Beruflich Dokumente
Kultur Dokumente
Abstract: In heart failure, diastolic dysfunction is respon- for different ventricular relaxation states (tP ranging from
sible for about 50% of the cases, with higher prevalence in 24 to 68 ms) and two RBP models characterized by linear
women and elderly persons and contributing similarly to and nonlinear pressure-flow characteristics. dQ/dtmin and tQ
mortality as systolic dysfunction. Whereas the cardiac sys- always correlated with the dP/dtmin and tP, respectively
tolic diagnostics in ventricular assist device patients from (r > 0.97). These relationships were influenced by the non-
pump parameters have been investigated by several linear pump characteristics during partial support and by
groups, the diastolic behavior has been barely discussed. the pump speed during full support. To minimize these
This study focuses on the determination of ventricular influences, simulation results suggest the evaluation of
relaxation during early diastole in rotary blood pump dQ/dtmin and tQ at a pump speed that corresponds to the
(RBP) recipients. In conventional cardiology, relaxation is borderline between partial and full support. In conclusion,
usually evaluated by the minimum rate and the time con- at least in simulation, relaxation can be derived from pump
stant of left ventricular pressure decrease, dP/dtmin and tP. data. This noninvasively accessible information could
Two new analogous indices derived from the pump flow contribute to a continuous estimation of the remaining
waveform were investigated in this study: the minimum cardiac function and its eventual recovery. Key Words:
rate and the time constant of pump flow decrease, dQ/dtmin Ventricular relaxationRotary blood pumpNoninvasive
and tQ. The correspondence between the indices was inves- diagnosticsNumerical simulations.
tigated in a numerical simulation of the assisted circulation
Complementary to the systolic properties of the heart failure is responsible for about 50% of heart
heart, diastolic properties play an important role in failure in the patient population and is more common
determining the overall pumping function of the in the elderly and in women (13). However, diastolic
heart. Therefore, diastolic dysfunction has raised dysfunction is not limited to patients with diastolic
growing concern in the clinical community especially heart failure; it also occurs often in patients with
because of an increasing prevalence of diastolic heart systolic heart failure (37). Diastolic abnormalities in
failure among heart failure patients (1). Diastolic systolic heart failure patients seem to represent even
more powerful predictors of outcome and mortality
than systolic abnormalities (8).
The diagnosis of diastolic dysfunction involves
doi:10.1111/j.1525-1594.2011.01392.x measurement of left ventricular relaxation, filling,
Received July 2011; revised August 2011. diastolic distensibility, and stiffness (3). These mea-
Address correspondence and reprint requests to Dr. Francesco surements involve either invasive procedures (to
Moscato, Center for Medical Physics and Biomedical Engineering, measure ventricular pressure and/or volume) or
Medical University of Vienna, AKH-4L, Waehringer Guertel
1820, 1090 Vienna, Austria. E-mail: francesco.moscato@ time-consuming echocardiography (to obtain mor-
meduniwien.ac.at phologic and/or functional parameters) (4,9).
aor_1392 470..478
VENTRICULAR RELAXATION DIAGNOSTICS USING RBP FLOW 471
Our group has previously developed an index for end-diastolic pressurevolume relationship
the evaluation of systolic function during rotary (EDPVR) was modeled with an exponential function
blood pump (RBP) support. This index is derived (15); the end-systolic pressurevolume relationship
noninvasively, requiring only signals available from (ESPVR) was modeled as the sum of the EDPVR
the pump (10). We have now looked for a comple- and a parabolic function (16). Parameters of the
mentary, noninvasive method to monitor continu- EDPVR and the ESPVR were derived to correctly
ously one of the components of the diastolic function: reproduce the cardiac output (CO) curves reported
the ventricular relaxation. Numerical simulations of in reference (17). The systemic and pulmonary arte-
the assisted circulation were used to investigate the rial systems were modeled by three-element Wind-
correspondence between two classical indices of ven- kessel models (18). Parameters of the systemic and
tricular relaxation (the minimum rate of ventricular pulmonary arterial systems were derived from arte-
pressure decrease, dP/dtmin, and the time constant of rial pressurevolume curves and arterial pulse pres-
ventricular pressure decrease, tP) and two indices sures (19) and further adjusted to match measured
derived from the pump flow waveform. Additionally, arterial input impedances (20). Each compartment
an analytical approach was employed to verify the and the whole model were validated according to the
correspondence between the indices derived from procedure reported in reference (12).
ventricular pressure and pump flow. In contrast to the model described in reference
(13), the normalized elastance function has been
MATERIALS AND METHODS simplified to a biexponential function (21). This
function allowed the adjustment of a parameter
The present study focuses on the evaluation of ven- closely related to the time constant of ventricular
tricular relaxation in RBP patients during early pressure relaxation, tP. Six different relaxation
diastole. In particular, left ventricular assistance is conditions were simulated with resulting values of
considered with the RBP inflow placed in the ven- tP = 24, 34, 43, 51, 60, and 68 ms. These values
tricular apex and outflow in the aorta. Knowledge of covered the whole range of normal and impaired
the pump flow is also assumed (either measured at relaxation (9,22,23).
the pump outflow or estimated from intrinsic pump A condition of systolic left ventricular failure was
signals). When relaxation is impaired, the ventricular simulated in combination with the six different ven-
pressure decrease is slowed. Since ventricular pres- tricular relaxation conditions. The hemodynamics of
sure pulsatility influences the pulsatility of the pump the unassisted cardiovascular system was character-
flow via the pump pressure-flow characteristics, we ized by a CO of 3.1 L/min, a heart rate of 95 bpm, an
analyzed the features of the flow signal that might ejection fraction of 23%, a mean aortic pressure
correlate to two classical pressure-derived relaxation (AoP) of 69 mm Hg, and a mean left atrial pressure
indices: the minimum rate of ventricular pressure (LAP) of 19 mm Hg.
decrease, dP/dtmin, and the time constant of ventricu- Two models of RBPs were simulated (24,25). The
lar pressure decrease, tP (11). Two indices derived first represented a pump with linear pressure-flow
from the pump flow were used and compared to the characteristics (Eq. 1). The second represented a
classical relaxation indices: the minimum rate of pump with nonlinear quadratic pressure-flow charac-
pump flow decrease, dQ/dtmin, and the time constant teristics (Eq. 2). These two models do not represent
of pump flow decrease, tQ. any specific RBP used in the clinical practice, but an
idealized theoretical behavior and an axial flow
Numerical simulations pump, respectively.
A complete mathematical model of the cardiovas-
dQ(t )
cular system was implemented in Matlab-Simulink AoP(t ) P(t ) = a 2 RP Q(t ) LP (1)
(The MathWorks, Inc., Natick, MA, USA). Details of dt
the model have been published elsewhere (12,13). dQ(t )
Briefly, the cardiovascular system was modeled as a AoP(t ) P(t ) = a 2 [ rP Q(t ) ] Q(t ) LP
dt
closed-loop hydraulic circuit comprising active atria (2)
and ventricles, systemic and pulmonary circulations,
and an autoregulatory feedback loop for the arterial Here, AoP(t) represented the aortic pressure, P(t)
systemic resistance that ensured the regulation of the represented the left ventricular pressure, Q(t) repre-
arterial systemic pressure. Left and right ventricles sented the pump flow, w represented the rotational
were modeled by nonlinear time-varying elastances speed of the impeller, a represented a speed-
and internal resistances (14). The left ventricular related coefficient (a = 8.5710-5 mm Hgs2/rad2), RP
a c
b d
FIG. 1. (a) Left ventricular pressure P(t) during relaxation obtained from a simulation plotted together with the relationship of Eq. 3; (b)
time derivative of the left ventricular pressure dP(t)/dt obtained from the same simulation; (c) pump flow Q(t) during relaxation obtained
from the same simulation plotted together with the relationship of Eq. 5 and the numerical fitting (bold line); (d) time derivative of the pump
flow dQ(t)/dt obtained from the same simulation.
represented the slope of the linear pressure-flow was 25 ms before Q(t) reached its minimum. These
characteristics (RP = 0.625 mm Hgs/mL), rP repre- two points corresponded approximately to the usual
sented the quadratic coefficient of the nonlinear start and end points used to calculate the time con-
pressure-flow characteristics (rP = 4.1610-3 mm Hg stant from the ventricular pressure decay: start when
s2/mL2), and LP represented the fluid inertance dP(t)/dt is minimal and end 5 ms before mitral valve
(LP = 0.020 mm Hgs2/mL). opening (26). Figure 1c shows the exponential fitting
Simulations were performed for both pump superimposed onto the simulated pump flow
models and the six relaxation conditions, with the waveform.
pump speed changing every 20 s from 7.5 to Plots of the dQ/dtmin versus dP/dtmin and of tQ
12.5 krpm in steps of 0.5 krpm. One representative versus tP were generated from simulation results
cycle for each pump speed was extracted from the from both the linear and nonlinear pump models, and
signals once the simulation reached steady state. The regressions coefficients were calculated.
dP/dtmin and the dQ/dtmin were calculated by taking Finally, tQ was evaluated at a constant tP (0.043 s)
the minimum of the first time derivative of pressure and average pump flow (6 L/min), using various con-
and flow, respectively (Fig. 1b,d). The tQ was calcu- ditions of ventricular preload and afterload. In par-
lated by fitting an exponential function to the ticular, the LAP was changed over a range of 50%
decreasing portion of the flow waveform, using a least of the baseline value during support (10 mm Hg) and
squares criterion. The start point for the fitting was at AoP over the range of 18% of the baseline value
the time when dQ(t)/dt was minimal, the end point during support (86 mm Hg).
Analytical relationships
For a more complete understanding of the interac-
tion between flow and pressure relaxation param-
eters, the relationships dQ/dtmin versus dP/dtmin and tQ
versus tP were derived analytically by using the fol-
lowing simplifications and assumptions:
1 The ventricular pressure decrease and its time
derivative were treated as exponential functions of
time (11):
P(t ) = ( P0 P ) e t P + P (3)
dP(t ) ( P P ) t P
= 0 e (4)
dt P
FIG. 3. One cycle of the pump flow waveforms calculated from
Here, P(t) represents the left ventricular pressure; simulations with linear pump characteristics at different pump
P0 represents the pressure at time t = 0+, when relax- speeds and at a ventricular relaxation tP of 0.043 s (dotted lines);
exponential fitting used to identify the tQ is superimposed (solid
ation starts and the first time derivative of pressure is lines).
minimal (dP/dtmin); P represents the pressure asymp-
tote for t ; and tP represents the time constant of
pressure decrease. Figure 1a shows the left ventricu- 4 A mono-exponential function Q ( t ) was used for
lar pressure obtained from a numerical simulation the determination of the flow time constant, tQ, in
plotted together with Eq. 3. Figure 1b shows the time the form of Eq. 5.
derivative of the ventricular pressure and the dP/dtmin
( t tn )
for the same simulation data.
Q ( t ) = (Qi Q ) e Q
+ Q (5)
2 The linear RBP model was considered (Eq. 1).
3 The AoP was considered constant. Here, tn is the time when the first time derivative of
pump flow is minimal, Qi is the value of the flow at
time tn, Q is the flow asymptote for t , and tQ is
the time constant of pump flow decrease. Figure 1c
shows the pump flow obtained from a numerical
simulation plotted together with the curve in Eq. 5
and the numerical fit. Figure 1d shows the time
derivative of the pump flow and the dQ/dtmin for the
same simulation data.
The analytical relationships between dQ/dtmin and
dP/dtmin and between tQ and tP were plotted along
with the simulation results for comparison.
RESULTS
Figures 24 refer to the linear pump
characteristics. In Fig. 2, the relationship between
dQ/dtmin and dP/dtmin for three different values of tP
are displayed. Squares represent the simulation data
and circles the relationship obtained from the ana-
FIG. 2. Relationship between dQ/dtmin and dP/dtmin calculated
from the simulation data (squares) and the analytical relationship lytical approach (Eq. 6). In Fig. 3, the simulated
(circles) of Eq. 6. Data from linear pump characteristics and three pump flow waves corresponding to the different
different ventricular relaxation conditions (tP) are shown. Each pump speeds and a tP equal to 0.043 s are shown.
point marker represents a different pump speed going from 7.5 to
12.5 krpm from the bottom left corner to the upper right corner of The exponential fit used to calculate the tQ
the picture. is superimposed. Figure 4 shows the relationship
FIG. 4. Relationship between tQ and tP. Box and whiskers rep- FIG. 5. Relationship between dQ/dtmin and dP/dtmin calculated
resent the calculated tQ from the simulation data at different from the simulation data. Data from the nonlinear pump charac-
speed steps (n = 11) and six different ventricular relaxation con- teristics and three different ventricular relaxation conditions (tP)
ditions (tP). Together with the simulation data, also the analytical are shown. Each point marker represents a different pump speed
relationship of Eq. 7 is shown. Data from the linear pump char- going from 7.5 to 12.5 krpm from the bottom left corner to the
acteristics are shown. upper right corner of the picture.
dQ 1 P dP
(6)
dtmin e LP RP dtmin
LP
Q 2 P (7)
RP
Values were calculated in all cases for a tP of 0.043 s and average pump flow of 6 L/min.
tP, time constant of ventricular pressure decrease; tQ, time constant of pump flow decrease;
LAP, left atrial pressure; AoP, aortic pressure.
and tQ versus tP relationships, calculated for the non- In both the case of a linear and quadratic pump
linear pump characteristics at each single relaxation characteristics, dQ/dtmin correlated well with dP/dtmin
state and pump speed, respectively, lead to values (Figs. 2 and 5, respectively). Impaired relaxation is
r > 0.97. said to be present if the absolute value of dP/dtmin is
Finally, in Table 1, the changes from baseline of the lower than 1100 mm Hg/s (27). However, the major
time constant of flow decay (tQ) are shown under drawback of dP/dtmin (and consequently of dQ/dtmin)
different preload and afterload conditions. Results is its sensitivity to ventricular end-systolic pressure
are reported as percentage changes for each condi- (27), which can be strongly influenced by the pump
tion studied, using both the linear and nonlinear speed during a full support condition (i.e., when
pump models. aortic valve stays always closed). For this reason,
during full support, both dQ/dtmin and dP/dtmin
decrease in absolute value for increasing pump speed
DISCUSSION
(Figs. 2 and 5). The condition of partial support (i.e.,
Proper monitoring of the assisted ventricle seems aortic valve opening during systole) also deserves
crucial to determine if and when weaning is possible, consideration. For the linear case (Fig. 2), both the
but also to preserve residual function in case of per- differential pressure and the flow pulsatility are not
manent support. Monitoring of ventricular function influenced much by the pump speed and so the
involves not only conventional ultrasound and cath- resulting dP/dtmin and dQ/dtmin tend to remain
eter examinations, but also continuous noninvasive constant. However, for the nonlinear case (Fig. 5),
monitoring from the pump available signals. Indeed, flow pulsatility is influenced by pump speed despite
systolic as well as diastolic ventricular properties, an almost constant differential pressure, with a reduc-
heart rhythm, and aortic valve condition (if it stays tion of pulsatility for increasing speeds. Therefore,
always closed during the heart cycle or not) all leave dQ/dtmin significantly decreased in absolute value for
a fingerprint on the pump flow signal. This study increasing speed despite an almost constant dP/dtmin.
investigated by numerical simulations the possibility This introduced a significant deviation of this rela-
of monitoring ventricular relaxation from the pump tionship from linearity.
flow only. This diagnostic tool could contribute to a Simulation results therefore suggest evaluating the
more comprehensive picture of the ventricular func- dQ/dtmin at a pump support level that corresponds to
tion and its changes during mechanical assistance. the borderline between partial and full support,
In addition to that, understanding and monitoring where neither the pump speed nor the pump nonlin-
diastolic dysfunction and its course (or at least ven- earity influences the dQ/dtmin and the dP/dtmin.
tricular relaxation, as in the present study) could Both in case of a linear and nonlinear pump model,
eventually encourage RBP use in patients with dias- tQ correlated with tP (Figs. 4 and 6). For the linear
tolic dysfunction alone. However, further studies of pump model, the relationship did not depend on
the peculiar hemodynamic condition of this class of either pump speed or partial/full support conditions,
patients and the interaction with a RBP are needed. as the small dispersion of the data in Fig. 4 at different
In the present study, it could be shown, at least in pump speeds shows. On the other hand, for the non-
numerical simulations, that the relaxation parameters linear pump model, the slope of this relationship
dQ/dtmin and tQ derived from the pump flow are ana- showed a strong dependency on pump speed (Fig. 6),
logues to the classical, pressure-derived ones dP/dtmin increasing for decreasing speed and especially during
and tP. partial support at low speed/flow (7.5 krpm in Fig. 6),
where the nonlinearity effects predominate. Never- this simplification does not have any effect on relax-
theless, for each given speed, a proportional relation- ation, which was the focus of the present study.
ship between tQ and tP exists and a coherent increase This study did not include the influence of mea-
or decrease of tP follows an increase or decrease of tQ. surement noise. In case of noisy signals, an approach
Results in Fig. 6 can be interpreted considering that based on signal averaging over multiple beats must
the quadratic pressure-flow pump characteristic has a be implemented (30). The multiple-beat averaging
slope that depends on the pump flow ([rP |Q(t)|] in method would allow a robust calculation of physi-
Eq. 2). Higher pump speeds lead to higher pump flows ologically relevant features from the noisy flow signal
that in turn result in a bigger pressure-flow character- (dQ/dtmin and tQ).
istic slope in comparison to a low speed/flow condi- Clearly, a comprehensive evaluation of the diastolic
tion. A bigger pressure-flow characteristic slope function must include, together with relaxation, the
determines a shallower relationship between tQ and tP aspects of myocardial diastolic distensibility, stiffness,
as can be inferred by Eq. 7 (where now RP is not and passive ventricular filling. An evaluation of these
constant but corresponds to the flow dependent term properties from the pump flow waveform seems diffi-
[rP |Q(t)|] in Eq. 2). Additionally, for very low pump cult but possible, as a wide variety of diastolic pump
speed/flow eventually leading to a condition of partial flow shapes can be observed in patients that reflect
support with backflow (7.5 krpm in Fig. 6), the nonlin- both ventricular filling and diastolic properties.
earity of the pressure-flow characteristics strongly Current investigations are ongoing to elucidate the
amplifies the flow pulsatility for slight changes in dif- relative changes in the pump flow shape at different
ferential pressure, which increases the slope of the speeds and how these changes relate to diastolic
relationship between tQ and tP. Backflow or even a stiffness and filling properties.Nevertheless,the evalu-
condition of very low pump flow should be therefore ation of relaxation alone can already provide infor-
avoided when evaluating the tQ. mation about diastolic function, as structures and
Simulation results therefore suggest that also the processes that alter relaxation can also result in mea-
evaluation of tQ shall be performed at a pump surable abnormalities in stiffness and vice versa
support level that corresponds to the borderline (4,7,31,32).
between partial and full support. Finally, the evaluation of ventricular relaxation
Finally, Table 1 shows a moderate influence of presented here requires the measurement of pump
preload on the evaluation of the tQ (6 and 9.5% for flow, which is not commonly available in clinically
linear and nonlinear pump characteristics for a 50% implanted pumps. However, an estimation of pump
change in LAP, respectively) but no influence of after- flow with adequate frequency response would suffice
load. for this diagnostic purpose.
Some additional aspects and limitations of the
present study should be considered.
The data presented here were obtained from
CONCLUSIONS
numerical simulations of assisted heart hemodynam-
ics, and, therefore, the results must be validated in Numerical simulations suggest that ventricular
subsequent animal and/or clinical studies. Neverthe- relaxation properties could be determined in rotary
less, the model was based on a careful literature blood pump recipients noninvasively by using two
review (1423) and clinical experience, and it is indices calculated from the pump flow waveform
believed to be a valid tool for elucidating the (dQ/dtmin and tQ). These indices corresponded, in the
complex interaction between ventricular relaxation, simulations, to the classical, pressure-derived relax-
pump characteristics, and the flow waveform. ation indices. A noninvasive assessment of relaxation
A simple biexponential model (21) for the normal- could contribute to the overall evaluation of heart
ized contractility function was used that led to function and its potential for recovery under RBP
monoexponential ventricular pressure decay and assistance.
rise. Although logistic (28) or even more complex
(26) models have been proposed for a better repre- Acknowledgments: This work was supported in
sentation of ventricular pressure decay, a monoexpo- part by the Oesterreichische Nationalbank Jubilae-
nential curve with nonzero asymptote adequately umsfonds zur Foerderung der Forschungs- und
describes the ventricular pressure decay in most clini- Lehraufgaben der Wissenschaft (Jubilaeumsfonds-
cal cases (9,22,23). The rising portion has been also projekt Nr. 13692).The authors thank Prof. Edward F.
modeled with an exponential curve even if this does Leonard from Columbia University, New York, NY,
not reflect real ventricular behavior (29). However, for his advice.
t
t
dQ ( t )
= (Q0 Q ) 2 e tn (A5) dQ ( t ) 2 (Q Q )
dt tn = 0 (A9)
dtmin e Q
Here, tn is the reciprocal of the natural frequency
(wn) and can be calculated as shown in Eq. A6. This This minimum is set equal to the one calculated in
term corresponds to the time when the dQ(t)/dt Eq. A7 thus obtaining:
reaches its minimum.
1 P dP 2 Q Q
0 (A10)
1 LP P e LP RP dtmin e Q
tn = = (A6)
n RP
From Eq. A10, one can finally obtain the relation-
By calculating the minima from Eqs. 4 and A5 and
ship between tQ and tP using the relationships
using the relationship between Q0 and P0 and Q and
between Q0 and P0 and Q and p, Eqs. A2 and A3,
p, Eqs. A2 and A3, one can finally derive the rela-
and the dP/dtmin calculated from Eq. 4.
tionship between dQ/dtmin and dP/dtmin.
dQ 1 P dP LP
(A7) Q 2 P (A11)
dtmin e LP RP dtmin RP