AJP - Heart Journal of Applied Physiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
 QUICK SEARCH:   [advanced]


     


Am J Physiol Heart Circ Physiol 289: H114-H130, 2005; doi:10.1152/ajpheart.01045.2004
0363-6135/05 $8.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Campbell, K. B.
Right arrow Articles by Slinker, B. K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Campbell, K. B.
Right arrow Articles by Slinker, B. K.

Dynamic myocardial contractile parameters from left ventricular pressure-volume measurements

K. B. Campbell,1 Y. Wu,1 A. M. Simpson,1 R. D. Kirkpatrick,1 S. G. Shroff,2 H. L. Granzier,1 and B. K. Slinker1

1Department of Veterinary and Comparative Anatomy, Pharmacology, and Physiology, College of Veterinary Medicine, Washington State University, Pullman, Washington; and 2Department of Bioengineering, University of Pittsburgh, Pennsylvania

Submitted 10 October 2004 ; accepted in final form 8 December 2004


    ABSTRACT
 TOP
 ABSTRACT
 Glossary
 METHODS
 RESULTS AND DISCUSSION
 APPENDIX A
 APPENDIX B
 GRANTS
 REFERENCES
 
A new dynamic model of left ventricular (LV) pressure-volume relationships in beating heart was developed by mathematically linking chamber pressure-volume dynamics with cardiac muscle force-length dynamics. The dynamic LV model accounted for >80% of the measured variation in pressure caused by small-amplitude volume perturbation in an otherwise isovolumically beating, isolated rat heart. The dynamic LV model produced good fits to pressure responses to volume perturbations, but there existed some systematic features in the residual errors of the fits. The issue was whether these residual errors would be damaging to an application where the dynamic LV model was used with LV pressure and volume measurements to estimate myocardial contractile parameters. Good agreement among myocardial parameters responsible for response magnitude was found between those derived by geometric transformations of parameters of the dynamic LV model estimated in beating heart and those found by direct measurement in constantly activated, isolated muscle fibers. Good agreement was also found among myocardial kinetic parameters estimated in each of the two preparations. Thus the small systematic residual errors from fitting the LV model to the dynamic pressure-volume measurements do not interfere with use of the dynamic LV model to estimate contractile parameters of myocardium. Dynamic contractile behavior of cardiac muscle can now be obtained from a beating heart by judicious application of the dynamic LV model to information-rich pressure and volume signals. This provides for the first time a bridge between the dynamics of cardiac muscle function and the dynamics of heart function and allows a beating heart to be used in studies where the relevance of myofilament contractile behavior to cardiovascular system function may be investigated.

heart function; muscle; mathematical model; cardiac fiber; force


LINKING MYOCARDIAL CONTRACTILE parameters with measurements taken in intact heart continues as a longstanding challenge in cardiovascular research. A widely applied and venerable approach has been to associate the isometric force-length relationship of isolated muscle with the various measures of the Frank-Starling mechanism in beating heart. Although this is a valid association, it is primarily an intuitive or qualitative notion that works well in reasoned explanations but does not offer a basis upon which to build a comprehensive system for quantitative connections. In the 1960s and 1970s, an intense effort was made to link muscle contractile behavior with whole heart behavior by relating the maximal velocity of unloaded cardiac muscle contraction with the time rate of change of pressure (dP/dt) during isovolumic contraction. This linkage was based on the Hill model of cardiac muscle and on the existence of a unique value for that model's series elastic element (55, 56). Whereas dP/dt continues as a valuable index for assessing the global contractile status of heart, the association between it and maximal shortening velocity could not be substantiated when it was shown that the Hill model was not a good representation of cardiac muscle and that the apparent series elastic element in this model was not uniquely valued (38, 40, 41). In the 1970s and 1980s, the apparent linear relationship between isochronal left ventricular (LV) pressure and volume led to the time-varying elastance [E(t)] concept for representing global LV mechanodynamic properties (47, 58, 61). Experimental evidence for the validity of E(t) led to efforts to link this global LV mechanical property to underlying contractile properties of muscle (47, 48, 59). One successful outcome of these efforts was the prediction and then the repeated confirmation of a strong empirical association between pressure-volume area and myocardial O2 consumption (47). However, it now appears that although E(t) is a useful descriptor of simultaneous LV pressure and volume events, it does not represent actual LV physical properties (6, 10, 14, 49, 52). Thus further attempts to link E(t) to contractile features of muscle are not likely to yield satisfactory results.

A long-term approach for linking heart and muscle has been to describe similarities in muscle and whole heart behaviors; similarities between the isometric force-length relationship of frog skeletal muscle and the isovolumic pressure-volume relationship of frog heart being the classical muscle-heart analogy (23). Many other similarity associations have been made of a broad scope of behaviors ranging from similarities in the end-shortening muscle force-length and end-systolic LV pressure-volume relationships (1, 20, 27, 48) to similarities in step and frequency responses of constantly activated heart and muscle (8, 11, 14, 15). Just as common has been the use of simple geometric transformations to derive muscle contraction relationships from LV measurements (5, 7, 44) or to reconstruct LV behaviors from muscle measurements (22, 39). Despite these many efforts, an unambiguous linkage with quantitatively verified associations has never been achieved.

Simultaneous with these experimentally based attempts were several modeling efforts in which elemental muscle contractile behavior was integrated mathematically with wall material properties, wall architecture and geometry, and chamber geometry in attempts to synthesize global organ function (25, 31, 33, 34, 36, 65). These efforts continue today with promise of eventual success (19, 26, 37), but because of the massive complexity of the problem, they are presently without practical results that may easily be implemented either experimentally or clinically.

A major problem in linking muscle contraction with LV mechanical behavior has been the reliance on inappropriate characterizations at the muscle level for making this link. For instance, the two most commonly used descriptors of muscle contraction, length-tension and force-velocity, are actually special cases with respect to contraction time and load, i.e., peak force during isometric contraction in the case of length-tension and initial shortening velocity against isotonic load in the case of force-velocity. These descriptors are not necessarily applicable to the dynamic history throughout a contraction event.

An alternative to length-tension and force-velocity descriptors of contraction is the dynamic stiffness of constantly activated muscle. Dynamic stiffness focuses on frequency-dependent force-length relations during small length changes and is profoundly sensitive to myofilament kinetic processes (3, 4, 30, 35, 46, 50, 51, 57, 62, 63, 66). Importantly, the frequency-domain expression of dynamic stiffness may be easily converted into an equivalent time-domain expression that allows prediction of the transient time course of muscle force in response to muscle length perturbations. Using the notion that myocardial dynamics are governed by both the dynamics of cross-bridge recruitment and the separate dynamics of cross-bridge distortion (12, 42), we recently constructed a simple differential equation representation of dynamic stiffness that accurately reproduces both transient and steady-state length-induced myocardial dynamic behaviors between 0.1 and 40 Hz (9). Interestingly, this model of muscle has the same mathematical form and dynamic time constants as an earlier LV model we developed from purely phenomenological evidence to describe the dynamic pressure-volume relationship in constantly activated heart (15). The implication of this model equivalence is that contractile force-length dynamics of myocardium are expressed in unaltered form in pressure-volume dynamics of the LV chamber. Thus the challenge became one of extending this analogy to beating heart. In this study, we show how to make this extension, allowing myocardial contractile parameters to be estimated from pressure-volume measurements taken in beating heart. This forges the long-sought link between myocardial contractile dynamics and whole heart pressure-volume behavior.


    Glossary
 TOP
 ABSTRACT
 Glossary
 METHODS
 RESULTS AND DISCUSSION
 APPENDIX A
 APPENDIX B
 GRANTS
 REFERENCES
 

Ai
Magnitude scalar for ith sine wave component of {Delta}V(t) signal

b
Rate constant of recruitment

c
Rate constant of distortion

D
Time derivative operator (=d/dt)

E(t)
Time-varying left ventricular (LV) elastance

E(t)-R
Elastance-resistance LV model

E{}
Dynamic elastance operator

E0
Zero-frequency LV elastance

E{infty}
Infinite-frequency LV elastance

EP
Passive elastance

e(i)
Residual errors

F(t)
Midwall fiber force

f0
Heart pacing frequency

fmin
Frequency of minimal stiffness

H{}
Dynamic operator that maps LV inputs to LV pressure

I{}
Dynamic interactance operator

I0
Zero-frequency LV interactance

I{infty}
Infinite-frequency LV interactance

L(t)
Length of midwall circumference

LBL
Baseline midwall circumference

ni
Frequency multiplier of f0 for ith sine wave component of {Delta}V(t) signal

P(t)
LV chamber pressure

Average pressure during isovolumic beat

PFf
Force-to-pressure transforming factor

PI(t)
LV pressure component due to interactance

Piso(t)
Isovolumic pressure

Pp(t)
Passive pressure

PR(t)
LV pressure component due to remainder terms in Taylor series

r
Rate constant of R{}

R{}
Dynamic operator that maps y(t) into PR(t)

R0
Magnitude of R{}

V(t)
LV chamber volume

VBL
Baseline LV chamber volume

Vw
LV wall volume

x(t)
Cardiac muscle distortion variable

y(t)
Inputs responsible for PR(t)

Q10
Ratio of reaction rate at one temperature and reaction rate at a temperature 10°C lower

{Delta}F(t)
Changes in cardiac muscle force

{Delta}L(t)
Changes in cardiac muscle length

{alpha}, {beta}
Viscoelastic parameters of passive left ventricle

{chi}0
Scalar parameter for elastance

{rho}0
Scalar parameter for resistance

{varepsilon}{}
Dynamic stiffness operator

{varepsilon}0
Zero-frequency cardiac muscle stiffness

{varepsilon}{infty}
Infinite-frequency cardiac muscle stiffness

{varepsilon}(j{omega})
Frequency-dependent muscle fiber stiffness

{phi}(t)
LV interactance recruitment variable

{eta}(t)
LV volume recruitment variable

{nu}(t)
Cardiac muscle recruitment variable

{psi}(t)
LV volume distortion variable

{zeta}(t)
LV interactance distortion variable

Building a Quantitative Link Between Cardiac Muscle Dynamics and LV Dynamics

Cardiac muscle dynamics: constant activation. The kernel for LV pressure-volume dynamics resides in the force-length dynamics of contracting cardiac muscle. To investigate these dynamics, we dissected bundles of fibers from the papillary muscle of rat hearts, removed the cell membranes with detergent to control activation at constant levels, and mounted these skinned fibers in an apparatus that allowed servo control of14fiber length (9). We then measured force development and response to length change during constant Ca2+ activation. Fiber length was changed over 45 s according to a command signal that specified small-amplitude, sinusoidal length variations in a frequency sweep from 0.1 to 40 Hz (Fig. 1).



View larger version (25K):
[in this window]
[in a new window]
 
Fig. 1. Force response ({Delta}F) from cardiac muscle fiber (top) during frequency-sweep length-change ({Delta}L) protocol (bottom). Measured force response (top; gray tracing) and model prediction from curve fitting Eqs. 13 to the measured force signal (black lines overlaying gray tracing) are compared. Expanded time scale at end of time period displays the last 0.2 s of the record during high-frequency (40 Hz) length change. Note the good model reproduction of amplitude and phase throughout the record from lowest to highest frequencies. Time point during frequency-sweep protocol of minimal-amplitude force response is shown (arrow). [Figure constructed from previously published data (8).]

 
Measured force-response signals to these imposed length changes were fit with a model developed from two fundamental considerations, including 1) a kinetic scheme for molecular contractile processes within the myofibers, and 2) the notion that a force generated by contractile units is equal to the number of parallel force generators in the force-generating state times the average elastic force per force generator (3, 12, 42). Change in the number of force generators was designated "recruitment," and change in the elastic force per generator was designated "distortion." Development based on the underlying molecular kinetic scheme yielded distinct dynamical equations for recruitment and distortion. We then reduced the recruitment-distortion model to incorporate one dynamic mode (one amplitude and rate constant) for recruitment and one dynamic mode (one amplitude and rate constant) for distortion. Expressed in differential equation format, the reduced model is


{zh40070540080e01}

(1)


{zh40070540080e02}

(2)


{zh40070540080e03}

(3)
In these equations, a dot over a variable represents the first time derivative; {Delta}L(t) is the measured change in length imposed upon the muscle; {Delta}F(t) is the model-predicted force in response to {Delta}L(t); {nu}(t) is the recruitment variable; and x(t) is the distortion variable. Both {nu}(t) and x(t) possess units of length. Parameters {varepsilon}0 and b represent the magnitude and rate constant of the slow recruitment response; parameters {varepsilon}{infty} and c represent the magnitude and rate constant of the fast distortion response. Both {varepsilon}0 and {varepsilon}{infty} possess units of stiffness.

When this model was fit to force responses in 118 records obtained from 19 fibers collected from 4 species with widely varying myofilament protein compositions, >98% of all the measured variation in the force response was explained (Fig. 1). Furthermore, the remaining unexplained variation was not correlated with the {Delta}L(t) input and thus could not be accounted for by any improvement in the model. Thus Eqs. 13 were accepted as suitable representation of force-length dynamics in constantly activated cardiac muscle.

To be consistent with mathematical developments, in the APPENDIX we consider the alternative expression for Eqs. 13 by writing the relation between {Delta}F(t) and {Delta}L(t) in input-output terms. For this, we use a dynamic operator that carries the physical units of stiffness. Employing the symbol D to represent differentiation with respect to time, D = (d/dt){}, Eqs. 13 may be written in the alternative but equivalent form

(4)
In Eq. 4, all terms within the square brackets constitute a dynamic operator that operates on the input, {Delta}L(t), to produce an output, {Delta}F(t). Let {varepsilon} represent the operator in square brackets. Then the dynamic force-length relation of cardiac muscle may be succinctly written as

(5)
The dynamic operator {varepsilon}{} has physical units of stiffness. Conversion of {varepsilon}{} to the frequency domain changes D to the complex frequency variable j{omega} and allows construction of the stiffness frequency spectrum.

Dynamic Muscle Stiffness (Constant Activation) is Converted to Dynamic LV Elastance (Constant Activation) by Geometric Transformation

Cardiac muscle fibers constitute the majority of the material of the LV wall and are arranged in a complicated pattern of spiraling sheets. However, in the midwall, fibers are circumferentially arranged and serially connected along the midwall circumference (Fig. 2). The centrality of these circumferential midwall fibers in the overall arrangement allows them to be taken as representative of all muscle fibers in heart providing that the distribution of material and physiological properties satisfies the condition of essential homogeneity as documented by uniform sarcomere-shortening patterns throughout myocardium (2, 21, 24, 43). Thus the force and length of an average circumferentially oriented fiber are representative of the force-length behavior of all the muscle fibers in the wall. By assuming spherical geometry, the transformation of force and length of the representative muscle fiber into pressure and volume of the LV chamber may be carried out by steps as indicated in Fig. 3.



View larger version (51K):
[in this window]
[in a new window]
 
Fig. 2. Muscle fibers are arranged circumferentially at the left ventricular (LV) midwall. Assuming a thick-walled sphere with homogenous material and physiological properties, a simple geometric transformation allows chamber volume [V(t)] to be converted to midwall length [L(t)], and force per unit area at the midwall circumference [F(t)] to be converted to chamber pressure [P(t)].

 


View larger version (9K):
[in this window]
[in a new window]
 
Fig. 3. Logic for transforming LV chamber volume [V(t)] into chamber pressure [P(t)] through the dynamic force-length [F(t)-L(t)] characteristics of cardiac muscle.

 
Reading Fig. 3 from left to right, there are two geometric transformation steps. A volume-to-midwall circumferential length transformation is given by

(6)
and a circumferential stress (force per cross-sectional area) to chamber-pressure transformation is given by

(7)

Although spherical geometry was used in the wall/chamber geometric transforming factors, small volume variation results in these factors being essentially constant regardless of the geometry. Thus small volume variations, because they do not cause appreciable variation in the transforming factors, allow a linear transformation of the model for force-length relations of the myofiber into an analogous model for pressure-volume relations in the left ventricle. For example, consider that measurements are made of LV pressure and volume. Then, for small {Delta}V around a baseline volume (VBL), the baseline length (LBL) of a representative midwall circumference will be

(8)
Furthermore, the force per unit cross-sectional area at the midwall is given by

where the pressure-force transforming factor <PFf> is (from Eq. 7)

(9)
With the use of these transformations, it can be shown that the relationship between a generic midwall muscle stiffness, {varepsilon} = dF/dL, and a generic LV chamber elastance, E = dP/dV, is

(10)
Thus the pressure-volume relationship of the left ventricle, given by E, may be converted to the force-length relationship of muscle, given by {varepsilon}, and vice versa.

Constant activation of muscle fibers to produce constant active force in heart occurs when Ba2+ is used as the activating agent (8). Because small volume perturbations allow linear transformation between the force-length relationship and the pressure-volume relationship according to Eq. 10, we can take advantage of the mathematical structure of {varepsilon} in Eq. 5 and by analogy write for the analogous property of constantly activated whole heart, i.e., the dynamic elastance, E

(11)
where E0 is the static (zero-frequency) elastance, E{infty} is the instantaneous (infinite-frequency) elastance, and b and c retain the definitions originally given in Eqs. 2 and 3. Thus the geometric transformation changes the values of the multipliers {varepsilon}0 to E0 and {varepsilon}{infty} to E{infty} but does not affect the values of the dynamic constants b and c.

With the definition for E{} given in Eq. 11, the pressure due to elastance (PE) and in response to a {Delta}V(t) perturbation in a constantly activated heart may be written as

(12)
Alternatively, it may be written in differential equation form analogous to Eq. 1 as

(13)
where {eta}(t) is the volume recruitment variable, analogous to the length recruitment variable {nu}(t) in Eq. 2, and {psi}(t) is the volume distortion variable, analogous to the lineal distortion variable x(t) in Eq. 2. The corresponding differential equations, which are analogous to Eqs. 2 and 3, are

(14)

(15)
In fact, we have previously acquired experimental evidence that Eqs. 1315 satisfactorily describe the steady-state pressure response to sinusoidal volume perturbations in constantly activated heart, and these are analogous to the equations that describe the steady-state force response to sinusoidal length perturbations of constantly activated papillary muscle (15).

Dynamic LV Elastance (Constant Activation) is Converted to Dynamic LV Interactance (Variable Activation) in Beating Heart

Now the challenge is to apply the above framework for myocardial-based dynamics of constantly activated left ventricle to beating heart, where activation is not constant. We summarize here the results of a more complete mathematical analysis of LV pressure-volume relationships during small volume perturbations of the otherwise isovolumic beating heart given in the APPENDIX. In this analysis, the pressure [P(t)] generated during a beat undergoing volume perturbation [{Delta}V(t)] depends on both {Delta}V(t) and the isovolumic pressure [Piso(t)] that would have developed had no {Delta}V(t) been administered.

In general, the dependence of P(t) on {Delta}V(t) and Piso(t) involves history (memory) of the system. To include history effects, it is useful to write the P(t) dependence in terms of a dynamic operator, H{}, i.e.

(16)
Equation 16 says that P(t) is the result of the mathematical operation H{} on two input functions, {Delta}V(t) and Piso(t). The dynamic operator H{} has a mathematical character similar to E{} in that it contains the time-derivative operator D as one of its primitive components; other primitive operators include scalars, summers, and multipliers.

A sequence of mathematical steps including a Taylor series expansion of Eq. 16 enables us to write an expression for the pressure response [{Delta}P(t)] to a small-amplitude volume perturbation in terms of the following: 1) an elastance pressure [PE(t)] due to the action of {Delta}V(t) alone as if {Delta}V(t) were applied to the left ventricle, generating a constant pressure equal to the mean of Piso(t); 2) an interactance pressure [PI(t)] due to the interaction of {Delta}V(t) and the variation of Piso(t) around its mean value[{Delta}Piso(t)]; and 3) a residual pressure [PR(t)] due to the sum of all the residual higher-order terms in the Taylor series.

Thus

(17)

In this, the elastance pressure PE(t) is given by Eqs. 1215 as if the heart is constantly activated to generate a pressure equal to the mean of Piso(t). The interactance pressure PI(t), which is of prime importance in this problem, may be couched in a form similar to our earlier expression, Eq. 12, for PE(t)

(18)
where the operator I{} operates dynamically on the product {Delta}V(t){Delta}Piso(t), which may be treated as a time signal defined as u(t) = {Delta}V(t){Delta}Piso(t). Furthermore, it is shown in the APPENDIX that I relates to E through differentiation with respect to Piso(t) as follows:

(19)
We make use of the specific formulation for E{} in Eq. 11 to derive a specific formulation for I{} from Eq. 19. We assume that of the E parameters, only the scaling coefficients E0 and E{infty} change with Piso, whereas the recruitment rate constant b and the distortion rate constant c do not. With this assumption, I may be represented as

(20)
where {partial}E0/{partial}Piso is the slope of E0 dependence on Piso (let I0 = {partial}E0/{partial}Piso), and {partial}E{infty}/{partial}Piso is the slope of i{infty} dependence on Piso (let I{infty} = {partial}E{infty}/{partial}Piso). Substitution yields

(21)

In the manner of the elastance example, we can now write the dynamic operations of Eq. 18 in differential equation format

(22)
where {phi}(t) is a time function with units of work (mmHg·ml) given by the first-order differential equation

(23)
and {zeta}(t) is a time function also with units of work (mmHg·ml) given by

(24)
The rate constants b and c are the same recruitment and distortion rate constants as in the elastance equations, Eqs. 14 and 15. The physical units of I0 and I{infty} are reciprocal volume (ml–1).

In the APPENDIX, it is shown that for the mean of the isovolumic pressure over the beat period () that

(25)
and

(26)

The remainder term [PR(t)], because it consists of lumping several higher-order terms in the Taylor series together, has few guidelines for representing its operator and the signal upon which it operates. Because it is a remainder of higher-order terms, it is assumed that PR(t) is small relative to PI(t), and its exact representation is less important to the problem than a correct representation of PI(t). Here we take an ad hoc approach and express the pressure due to these remainder terms, PR(t), in a dynamic form, i.e.

(27)
where it is understood that R{}, like E{} and I{}, is a dynamic operator, and y(t) is an appropriate signal to be defined.

We imposed the constraints that R{} be of low dynamic order and that it add no more than two unknown parameters to the overall model. Furthermore, likening PR(t) to a deactivation effect (19, 28, 29), we assumed that deactivation comes about as a result of distortion. Distortion is expressed in PE(t) and PI(t) by the terms E{infty}{psi}(t) and I{infty}{zeta}(t). We caused this deactivation to be independent of direction by assigning the forcing function y(t) to the square of the sum of these distortion-related terms

(28)
With this, R{} was represented as

(29)
where R0 is a magnitude parameter and r is a rate constant. The corresponding differential equation is

(30)

Summary of Quantitative Muscle-Left Ventricle Linkage

A dynamic kernel for LV dynamics was derived from dynamic force-length relations in constantly activated cardiac muscle fibers. Geometric transformation was used to convert dynamic force-length relationship of a representative midwall myofiber into dynamic pressure-volume relationship in the left ventricle of constantly activated heart. Applying the calculus of variation and a Taylor series expansion to the dependence of LV pressure on both the volume perturbation and the isovolumic pressure during beating (see APPENDIX), the pressure response to a volume perturbation was found to consist of three components as follows: a component due to the dynamic elastance referenced to the mean isovolumic pressure, a component due to the interactance between the volume variation and the pressure variation around its mean, and a component due to a remainder term consisting of higher-order terms in the Taylor series.

Summarizing the model equations

(a)

(b)

(b1)

(b2)

(c)

(c1)

(c2)

(d)
We refer to Eqs. ad as the dynamic LV model. The dynamic LV model equations contain five state variables [{eta}(t), {psi}(t), {varphi}(t), {zeta}(t), and PR(t)] and eight parameters including five magnitude-scaling parameters (E0, E{infty}, I0, I{infty}, and R0) and three rate constants (b, c, and r). Equations 25 and 26 demonstrate that E0 and E{infty} are not independent of I0 and I{infty}. Thus there are only six free parameters in the model equation set. (Note that as a consequence of the interdependence of E0 and I0 and E{infty} and I{infty}, a simpler rendition of Eqs. ad can be written; see APPENDIX. However, this simpler rendition obscures the origins of the components of the pressure response, which are important for understanding the model and how it arises.)


    METHODS
 TOP
 ABSTRACT
 Glossary
 METHODS
 RESULTS AND DISCUSSION
 APPENDIX A
 APPENDIX B
 GRANTS
 REFERENCES
 
Experimental Protocol

Hearts for these studies were obtained from rats according to a protocol approved by the Washington State University Institutional Animal Care and Use Committee. All animals in this study received humane care in compliance with the animal use principles of the American Physiological Society and the Principles of Laboratory Animal Care formulated by the National Society of Medical Research and the National Institutes of Health's Guidelines for the Care and Use of Laboratory Animals (NIH Publication No. 85-23, Revised 1985).

Hearts were isolated from young adult male rats (2–6 mo of age; 300–500 g body wt) after administration of anesthesia [that contained (in mg/kg im) 50 ketamine, 5 xylazine, and 1 acepromazine]. Upon excision of the heart, the aorta was quickly cannulated and the heart was immediately perfused through the aortic cannula with Krebs-Henseleit solution (Ca2+ concentration, 1.25 mM) that contained high levels of dissolved O2 (PO2 > 600 mmHg) at a constant perfusion pressure of 100 mmHg. Hearts were mounted onto an experimental system that consisted of a constant-pressure perfusion subsystem, an environmental control subsystem, a pacing-control subsystem, and a volume servo subsystem. A latex balloon attached to the obturator of the volume servo subsystem and sized so as to not develop measurable pressure when inflated to 500 µl was inserted into the LV chamber through the mitral orifice. The mitral annulus was secured to the obturator. Mounted and perfused hearts were then submerged in perfusate in a temperature-controlled environmental chamber that kept the epicardial surface wet and allowed for field stimulation. Electrical pacing was initiated using a computer-controlled stimulator and field electrodes placed in the bath of the environmental chamber on either side of the heart. The balloon was inflated to a reference volume [VBL], which was defined as the volume necessary to achieve an end-diastolic pressure of ~5 mmHg. LV pressure [P(t)] was measured with a 3-Fr catheter-tip Millar pressure transducer that was passed through a port in the volume servo system through the lumen of the obturator and positioned in the balloon within the LV chamber. Experiments were conducted at both 37 and 25°C by adjusting the temperature of the perfusate and the water circulating through the jackets of the experimental subsystems.

The LV balloon was connected to a computer-controlled volume servo system with a displacement piston in a servo chamber. Movement of the piston displaced volume out of or into the LV balloon according to the servo command. This allowed dynamic control and perturbation of LV volume with simultaneous measurement of LV pressure.

The pace period was set at 500 ms in experiments conducted at 37°C and 1,500 ms in experiments conducted at 25°C. Once stable isovolumic beating was achieved at VBL, a single-beat Frank-Starling protocol (16) was administered to evaluate both systolic and diastolic LV functions. Criteria for acceptable preparations included the functional indices of developed pressure ≥ 100 mmHg and passive stiffness ≤ 0.7 mmHg/µl at VBL. After we established that functional criteria were met, a volume-perturbation protocol was administered.

The volume-perturbation protocol was designed to provide pressure-response data from which model parameters could be estimated. Because dynamic behaviors associated with recruitment occur at low frequencies, whereas dynamic behaviors associated with distortion occur at higher frequencies (see Fig. 1), the perturbation protocol for parameter estimation necessarily needed to generate information at both low and high frequencies. To satisfy this requirement, six {Delta}V(t) signals were constructed, each of which consisted of one of two frequency compositions and one of three mean values. Each of the two frequency composites consisted of five summed sinusoids (Table 1). Separate records were taken of the response to each applied signal

(31)
where {Delta}j = –0.02VBL, 0, or 0.02VBL, and f0 = 1/(pace period). The magnitude scalar (Ai) and frequency multiplier of f0 (ni) for the ith sine wave component of the two different composite frequency signals were as indicated in Table 1.


View this table:
[in this window]
[in a new window]
 
Table 1. Frequency composition of {Delta}V(t)

 
Because the frequency of the slowest component in each frequency composite was 0.5 f0, the volume perturbation in both signals with differing frequency compositions lasted over 2 beats. In addition to recording the volume-perturbed beats, it was necessary to record the Piso(t) value in the beat immediately preceding the volume-perturbed beats. This was taken to be the Piso(t) value that would have been generated by the two perturbed beats if {Delta}V(t) had not been applied. With two frequency compositions and three {Delta}j values, an ensemble of six pressure-response records was generated.

Removing Passive Pressure Response From Measured Data

Pacing periods were chosen to be long enough to allow a well-defined diastolic period where pressure existed at the diastolic (passive) level without increasing or decreasing. During these diastolic periods, the pressure response to the volume perturbations was considered to be entirely due to passive LV properties. The passive left ventricle was modeled as a generalized viscoelastic body according to

(32)
where Pp(t) is passive pressure, EP is the DC passive elastance, and {alpha} and {beta} are viscoelastic parameters.

This passive-pressure model was fit (by nonlinear least squares; see below) to just that portion of the pressure response during the identified diastolic periods [Piso(t) < 10 mmHg; two diastolic periods in each record of two beats] and the passive parameters (EP, {alpha}, and {beta}) were thus estimated. The passive LV properties were taken to be in parallel with the active LV properties. With the use of Eq. 32 and the estimated passive parameters, that part of the total response due to passive properties was predicted throughout the data record, and the predicted passive pressure was then subtracted from the total {Delta}P(t) response to leave just the active response. The active response was the signal to which the following data-fitting and parameter-estimation techniques were applied.

Data Fitting, Parameter Estimation, and Model Evaluation

The dynamic LV model was fit to the active part of the measured pressure response by solving the model equations (Eqs. ad) for {Delta}P(t) using measured {Delta}V(t) and {Delta}Piso(t) as the forcing functions according to their roles in each model component. The differential equations were solved for the respective state variables by numerical integration (fourth-order Runge-Kutta) using an integration step size equal to the sampling interval of 0.001 s.

Model-generated {Delta}P(t) was fit to measured {Delta}P(t) by adjusting the six free model parameters (E0, b, E{infty}, c, R0, and r) using a modified Levenberg-Marquardt algorithm (MINPACK; Argonne National Laboratory) to minimize the sum of square residual errors. Outcomes from this fitting procedure included the following: 1) the set of parameters (E0, b, E{infty}, c, R0, and r) that provides the best fit of the model to the measured pressure response; 2) the standard errors of the estimates for each of these parameters; 3) the model generation of best fit, {Delta}(t); and 4) the time series of residual errors, e(i) = {Delta}(i) {Delta}P(i), where the i index indicates sampled values of {Delta}(t) and {Delta}P(t).

Model evaluation consisted of indices of descriptive validity, i.e., measures of how well the model fit the data. Two measures of descriptive validity were used, including 1) the linear regression of {Delta}(t) on {Delta}P(t), {Delta}(t) = b{Delta}P(t) + p0, and an evaluation of the regression parameters for their ideal values of b = 1 and p0 = 0; and 2) a relative error, er, defined as

where, for npts data points in the sample,

is the mean square error of residuals and

is the total variance in {Delta}P(t). Note that because the residual errors in this model-fitting exercise did not prove to be randomly distributed, the variance in the data that was not explained by the model was not simply 1 – R2. Thus er was valuable as an independent measure of goodness of fit.

At the end of the experiment, all atrial and great vessel tissue was trimmed from the heart, and the heart was blotted dry. The right ventricular free wall was trimmed from the heart, and the septum plus the LV free wall was weighed. The dissected right ventricular free wall was added to the septum plus the LV free wall, and all ventricular tissue was weighed.


    RESULTS AND DISCUSSION
 TOP
 ABSTRACT
 Glossary
 METHODS
 RESULTS AND DISCUSSION
 APPENDIX A
 APPENDIX B
 GRANTS
 REFERENCES
 
Volume-Perturbation Protocol Resulted in Pressure-Response Signals With Rich Dynamic Content

The volume-perturbation protocol was successful in generating pressure responses with dynamically rich information content. An example of typical pressure responses obtained from the volume-perturbation protocol (at 25°C; pacing period, 1.5 s) is given in Fig. 4. For clarity, this example shows only a subset of two of the full ensemble of six records to which the model was fit ({Delta}j = 0, frequency compositions 1 and 2). Note that at this pacing rate, there is ~0.5 s of diastole during which passive LV properties could be estimated using Eq. 32. The {Delta}P(t) signal shown in Fig. 4 (bottom) consisted of just the active part of the response (i.e., the passive response has been removed). Note the difference in {Delta}P(t) trajectory on the first and second beats in each of the two records, as these correspond to the positive- and negative-going halves of the slowest frequency sinusoid (f0, 0.5) in the volume-perturbation signal (Table 1). Also, note the different shape and form of {Delta}P(t) when {Delta}V(t) consisted of frequency composition 1 (Fig. 4, left) compared with when it consisted of frequency composition 2 (Fig. 4, right). With the combination of the trajectories and shapes represented in the full set of six records in the volume-perturbation protocol, sufficient information was present in the combined records to estimate model parameters that selectively influenced low- and high-frequency behaviors.



View larger version (18K):
[in this window]
[in a new window]
 
Fig. 4. Two 3-s records of composite frequency volume-perturbation protocol. Pacing frequency (f0) equaled 0.67 Hz. Volume perturbation consisted of a different combination of five frequencies between 0.5 and 32 f0 in each record (left vs. right). Top: pressure [P(t)] was recorded during isovolumic beating (thin lines) and volume perturbation (thick lines). Middle: volume-perturbation signals. Bottom: pressure response [{Delta}P(t)] to volume perturbation was calculated as the difference between pressure during volume perturbation and isovolumic pressure; different shapes of {Delta}P(t) (left vs. right) are due to different frequency contents of {Delta}V(t).

 
Dynamic LV Model Accounts for Most Features and Majority of Variance in Pressure Response to Volume Perturbation

The dynamic LV model, when fit to the data, reproduced all identifiable qualitative features in the response records. Qualitative comparison of a measured response during a single beat with a model-generated response for that beat demonstrated feature-by-feature reproduction (Fig. 5) including 1) large-amplitude responses during systole and virtually no response during diastole, 2) variation in response amplitude over the time course of the systolic interval, and 3) characteristic differences in shapes of the responses to volume perturbations with differing frequency content. When comparison is made between the first and second beats in the record, the trajectory of response during beat 1 (during the positive half of the slowest volume sine wave in the perturbation signal) was always above that of response during beat 2 (during the negative half of the slowest volume sine wave in the perturbation signal) in both the measured and model-generated signals. When comparison was made of the responses in records at different {Delta}j, the trajectory of response at {Delta}j = 0.02VBL was always above that at {Delta}j = 0, which was above the trajectory at {Delta}j = –0.02VBL in both measured and model-generated records.



View larger version (36K):
[in this window]
[in a new window]
 
Fig. 5. Comparison of measured and model-generated pressure responses. Systolic and diastolic periods of the beat are identified; diastole includes late relaxation. In each graph, the first (solid line) and second (dashed line) beats of the two beats in each response record are shown to demonstrate the differences in response trajectories of these two beats. Differences in responses to volume perturbations consisting of frequency compositions 1 and 2 are also reproduced by the model. See text for additional explanation.

 
The fit of the dynamic LV model to the full ensemble of six pressure-response records (three {Delta}j times two frequency compositions) generated the goodness-of-fit measures in Table 2. An example of a fit to an ensemble of six records (animal 29; 37°C; Table 2) is given by the overlay of model-generated responses on measured responses in Fig. 6. For much of the systolic period, it was difficult to discern the difference between measured and model-generated signals. This means that the trajectory of the low-frequency component of the measured response was well reproduced, as was the timing of the peaks and valleys in the high-frequency aspects of the response. Importantly, the reproduction of all six pressure-response records in Fig. 6 was with a single set of parameters that were estimated from fitting to all six records simultaneously.


View this table:
[in this window]
[in a new window]
 
Table 2. Goodness-of-fit statistics

 


View larger version (51K):
[in this window]
[in a new window]
 
Fig. 6. Fit of dynamic LV model (solid lines) to measured pressure responses (broken lines) over two beats. One set of parameters was used in the model generation of all six responses. Responses to {Delta}V perturbations of one composition of five frequencies around three mean ({Delta}; rows) values (left) and responses to another composition of five frequencies around the same three mean values (right) are shown. VBL, baseline LV chamber volume. See text for details.

 
Quantitative statistics from the fitting procedure are summarized in Table 2. The mean R2 from fits to records at 37°C was 0.83; the mean R2 from fits to records at 25°C was 0.80. The mean er (mean sum of squared residual errors relative to signal variance) was 0.17 from fits of records at 37°C and was 0.21 from fits of records at 25°C. The difference between mean goodness-of-fit parameters at 37 and 25°C was largely due to a relatively poor fit (R2 = 0.65; er = 0.39) in 1 of the 10 hearts (animal 34) in the 25°C dataset. Although this heart appeared to be an outlier, we had no objective reason for excluding it. Other than this one heart, fits to pressure responses at both temperatures were equally good. Consistent with the close overlay of measured and model-generated pressure responses in Fig. 6, the goodness-of-fit statistics indicated that there was only a small amount of residual variance in the data that was not accounted for by the model.

To appreciate the capability of the dynamic LV model to closely reproduce a set of dynamically complex pressure-response records, i.e., records with representation of both low and high frequencies (Fig. 6), comparison needs to be made to the best fit to this same set of data that can be obtained using an alternative model. For this alternative, we choose the broadly applied time-varying elastance-resistance [E(t)-R] LV model. In accord with previous work (13, 29, 34, 53, 54, 60, 64), we couched the E(t)-R formulation as

(36)
Because time variation in E(t) and i(t) follows that of Piso(t) (53, 54), Eq. 36 was written in terms of Piso(t) and two parameters including a scalar for elastance ({chi}0) and another for resistance ({rho}0)

(37)
The same optimization techniques used to fit the dynamic LV model to the pressure-response records were used to fit with the E(t)-R model and to estimate the two parameters {chi}0 and {rho}0. The results from the fitting procedure (Fig. 7) exhibit a clear separation between measured and E(t)-R-predicted responses throughout systole. Not only did the low-frequency trajectories of the model-generated and measured pressure responses differ, but also, the timing of the peaks and valleys of the higher frequency components within the model-generated and measured responses did not coincide as they did with the dynamic LV model. The average R2 from fitting the measured pressure responses with the E(t)-R model was only 0.37 compared with the R2 of >0.80 for the dynamic LV model. Although the E(t)-R model possessed only two parameters and the dynamic LV model possessed six, we argue that the E(t)-R model could not reproduce the measured response pattern with acceptable fidelity not because of a lack of parameters, but because it did not possess the dynamic mechanisms necessary for such reproduction. The point to be made in comparing Figs. 6 and 7 is not to denigrate the E(t)-R model, but rather, to demonstrate the challenge of using a model to reproduce a dynamically rich pressure-response signal from a beating heart, and to emphasize that we have largely met that challenge with our present dynamic LV model.



View larger version (56K):
[in this window]
[in a new window]
 
Fig. 7. Fit of elastance-resistance [E(t)-R] model (solid lines) to the same pressure responses (broken lines) shown in Fig. 6. Inability of the E(t)-R model to reproduce the measured response demonstrates the challenge of representing LV properties responsible for a broad range of dynamic behaviors. One set of parameters was used in the model generation of all six responses.

 
Despite the overall good fit by the dynamic LV model, we found that the slopes of the regression lines of model-generated vs. measured pressure response were <1 in every heart at both temperatures (see Table 2). This implied that there was a systematic character to the residual errors from the model fit. Indeed, when plotted as a function of time, the residual errors appeared to be nonrandomly distributed during the relaxation phase of the cardiac cycle. These errors briefly became predominantly negatively valued at the time of minimum dPiso/dt and then briefly became predominantly positively valued during the late phases of relaxation (data not shown). The issue raised by these systematic residual errors is whether their existence becomes damaging to the intended application of the model, i.e., the use of the model to allow estimates of cardiac muscle-contraction parameters from pressure and volume measurements taken in whole heart. To address this issue, we compared muscle-contraction parameters estimated from this study, using measurements taken in beating heart, with corresponding contraction parameters previously obtained in a study that used constantly activated, isolated cardiac muscle fibers (9). First, however, model parameters need to be organized to facilitate the comparison.

Dynamic Parameters May Be Separated Into Temperature-Insensitive (Magnitude Scaling) and Temperature-Sensitive (Kinetic Rate Constant) Groups

The estimates of parameters for all hearts and the averages of these estimates are given in Table 3. Because the most relevant of the six estimated model parameters (E0, b, E{infty}, c, R0, and r) were the four derived from dynamic behavior of muscle (E0, b, E{infty}, c), we focused on these four. These four parameters may be partitioned two ways, as follows: 1) those associated with the relatively slow dynamics of recruitment of force-generating units (i0 and b) and those associated with the relatively fast dynamics of distortion of force-generating units (E{infty} and c), or 2) parameters that scale the magnitude of the recruitment and distortion components of the response (E0 and E{infty}) and parameters that govern the speed (kinetics) of the two components of the response (b and c). Magnitude-scaling parameters (E0 and E{infty}) represent different aspects of the response than those aspects represented by the kinetic parameters (b and c). We asked how the magnitude-scaling parameters differ from the kinetic parameters in their dependence on temperature.


View this table:
[in this window]
[in a new window]
 
Table 3. Estimated parameters of dynamic LV model

 
In comparing the magnitude-scaling parameters at different temperatures, it was necessary to account for the influence of temperature on pressure development. Based on the average pressure (), over the course of an isovolumic beat, hearts at 25°C generated more pressure ( = 60.4 mmHg) than those at 37°C ( = 49.3 mmHg). To account for this difference in the comparison, we plotted E0 and E{infty} vs. and analyzed differences in the E0 vs. and E{infty} vs. relationships at the two temperatures using ANOVA. No demonstrable differences in each of these relationships at the two temperatures were found; P = 0.21 for E0 and P = 0.12 for E{infty}. Therefore, we concluded that the magnitude-scaling parameters E0 and E{infty} were temperature independent. Thus estimates of these parameters obtained in the heart at 37 and 25°C in this study could be compared with estimates of analogous parameters obtained from earlier muscle experiments conducted at other temperatures.

In contrast, the kinetic parameters b and c were strongly temperature dependent. The average value of b at 37°C (14.4 s–1) was 2.4 times the average value at 25°C (6.01 s–1), and the average value of c at 37°C (128.0 s–1) was 2.3 times the average value at 25°C (56.3 s–1); P < 0.001 by t-test for both b and c. Thus when comparing with equivalent kinetic parameters obtained from earlier cardiac muscle studies conducted at other temperatures, the temperature dependence of these kinetic parameters had to be taken into account.

Magnitude-Scaling LV Parameters Are Consistent With Magnitude-Scaling Muscle Parameters

Because the magnitude-scaling parameters were not affected by temperature, we lumped E{infty} values determined at 37 and 25°C together. E{infty} is referenced to the mean pressure level during isovolumic beating as if the heart muscle were constantly activated to produce the level of myocardial force commensurate with the mean pressure () during the beat. Geometric transformation (Eq. 7) was used to convert to a representative midwall force (F). Additional geometric transformation (Eq. 10) was used to transform the estimated chamber E{infty} to the corresponding midwall {varepsilon}{infty}. The resulting {varepsilon}{infty} vs. F relation from calculations in all 20 hearts is given by the data points in Fig. 8. Theoretically and experimentally (9), there is a linear relationship between steady-state active force produced by a constantly activated muscle and the infinite frequency stiffness ({varepsilon}{infty}) of the muscle fiber. The regression line describing this linear relationship for data previously collected in 19 cardiac muscle fibers at varying levels of activation is plotted in Fig. 8 along with the 95% tolerance limits (9) for observations around the regression line. Of the 20 {varepsilon}{infty} vs. F points derived by geometric transformation of E{infty} and from the 20 hearts studied, 18 fall on or within the 95% tolerance limits for the muscle fiber regression. That is, the observations of {varepsilon}{infty} vs. F obtained from estimates in beating heart using our dynamic LV model appear to be members of the same population as defined by the relationship between {varepsilon}{infty} and F found from experiments using constantly activated, isolated cardiac muscle fibers. This remarkable result leads to the conclusion that parameters estimated from fitting the dynamic LV model to dynamic pressure-volume behaviors may be reliably used to estimate stiffness vs. force relations in representative midwall muscle fibers.



View larger version (13K):
[in this window]
[in a new window]
 
Fig. 8. Plots of infinite-frequency (instantaneous) muscle fiber stiffness ({varepsilon}{infty}) vs. muscle fiber force (F). Regression from fit to data obtained in isolated muscle (solid line) and 95% tolerance limits on the observations around the regression line (dashed lines) are shown. Data points were derived by geometric transformation of model-derived LV infinite-frequency elastance and measured mean pressure during a heart beat.

 
Kinetic LV Parameters Are Consistent With Kinetic Muscle Parameters

The estimates of kinetic parameters b and c obtained in these beating-heart studies, which do not change with geometric transformation, are also in good agreement with findings from isolated muscle studies. Representative values for b and c obtained in rat skinned cardiac muscle fibers at 20°C were 4.2 and 31.9 s–1, respectively (data taken from results reported in Ref. 9). For the rat hearts of these studies, average estimates of b and c were 6.1 and 54.1 s–1, respectively, at 25°C and were 14.3 and 139.9 s–1, respectively, at 37°C. These values, obtained at different temperatures in beating heart and in constantly activated isolated muscle fibers, were considered together in an Arrhenius plot. The individual plots for b and c (Fig. 9) demonstrate that estimates from beating heart and constantly activated isolated muscle fall along one line for each kinetic constant. The calculated Q10 values from these Arrhenius lines were 2.02 for b and 2.21 for c. These Q10 values are characteristic of myofilament kinetic behaviors. Thus we conclude that isolated beating heart can be used with our dynamic LV model to estimate parameters of the same underlying myofilament kinetic processes as could be evaluated using constantly activated isolated muscle.