Simulating the swelling and deformation behaviour in soft tissues using a convective thermal analogy
 John Z Wu^{1} and
 Walter Herzog^{2}Email author
DOI: 10.1186/1475925X18
© Wu and Herzog; licensee BioMed Central Ltd. 2002
Received: 1 August 2002
Accepted: 19 December 2002
Published: 19 December 2002
Abstract
Background
It is generally accepted that cartilage adaptation and degeneration are mechanically mediated. Investigating the swelling behaviour of cartilage is important because the stress and strain state of cartilage is associated with the swelling and deformation behaviour. It is well accepted that the swelling of soft tissues is associated with mechanical, chemical, and electrical events.
Method
The purpose of the present study was to implement the triphasic theory into a commercial finite element tool (ABAQUS) to solve practical problems in cartilage mechanics. Because of the mathematical identity between thermal and mass diffusion processes, the triphasic model was transferred into a convective thermal diffusion process in the commercial finite element software. The problem was solved using an iterative procedure.
Results
The proposed approach was validated using the onedimensional numerical solutions and the experimental results of confined compression of articular cartilage described in the literature. The timehistory of the force response of a cartilage specimen in confined compression, which was subjected to swelling caused by a sudden change of saline concentration, was predicted using the proposed approach and compared with the published experimental data.
Conclusion
The advantage of the proposed thermal analogy technique over previous studies is that it accounts for the convective diffusion of ion concentrations and the Donnan osmotic pressure in the interstitial fluid.
Background
The extracellular matrix (ECM) of articular cartilage is negatively charged and hydrated [1, 2]. Under normal physiological conditions, the fixed negative charge in ECM is electrically neutralized by mobile cations and anions contained in the synovial fluid. The difference in ion concentration between the ECM and synovial fluid introduces a fluid pressure difference between them, the socalled Donnan osmotic pressure [3]. At equilibrium, the swelling pressure, which consists of the Donnan osmotic pressure and a chargetocharge expansion stress within the tissue, is counteracted by structural elements in the solid matrix.
It is well accepted that the swelling of soft tissues is associated with mechanical, chemical, and electrical events. The triphasic theory [4], which was generalized by Gu et al. [5] to include multiple ion species, is the only realistic model currently available to describe the mechanics of articular cartilage swelling. Gu et al. [6] and Gu [7] used the triphasic theory to analyse the coupling between fluid and ion transport, streaming potential, and deformation of cartilage during compression tests. Bryant and McDonnell [8] applied the triphasic theory to analyse corneal swelling under steadystate conditions in which the effects of fluid flow were neglected. Recently, Hon et al. [9] developed a rigorous numerical algorithm of the triphasic model for twodimensional problems based on radial basis functions. However, all numerical/analytical approaches mentioned above are problemoriented, and cannot be used for solving realistic biological problems, which are usually nonlinear, threedimensional, irregular and complex in geometry, and undergo large deformations.
In order to apply the triphasic theory to biological systems, or applied biomechanical problems, the triphasic constitutive equations must be implemented into a finite element model (FEM). Simon et al. [10] proposed a poroelastic finite element formulation that included transport and swelling effects in soft tissue. This model might be considered a special case of the triphasic model by Lai et al. [4]. Simon et al. [10] developed a onedimensional FEM that was used to demonstrate the properties of triphasic materials. However, it could not be used for practical problems in two or threedimensional space. Sun et al. [11] developed sophisticated FEM formulations that included the coupling between mechanical, electrical and chemical events in articular cartilage. However, the approach proposed by [11] still in a conceptual stage and needs substantial research and development until it can be used for solving realistic biological problems.
Because the process of diffusion is mathematically equivalent to the process of thermal transfer, the transport of ions in cartilage may be analysed using a thermal analogy. Myers et al. [12] simulated swelling of articular cartilage using a thermal analogous technique. Kaufmann et al. [13] analysed poroelastic problems with swelling deformation using a commercially available FEM software package (ABAQUS). In these last two studies, the diffusion process of ion concentrations was analysed by neglecting convection, i.e., by assuming that the velocity of the tissue and the diffusion of ions are decoupled. For most substances, the diffusion process is strongly coupled with the velocity of the media. Therefore, the coupling between the velocity of the tissue and the diffusion of ions may have nonnegligible effects in physiological problems.
Understanding the mechanisms of swelling and deformation is important when investigating the mechanics and adaptative behaviour of articular cartilage. The triphasic theory [4, 5] can be used to describe swelling and transport of ions in articular cartilage. However, a commercial finite element tool based on the triphasic theory, which bioengineers can use to analyse swelling and deformation behaviour for reallife biophysical problems, does not exist. The purpose of this study was to implement the triphasic theory [4, 5] into a commercially available finite element software package (ABAQUS), and to validate the proposed approach using the onedimensional numerical solutions by Simon et al. [10], and the experimental results of confined articular cartilage compression obtained by Eisenberg and Grodzinsky [14]. In order to solve the problem using ABAQUS, the triphasic model was converted into a thermomechanical interaction problem with convective heat transfer. It was assumed that the volume occupied by ions is small and that the effect of electric current on the flow of ions can be ignored.
Method
Governing Equations
Articular cartilage was assumed to obey the triphasic model proposed by Lai et al. [4] and Gu et al. [5]. For infinitesimal deformation, the constitutive equation is expressed as
where I is the unit tensor; σ^{ s }, σ^{ f }, and σ^{ t }represent the stresses in the solid phase, in the interstitial fluid, and in the total tissue; λ and μ are the first and the second Lamé constants; p and ε are fluid pressure and strain of the solid phase, respectively. The stress in the solid phase (σ^{ s }) is decomposed here into three parts: a component associated with the fluid pressure (Φ^{ s } p I), a chemical expansion stress (T _{ c } I), and an elastic stress associated with the deformation of the matrix. Φ^{ s }and Φ^{ f }are the solid and fluid volume fractions, respectively. Assuming that the volumetric fraction of ions, Φ^{} and Φ^{+}, is negligible, as suggested by Gu et al. [5] and Sun et al. [11], one can write the equation of the mixture in the form:
Φ^{ s }+ Φ^{ f }+ Φ^{} + Φ^{+} ≈ Φ^{ s }+ Φ^{ f }= 1 (2)
The ECM is negatively charged, with a fixed charge density of c^{ F }. Everywhere within the tissue under physiological conditions, the electroneutrality condition is maintained by
c^{+} = c^{} + c^{ F } (3)
where c^{+} and c^{} are concentrations of mobile cations and anions, respectively. c^{+}, c^{}, and c^{ F }are measured in moles per unit fluid volume.
The conservation of momentum is satisfied everywhere in the tissue
where υ^{ s }and υ^{ f }are the velocities of solid and fluid, k is the hydraulic permeability (open circuit). The continuity condition holds everywhere in the tissue, such that
Darcy's law is obtained from the second equation of the conservation of momentum (4).
W _{ i }= Φ^{ f }(υ^{ f } υ^{ s }) = k p (6)
where W _{ i } is the effective fluid flux.
Fluid pressure p is composed of hydrostatic pressure, p _{0}; and osmotic pressure [15], Δπ:
p = p _{0}  Δπ (7)
The osmotic pressure is associated with the chemical potential and is expressed explicitly
π = ξRT [c + A _{0} c^{2} + A _{1} c^{3} + ··· ];
Δπ = π  π_{0} (8)
where π_{0} is the initial state of the system; ξ is the osmotic coefficient describing the character of the ion transport through the semipermeable membrane (tissue). ξ approaches 1 for an ideal membrane, i.e., no ions can pass through the tissue. R and T are the gas constant and absolute temperature, respectively. c is the effective NaCl concentration within the tissue which is related to the concentration of c^{} and c^{+} by [14]:
In a first order approximation, Eq. (8) takes the form of van't Hoff's equation:
π = ξR·T·c;
Δπ = π  π_{0}. (10)
The dependence of the chemical expansion stress, T _{ c }, on the ion concentration has not been quantified based on a strict theory. Eisenberg and Grodzinsky [14] proposed a chemical expansion stress as a function of c
where β_{0} and c _{β} are material parameters.
Another form of T _{ c }as a function of c^{} and c^{ F }was proposed by Lai et al. [4]
where a _{0} and κ are material parameters; γ and γ* are the activity coefficients of NaCl in the tissue and the external solution, respectively.
Conservation of mass requires
where ρ^{α} is the the apparent density for component α; the superscripts α = s, f, +,  represent the solid and fluid phase, and cations and anions, respectively; t is time.
Using the relations [4] ρ^{α} = c^{α} M^{α} (α = +, ) with M^{+} and M^{} being the atomic weight of Na^{+} and Cl^{}, respectively, Eq.(13) can be written for electrolytes as

D^{α} c^{α} = c^{α} (υ^{α}  υ^{ s }) α = , + (15)
where D^{α} is the diffusion coefficient for component α.
Fick's diffusion law (15), combined with mass conservation (Eq.(14)), results in the governing equation of the convective diffusion process
where the effects of the volume strain rate of the solid phase on the diffusion process have been neglected. The diffusion problem (Eq. (16)) can be solved using a convective, thermal analogy, i.e., c^{α} → θ with θ denoting temperature.
The equations of motion for fluid and electrolytes (c^{+} and c^{}) in the triphasic model of Lai et al. [4] [their Eqs. (30–32)] are reduced to Darcy's law (Eq. (6)) and Fick's law (Eq. (15)) when the effects of the electrical potential, the friction between ions and fluid, the friction between ions and anions, and the effects of the volumetric deformation rate of the solid phase on the fluid phase are neglected.
Solution Procedure
For a practical problem with three components, only c^{} needs to be solved, because c^{+} can be obtained from Eq. (3). In order to solve the triphasic problem, the triphasic constitutive relation (1) was converted into a biphasic equation [2] with thermal expansion,
ε = ε^{ e }+ ε^{ c }= ε^{ e }+ K Δc I (17)
where ε is the total tissue strain, ε^{ e }is the elastic strain associated with mechanical loading, and ε^{ c }is the strain associated with the chemical expansion stress ("thermal strain"); K = K(c) is a "thermal expansion coefficient", defined as
where E is the Young's modulus of the tissue.
 (a)
The initial c distribution within the tissue is obtained by solving the diffusion equation (16) using υ^{ s }= 0. The problem is equivalent to the diffusion of c through the tissue in a steadystate. The boundary condition, i.e., the distribution of c on the surface of the tissue, is known.
 (b)
The triphasic problem [Eqs. (1)] is solved using the distribution of c in the tissue obtained in the previous step. The problem is solved using a thermal analogous procedure, as discussed previously. From this step, the distributions of the stress/strain and the displacement rate of the solid phase (υ^{ s }) were obtained.
 (c)
The diffusion equation (16) is solved again using the updated υ^{ s }obtained in the previous step. This is realized using a convective thermal analogous procedure, as discussed previously.
The iterations [steps (b) and (c)] are continued until the problem has converged. The distributions of the stress/strain and the displacement rate of the solid phase (υ^{ s }) were extracted from the simulation results via a customized subroutine, and the iterations were run manually. In the present simulations, the effective salt concentration in the bath is assumed to be constant.
Numerical Tests
Parameter  Unit  Test A  Test B  Test C 

Specimen Dimension:  
Diameter, d  [mm]  30.00  6.40  6.40 
Thickness, h  [mm]  5.00  0.540  0.565 
Constant Parameters:  
Poisson's Ratio  []  0.10  0.10  0.10 
Diffusion Coefficient, D  [m^{2}/s]  0.20·10^{}  0.59·10^{9}  0.82·10^{9} 
Hydraulic Permeability, k  [m^{4}/N·s]  1.51·10^{10}  1.10·10^{15}  1.50·10^{15} 
Osmotic Parameter, ξRT  [MPa/M]  3.0·10^{4}  3.0·10^{4}  3.0·10^{4} 
Test A: Constant Aggregate Modulus  

H _{ A }  [MPa]  0.98 
Tests B & C: Variable Aggregate Modulus H _{ A }= + (  ) exp (c/c _{ H })  
 [MPa]  0.27 
 [MPa]  1.17 
c _{ H }  [M]  0.12 
Parameters for the Chemical Expansion Stress used in the Simulations. The parameters of the chemical expansion stress [i.e., Eq. (12)] for Test A were from Lai et al. [4] and Gu et al. [5]. The parameters for the chemical expansion stress [i.e., Eq. (11)] for Tests B and C were from Eisenberg and Grodzinsky [14]
Test A  

 
a _{0}  [MPa/c]  0.05 
κ  [M^{1}]  7.50 
γ  []  1.00 
γ*  []  1.00 
c ^{ F }  [M]  0.10 
Tests B & C  
 
β_{0}  [kPa]  31.0 
c _{β}  [M]  0.032 
Numerical Tests B and C were designed to reproduce the confined compression experiments by Eisenberg and Grodzinsky [14]. The articular cartilage specimens had a diameter of 6.40 mm, and had an original thickness of 0.540 mm and 0.565 mm for Tests B and C, respectively. The specimens were compressed initially by an amount of δ_{0} = 0.0560 mm and 0.0185 mm for Tests B and C, respectively. The initial NaCl concentrations in the bathing solution were c* = 0.10 and 0.005 M for Tests B and C, respectively. The specimen was kept in this state for 2000 seconds until the predicted force reached a steadystate (F _{0} = 2.31 N and 2.28 N for Tests B and C, respectively). Once the steadystate force had been reached, the NaCl concentration was changed suddenly from the initial value (c*) to 0.15 M and 0.10 M, for tests B and C, respectively, while the compression of the specimen was kept constant. The predicted and the measured [14] forces were normalized to the steadystate force F _{0}, and were plotted as a function of time. The aggregate modulus, H _{ A }, was assumed to be linearly dependent on c. Other parameters (D, k, and v) were assumed to be constant (Tables 1, 2, 3) [14].
Van't Hoff's equation [Eq. (10)] was used in all simulations to evaluate the osmotic pressure.
Results
The interaction between mechanical and chemical stresses and deformations during cartilage deformation was demonstrated in our numerical simulations. For example, a sudden change in the deformation state of the tissue caused a change in ion concentration across the thickness of the cartilage specimen (Fig. 2a and 2b), while a change in NaCl concentration caused to a change in stress within the articular cartilage (Fig. 4a,4b,4c).
Discussion and Conclusion
Experimental animal models have shown that articular cartilage becomes thicker [17, 18], softer [19], and more permeable [20] in the initial stages of osteoarthritis (OA). It is assumed that the adaptation and degeneration of articular cartilage is associated with the stressstrain state of articular cartilage during normal, everyday loading. The stressstrain state of articular cartilage is associated with the swelling and deformation of the cartilage, which is influenced by the changing mechanical and chemical environment. A general FEM tool using the triphasic theory, which bioengineers can use to analyse practical swelling and deformation behaviour for physiologically relevant problems, does not exist. Because of the mathematical identity between thermal and mass diffusion processes [21], it is possible to transfer the triphasic model into a suitable form for implementation into commercial finite element software to simulate the swelling and deformation behaviour of articular cartilage. The main difference between the proposed approach and the triphasic model [4] is that the electrical potential and its influence on the transfer of electrolytes (i.e., NaCl concentration) was neglected in the current analysis. The advantage of the proposed thermal analogous technique over previous studies [12, 13] is, that it accounts for the convective diffusion of NaCl concentrations and the Donnan osmotic pressure in the interstitial fluid.
Our simulations (Test A) (Figs. 2b and 3) on the timedependent distributions of concentrations on the central line of the specimen agree well with the onedimensional finite element results obtained by Simon et al. [10]. The displacement at the center of the specimen (z/h = 0.5) showed a sudden change (Fig. 2a) when the prescribed displacement at the specimen surface was changed using a step function (Fig. 2a). This behaviour was not predicted by Simon et al. (their Fig. 3) [10], who gave no explanation why they did not predict this behaviour which should occur when considering the articular cartilage as a continuous material.
In order to reach an initial compression force (F _{0}), the cartilage specimen must be compressed long enough for the force to reach a steadystate. It was found that the amount of initial compression to reach a given compression force in our numerical tests was smaller than the corresponding compression in the experiments by Eisenberg and Grodzinsky [14]. This difference in the magnitude of compression between theory and experiment is likely associated with the fact that in real tests, there is space between the lateral surface of the cartilage specimen and the chamber wall; while in our numerical simulations, the lateral boundary of the specimen was assumed to be fixed ideally.
The proposed approach differs from the poroelastic finite element formulation by Simon et al. [10] in that the fluid pressure, which is the sum of the hydrostatic and Donnan osmotic pressure, is governed by Darcy's law; while in the model by Simon et al. [10], the effect of the Donnan osmotic pressure was included into the fluid pressure through a "generalized Darcy's" law [Eq. (12b) in [10]].
Compared to the general triphasic model by Lai et al. [4] and Gu et al. [5], the effects of the electrical potential, the friction between ions and fluid, the friction between cations and anions, and the effects of the volumetric deformation rate of the solid phase on the fluid phase were neglected in the proposed approach. The effects of friction between cations and anions and the volumetric deformation rate of the solid phase on the fluid phase are believed to be negligible [5, 11].
The major difference between the proposed approach and the rigorous formulation by Sun et al.'s [11] is that the friction between ions and fluid was neglected in our model formulation. However, the effects of the friction between ions and fluid are included in the diffusion coefficient (D^{α}) in our model. Our model represents a reasonable approximation of complex biological systems and the associated problems, since the friction between ions and fluid cannot be measured directly, while the diffusion coefficient, which includes the effects of ions/fluid friction, can be determined experimentally.
In the present simulations, a constant osmotic coefficient, hydraulic permeability, and diffusion coefficient was used. However, in reality, these coefficients are not constant; they change as a function of the deformation state of the matrix, the ion concentration, and the electrical potential [6, 7]. Deformationand ion concentrationdependent coefficients (osmotic coefficient, hydraulic permeability, and diffusion coefficients) can be incorporated into the proposed approach.
In summary, we have developed an approach that converts the triphasic analysis into a problem of convective thermal diffusion coupled with a poroelastic thermal stress analysis. The proposed approach makes it possible for bioengineers to analyse triphasic physiological systems using commercially available FEM software.
Declarations
Acknowledgement
This study was supported by The Alberta Heritage Foundation for Medical Research, NSERC of Canada, the Medical Research Council of Canada, and The Arthritis Society of Canada.
Authors’ Affiliations
References
 Edwards J: Physical characteristics of articular cartilage. Proc Inst Mech Eng 1967, 181: 16–24.Google Scholar
 Mow VC, Kuei SC, Lai WM, Armstrong CG: Biphasic creep and stress relaxation of articular cartilage in compression?: theory and experiments. J Biomech Eng 1980, 102: 73–84.View ArticleGoogle Scholar
 Tombs MP, Peacocke AR: The osmotic pressure of biological macromolecules Clarendon Press, Oxford 1974.Google Scholar
 Lai WM, Hou JS, Mow VC: A triphasic theory for the swelling and deformation behaviours of articular cartilage. J Biomech Eng 1991, 113: 245–258.View ArticleGoogle Scholar
 Gu WY, Lai WM, Mow VC: A mixture theory for chargedhydrated soft tissues containing multielectrolytes: passive transport and swelling behaviours. J Biomech Eng 1998, 120: 169–180.View ArticleGoogle Scholar
 Gu WY, Lai WM, Mow VC: Transport of fluid and ions through a porouspermeable chargedhydrated tissue, and streaming potential data on normal bovine articular cartilage. J Biomech 1993, 26: 709–723.View ArticleGoogle Scholar
 Gu WY: A study of mechanoelectrokinetic properties of chargedhydratedsoft tissues and biomechanics of articular cartilage. Ph.D. Thesis, Columbia University 1994.Google Scholar
 Bryant MR, McDonnell PJ: A triphasic analysis of corneal swelling and hydration control. ASME J Biomech Eng 1998, 120: 370–381.View ArticleGoogle Scholar
 Hon YC, Lu MW, Xue WM, Zhou X: Numerical algorithm for triphasic model of charged and hydrated soft tissues. Computational Mechanics 2002, 29: 1–15. 10.1007/s0046600203071View ArticleGoogle Scholar
 Simon BR, Liable JP, Pflaster D, Yuan Y, Krag MH: A poroelastic finite element formulation including transport and swelling in soft tissue structures. ASME J Biomech Eng 1996, 118: 1–9.View ArticleGoogle Scholar
 Sun DN, Gu WY, Guo XE, Lai WM, Mow VC: A mixed finite element formulation of triphasic mechanoelectrochemical theory for charged, hydrated biological soft tissues. Int J Num Meth Eng 1999, 45: 1375–1402.View ArticleGoogle Scholar
 Myers ER, Lai WM, Mow VC: A contimuum theory and an experiment for the ioninduced swelling behaviour of articular cartilage. ASME J Biomech Eng 1984, 106: 151–158.View ArticleGoogle Scholar
 Kaufmann MV, Simon BR, Meyer PJ, Baldwin AL: Abaqus finite element analysis of a porohyperelastic material with swelling by thermal analogy. Advances in Bioengineermg 1995, ASME BEDVol. 31: 103–104.Google Scholar
 Eisenberg SR, Grodzinsky AJ: The kinetics of chemically induced nonequilibrium swelling of articular cartilage and corneal stroma. ASME J Biomech Eng 1987, 109: 79–89.View ArticleGoogle Scholar
 Katchalsky A, Curran PF: Nonequilibrium Thermodynamics in Biophysics. Harvard University Press, Cambridge 1967.Google Scholar
 Wu JZ, Herzog W, Epstein M: Evaluation of the finite element software ABAQUS for biomechanical modelling of biphasic tissues. J Biomech 1998, 31: 165–169. 10.1016/S00219290(97)001176View ArticleGoogle Scholar
 Kiviranta I, Jurvelin J, Tammi M, Saamanen AM, Helminen HJ: Weightbearing controls glycosaminoglycan concentration and articular cartilage thickness in the knee joints of young beagle dogs. Arthritis Rheum 1987, 30: 801–809.View ArticleGoogle Scholar
 Herzog W, Adams ME, Matyas JR, Brooks JG: A preliminary study of hindlimb loading, morphology, and biochemistry of articular cartilage in the acldeficient cat knee. Osteoarthritis and Cartilage 1993, 1: 243–251.View ArticleGoogle Scholar
 Jurvelin J, Kiviranta I, Tammi M, Helminen HJ: Softening of canine articular cartilage after immobilization of the knee joint. Clin Orthop 1986, 207: 246–252.Google Scholar
 Sah RLY, Grodzinsky AJ, Plaas AHK, Sandy JD: Effects of static and dynamic compression on matrix metabolism in cartilage explants. In Articular Cartilage and Osteoarthritis (Edited by: Kuettner KE, Schleyebach R, Peyron JG, Hascall V). New York, Raven Press 1991, 373–383.Google Scholar
 Welty JR, Wicks CE, Wilson RE: Fundamentals of momentum, heat and mass transfer Wiley, New York 1976.Google Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article: verbatim copying and redistribution of this article are permitted in all media for any purpose, provided this notice is preserved along with the article's original URL.