’ shape of the ventricle. At last the volume curve is
interpolated to obtain a smooth volume curve. The resulting volume curve is represented in figure 3.
7
Pressure measurement
The left ventricular pressure is measured invasively through a miller catheter (5 French; 1.67mm). The
catheter is positioned during echo measurements to obtain good placement. The pressure is measured at a
samplerate of 383.4Hz. The pressure data is acquired by the CardiacSoft system version 3.4.1 by
Sonometrics.
For PV loop generation it is necessary to obtain pressure data of one cardiac cycle. Data from one cardiac
cycle is obtained by comparing the left ventricular and the aortic pressures. The interval between two points
where the left ventricular pressure rises over the aortic pressure is taken as one cardiac cycle. To smooth
the pressure data a 14 points convolution filter is applied. This filter sufficiently removes the measurement
noise from the signal and maintains the characteristic shape of the pressure curve.
Combining the Pressure and the Volume data sets
To obtain PV loops the pressure and the volume data during one cardiac cycle must be combined. This is
most easily done when measurements of both signals are triggered on the ECG signal and performed by the
same equipment. Unfortunately this feature did not function properly on the equipment that was used
during this project. For that reason the measurements needed to be performed on two different
measurement systems. The pressure is measured during the whole episode in which the echo measurements
are performed. In this way a normal shaped pressure curve can be chosen from the complete episode and
can be used during the PV loop generation. PV loop generation can only be done when the pressure and the
volume signals contain the same number of data points. Since the measurements of the pressure and the
volume are done at different sample frequencies, the data must be processed to level out the number of data
points. The shorter of the two signals is subjected to an interpolation algorithm to make it fit to the longer
of the two signals.
The two signals can be combined together when characteristic points of the pressure and the volume curves
are taken into account. Theoretically the pressure and volume curves can be subdivided in 4 phases. These
phases are depicted in figure 4. a: the ventricular filling phase. b: the isovolumetric contraction phase. c: the
ventricular ejection phase. d: the isovolumetric relaxation phase.
Figure 4: Theoretical relation between pressure and volume
Emax calculation
ESPVR is defined as the end systolic pressure volume relation. Emax is the slope of the ESPVR. Since
Emax is the parameter of interest, the combining of the pressure and volume curves is guided mainly by the
isovolumetric relaxation phase at end systole. The overall shape of the PV loop is another guideline to the
most optimal pressure volume coupling. The ESPVR line connects the points with the highest P/V ratio of
a series of PV loops that are drawn in different load situations of the left ventricle. Emax can be calculated
Combining the pressure and the volume
curve is done manually. Based upon the
characteristic phases of the pressure and
the volume curves a criterion is defined to
combine the pressure and the volume data.
In the criterion the focus is on the
isovolumic contraction and relaxation
phases. The most optimal PV loop is
assumed to be the PV loop when both the
isovolumic contraction and relaxation
phases are most vertical. This process is
optically optimized to obtain the most
reliable shape.
8
from the fitting of a linear function through the points of maximum P/V ratio. To be able to draw this line
at least 2 PV loops must be available. In this project three load situations of the left ventricle are examined.
These are: Head up (low LV volume), Horizontal (mean LV volume) and Head down (high LV volume).
The different loads result in different positions and shapes of the PV loops.
The calculation of Emax depends highly on the coupling of the pressure and the volume curve. The error
that can be introduced by the coupling needs to be investigated. This is done by shifting the pressure curve
with respect to the volume curve. Both a forward shift of the pressure as a backward shift of the pressure
over 52 milliseconds (20 datapoints) is applied. In that case each PV loop of each volume stage has 3
versions (backward shifted, normal, and forward shifted pressure curve). The influence of this shift of the
pressure curve on the PV loop can then be investigated.
Statistics
In the process of PV loop generation and Emax calculation some major observer influences can be noticed.
The manual tracing of the endocardial wall determines the values and shape of the volume curve. The
coupling of the pressure and the volume curves determines the shape of the PV loop. Both the tracing and
the coupling influence the Emax value. To examine the influences of these observer decisions, inter- and
intra observer analysis are performed on the determined ESV and EDV and on the Emax value. An
ANOVA test is done to examine whether the influences of the observers and the coupling are significant.
The processing of the echo data to determine the EDV and ESV, and to calculate the volume curve, is
done by three observers. Each observer processes the data three times. An ANOVA test is applied to
determine the significance of the variance that is introduced by the observers. The effect of the error that is
introduced when the pressure and volume curves are coupled is investigated through a fixed shift of the
pressure curve with respect to the volume curve. Fixed shift intervals are applied on each PV loop and the
influence of this on the Emax value is calculated through an ANOVA test.
The Emax value is calculated from three PV loops and three shift intervals are applied to each PV loop.
Therefore the PV loops that are used to construct Emax are selected from a total amount of 9 PV loops. For
each repetition of an observer 3 PV loops to the power of 3 shifts (33 = 27) Emax values can be calculated.
9
RESULTS
Volume results
When the echo data is processed a volume curve is acquired as depicted in figure 5. This curve is the result
of an interpolation routine. In figure 5 the volumes of all the frames are marked in the total volume curve.
Figure 5: interpolation of a volume curve
Pressure results
The Left ventricular pressure curve that is obtained during the echo measurement is filtered. The raw signal
and the filtered signal are represented in figure 6.
Figure 6: Filtered and raw pressure signal of a single heartbeat
Combining the pressure and the volume data sets results
When the pressure and the volume curves are combined, a most optimal fit is made based upon the shapes
of the pressure and the volume curves and of the shape of the PV loop. These are shown in figure 7.
The echoframes are measured with a
frequency of 22 Hz. The heart rate
during the measurement is 60 bpm.
The EDV is 80 ml and the ESV is
44 ml. The Ejection fraction is 45%.
No isovolumic phases at end systole
and end diastole can be recognized
in the volume curve.
From the figure can be noticed that
both the diastolic pressure and the
systolic pressure have low values.
The diastolic pressure is
approximately 10 mmHg and the
systolic pressure is approximately
100 mmHg. The pressure range
therefore is approximately 90
mmHg. The applied convolution
filter smoothes the signal.
10
Figure 7: Relation between pressure and volume (left) and the corresponding PV loop.
When the pressure and volume curves that are presented on the left of figure 7 are combined, the PV loop on the
right hand side is obtained. It can be seen that the pressure range of 90 mmHg in the pressure curve corresponds to
the pressure range of the PV loop. The ejected volume of 36 ml in the volume curve corresponds to the ejected
volume in the PV loop as well.
The effect of the error that is introduced when the pressure and volume curves are coupled is investigated through
the shift of the pressure curve with respect to the volume curve. This is shown in figure 8.
Figure 8: Shifted pressure curves with respect to volume curve (left) and PV loop when pressure curve shifted with respect to volume curve (right). The applied shifts are -52 ms (black) and 52 ms (red) with respect to the green curve (0 ms).
It can be noticed from figure 8 that a shift of the pressure curve over 52 ms influences the PV loop in all phases
where the pressure is not constant. The end systolic points of the PV loops and therefore also the Emax values
alter due to the applied pressure shift. This can clearly be noticed from the different positions of the upper left
corners of the PV loops that are shown at the right side of figure 8.
Emax calculation results
From at least two PV loops that are recorded at different volume loadings the ESPVR can be estimated. The slope
of the estimated ESPVR is the Emax value. This is depicted in figure 9 and 10.
11
Figure 9: Emax based on three filling volumes of the left ventricle (animal 1)
Figure 10: Emax based on two filling volumes of the left ventricle (animal 2)
The line of ESPVR is fitted through the points at end systole according to a linear function. In figure 9 the fit
could be based on the ESPVR of three PV loops that were recorded at different volume loadings. In figure 10 the
head down loading situation prominently affects the ESPVR and Emax. It was decided to exclude this loading
situation from the algorithm to calculate ESPVR. ESPVR of animal 2 was derived from two PV loops.
Statistical results
Volume
Since the measurements are performed on two animals, the intra- and interobserver analysis are performed on two
animals independently. The outcome of the intra- and interobserver analysis are represented in Table 1 and 2.
Here the mean and standard deviation of the end diastolic and end systolic volumes are presented. An ANOVA
test is performed to indicate whether the differences between the means of the different groups are significant.
This is indicated with the p value. When a threshold of 0.05 is applied it can be stated that there is 95% confidence
that the means of the compared populations differ significantly when the p value is under 0.05. The outcomes of
the ANOVA tests of animal 1 and 2 are depicted in table 1 and 2 respectively. The p (pos) column indicates the
difference between the mean values that are measured at the different positions (head up, horizontal, head down)
The p (rep) column indicates the difference between the mean values that are measured in each repetition. This is
the intra observer reliability. The p (obs) row indicates the difference between the mean values that are measured
by each observer in each position. This is the inter observer reliability.
Observer EDV mean (sd) [ml] ESV mean (sd) [ml]
head up horizontal head down p (pos) p (rep) head up horizontal head down p (pos) p (rep)
Table 2: Mean EDV and ESV at different positions measured by different observers (animal 2)
The EDV and ESV values are also represented in the box plots of figure 11. Here the mean values of the three
observers are presented in the boxes. The volume is placed along the vertical axis, and the different positions are
placed along the horizontal axis. EDV is depicted in blue and ESV is depicted in green. The difference between
EDV and ESV in the different positions can clearly be seen. The difference between animal 1 and animal 2 can
also be seen when the figures on the left and the right side are compared.
100
80
60
40
20
Position
Head downHorizontalHead up
ESV
EDV
Vo
lum
e [
ml]
Position
Head downHorizontalHead up
140
120
100
80
60
40
ESV
EDV
Vo
lum
e [
ml]
Figure 11: The mean EDV and ESV of the different observers are represented in the boxes. Animal 1 is depicted in the boxplot on the left and Animal 2 is depicted in the boxplot on the right.
Emax
The influences of the observers on the calculated Emax value is determined by an ANOVA test. The influences of
the observers through the volume measurements and the influences of the coupling through a shift parameter are
indicated. The results of this analysis are presented in Table 3 and 4.
Observer Emax mean (sd) [mmHg/ml] Emax mean (sd) [mmHg/ml]
no shift min max p (shift) no shift min max P (shift)