MPEMathematical Problems in Engineering1563-51471024-123XHindawi Publishing Corporation83480710.1155/2012/834807834807Research ArticleAnalytical Model Based on a Cylindrical Geometry to Study RF Ablation with Needle-Like Internally Cooled ElectrodeLópez MolinaJuan A.1RiveraMaría J.1BerjanoEnrique2WongKwok W.1Departamento de Matemática AplicadaInstituto Universitario de Matemática Pura y AplicadaUniversitat Politècnica de València46022 ValenciaSpainupv.es2Biomedical Synergy, Electronic Engineering Department, Universitat Politècnica de València, 46022 ValenciaSpainupv.es20121112012201204072011290920112012Copyright © 2012 Juan A. López Molina et al.This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Radiofrequency (RF) ablation with internally cooled needle-like electrodes is widely used in medical techniques such as tumor ablation. The device consists of a metallic electrode with an internal liquid cooling system that cools the electrode surface. Theoretical modeling is a rapid and inexpensive way of studying different aspects of the RF ablation process by the bioheat equation, and the analytical approach provides an exact solution to the thermal problem. Our aim was to solve analytically the RF ablation transient time problem with a needle-like internally cooled cylindrical electrode while considering the blood perfusion term. The results showed that the maximal tissue temperature is reached 3 mm from the electrode, which confirms previous experimental findings. We also observed that the temperature distributions were similar for three coolant temperature values (5C, 15C, and 25C). The differences were only notable in temperature very close to the probe. Finally, considering the 50C line as a thermal lesion mark, we found that lesion diameter was around 2 cm, which is exactly that observed experimentally in perfused hepatic tissue and slightly smaller than that observed in nonperfused (ex vivo) hepatic tissue.

1. Introduction

Radiofrequency (RF) ablation with internally cooled needle-like electrodes is widely used for medical techniques such as tumor ablation , treatment of autonomously functioning thyroid nodules , and cardiac ablation to cure arrhythmias . The device consists of an internally liquid cooled metallic electrode that cools the electrode surface (see Figure 1(a)). Theoretical models provide a fast and inexpensive way of studying different aspects of the RF ablation process with the bioheat equation . Unlike numerical solutions such as those based on the finite element method, the analytical approach provides an exact solution to the thermal problem in a simplified scenario. In the case of a needle-like electrode, the model is based on an ideal conductor of infinite length totally immersed in homogeneous tissue (see Figure 1(b))). As far as we know, only Haemmerich et al.  have developed an analytical model of RF ablation with cooled-needle cylindrical electrodes, but this model had two important limitations: (1) it only considered the steady-state solution and (2) did not include the blood perfusion term, which is crucial in RF ablation of well-perfused organs such as the liver. A similar study including the blood perfusion term in the steady-state case was developed by Yue et al. in . The aim of our study was thus to solve analytically the transient time problem of RF ablation with a needle-like internally cooled cylindrical electrode in well-perfused tissue, that is, considering the blood perfusion term, which is essential in the bioheat equation. As far as we are aware, this is the first full analytical solution obtained for this problem.

(a) Cluster of three internally cooled needle-like electrodes used to ablate biological tissues by means of RF currents. (b) Analytical model representing a simplified scenario of an ideal conductor with infinite length totally immersed in homogeneous tissue. The internal cooling was modeled by means of a Dirichlet’s thermal boundary condition with a constant temperature (TC) which corresponds with the coolant temperature inside the electrode. Accordingly, the theoretical model has one dimension (1D), that is, axis r.

2. Governing Equations

Temperature distribution in the tissue during RF ablation is mathematically obtained by solving the Pennes bioheat equation :ηcTt=(kT)+S-ηbcbωb(T-Tb), where η,  c, and k are the density, specific heat, and thermal conductivity of the tissue; respectively, ηb,  cb, and ωb are the density, specific heat, and perfusion coefficient of the blood, Tb is the blood temperature, and S represents the heat sources. Assuming all quantities η,  ηb,  c,  cb,  k, and ωb to be constant and that the heat source is independent on the polar angle θ and working in cylindrical coordinates (r,θ), (2.1) becomesηcTt(r,t)=k(2Tr2(r,t)+1rTr(r,t))+S(r,t)-ηbcbωb(T(r,t)-Tb).

The heat source used in the study (S) was the electrical power density (W/m3) dissipated into the tissue throughout the temporal interval [0,[. Due to the cylindrical geometry of the theoretical model, we employed the same formulation used by Haemmerich et al. , and hence we haveηcTt(r,t)=k(2Tr2(r,t)+1rTr(r,t))+j02r02σr2-ηbcbωb(T(r,t)-Tb), where j0 is the current density at the conductor surface, σ is the electrical conductivity, and r0 is the electrode radius.

The initial and boundary conditions in the case of an internally cooled electrode areT(r,0)=Tb,r>r0,limrT(r,t)=Tb,t>0,T(r0,t)=TC,t>0, where TC is the temperature fixed by the refrigeration of the electrode.

We change to dimensionless variablesρ=rr0,ξ=αtr02,V(ρ,ξ)=σkj02  r02(T(r0ρ,r02ξα)-Tb), (where α=k/cη) which leads us to the problem-(2Vρ2+1ρVρ)+Vξ+βV=1ρ2, (where we have defined β=ηbcbωbr02/k) with the following initial and boundary conditions:V(ρ,0)=0,ρ>1,limρV(ρ,ξ)=0,ξ>0,  ξ>0,V(1,ξ)=σk(TC-Tb)j02r02:=(-B), (Dirichlet’s boundary condition).

3. Preliminary Results about Bessel’s Functions

To solve the initial boundary value problem (2.8), (2.9), (2.10), (2.11), we shall need to use deep well-known properties of Bessel’s and modified Bessel functions of complex argument, which we present now to facilitate reading the paper. First we recall the expression of the modified Bessel functions I0(z) and K0(z) of first and second class and order 0I0(z)=m=0z2m22mm!Γ(m+1),K0(z)=-I0(z)log(z2)+m=01+(1/2)++(1/m)-γ22m(m!)2  z2m, and the expression of the Bessel function of second class and order 0Y0(z)=2π(J0(z)log(z2)-m=0(-1)m1+(1/2)++(1/m)-γ22m(m!)2z2m), (γ is the Euler-Mascheroni constant) which implies thatK0(z)=-I0(z)log(z2)+Z(z),Y0(z)=2π(log(z2)J0(z)-R(z)), where Z(z) and R(z) are even holomorphic functions on . Remark thatZ(zi)=R(z). Moreover, we shall need the following relations:K0(z)=-K1(z),zC,  I0(z)=J0(zi).

Finally we recall the asymptotic expansions for |z| of K0(z) and the modified Bessel functions of first class and integer order Iν(z) (see , 7.2.3), which will be necessary for delicate computations in the following sections:K0(z)~π2ze-z(1+O(1z)),if  |Arg(z)|3π2,Iν(z)~ez2πz(1+4ν2-12222z+(4ν2-12)(4ν2-32)242!(2z)2+O(1z3))+e-z+(ν+(1/2))πi2πz(1+O(1z)),if  -π2<Arg(z)<3π2,Iν(z)~ez2πz(1+4ν2-12222z+(4ν2-12)(4ν2-32)242!(2z)2+O(1z3))+e-z-(ν+(1/2))πi2πz(1+O(1z)),if  -π2<Arg(z)<3π2,  if  -3π2<Arg(z)<π2, for every ν{0}.

4. Resolution of the Initial-Boundary Value Problem

Taking Laplace’s transforms D(ρ,s,β):=𝔏[V(ρ,ξ)](ρ,s,β) with respect to ξ and using (2.9), (2.10), and (4.2), we obtainρ2d2Ddρ2+ρdDdρ-(s+β)ρ2D=-1s,limρD(ρ,s,β)=0,D(1,s,β)=-Bs. The homogeneous equation associated to (4.1) is a modified Bessel equation of order 0 with general solutionD(ρ,s,β)=C1(s)I0(ρs+β)+C2(s)K0(ρs+β). Since the Wronskian determinant of I0(z) and K0(z) is W(I0(z)K0(z))=-1/z (formula (19) of  3.71 in ), the method of variation of parameters gives usC1(s)=-1sK0(s+βρ)ρ,C2(s)=1sI0(s+βρ)ρ.

So we takeC1(s)=-1s1ρK0(s+βv)vdv+M1(s),C2(s)=1s1ρI0(vs+β)vdv+M2(s), where M1(s) and M2(s) are functions independent on ρ to be chosen in such a way that (4.2) and (4.3) hold.

First, we remark the expected but nontrivial fact thatL1:=limρI0(ρs+β)ρK0(vs+β)vdv=0. In fact, by L’Hôpital’s rule and formula (7) in Section  3.71 of , we haveL1=limρ-1ρs+βK0(ρs+β)I02(ρs+β)I1(ρs+β). Using the asymptotic expansions of I0(x),  I1(x), and K0(x) given in Section  7.23 in , we obtainlimxK0(x)I02(x)I1(x)=limx2πxe2x(1+O(1/x))(1+O(1/x))22πxexex(1+O(1/x))=0, and (4.7) follows easily. Analogously,L2:=limρK0(ρs+β)1ρI0(vs+β)vdv=limρI0(ρs+β)K02(ρs+β)ρs+β  K1(ρs+β)=limxe-x(1+O(1/x))(1+O(1/x))22x2e-x(1+O(1/x))=0.

As a consequence of (4.7) and (4.10), we need to chooseM1(s)=1s1K0(vs+β)vdv. Then, to satisfy (4.3), we obtainM2(s)=-BsK0(s+β)-I0(s+β)sK0(s+β)  1K0(vs+β)vdv, obtaining finallyD(ρ,s,β)=1sI0(ρs+β)ρK0(vs+β)vdv+1sK0(ρs+β)1ρI0(vs+β)vdv-B  K0(ρs+β)sK0(s+β)-I0(s+β)  K0(ρs+β)sK0(s+β)1K0(vs+β)vdv=ρL-1[f1](ρ,ξ,v)vdv+1ρL-1[f2](ρ,ξ,v)vdv-BL-1[f3](ρ,ξ,v)-1L-1[f4](ρ,ξ,v)vdv.

To compute the inverse Laplace transform 𝔏-1[D(ρ,s)], we proceed in several steps. All the involved functions have a branch point in s=-β, so we will use the Bromwich’s contour of Figure 2 where γ0 denotes a fixed positive real number such that all the used functions are holomorphic on the set Re(s)γ0. We denote for subsequent use 𝒟β:=]-,-β].

Bromwich’s integration contour.

4.1. Computation of <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M84"><mml:msup><mml:mrow><mml:mi>𝔏</mml:mi></mml:mrow><mml:mrow><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mrow><mml:mo stretchy="false">[</mml:mo><mml:mrow><mml:msub><mml:mrow><mml:mi>f</mml:mi></mml:mrow><mml:mrow><mml:mn>1</mml:mn></mml:mrow></mml:msub></mml:mrow><mml:mo stretchy="false">]</mml:mo></mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>ρ</mml:mi><mml:mo>,</mml:mo><mml:mi>ξ</mml:mi><mml:mo>,</mml:mo><mml:mi>v</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math></inline-formula>

As vρ in the integralρL-1[f1](ρ,ξ,v)vdv, we begin computing, for every vρ,F1(ρ,ξ,v):=L-1[f1](ρ,ξ,v):=L-1[I0(ρs+β)K0(vs+β)s](ρ,ξ,v).

As vρ, it follows easily from the asymptotic expansions (3.8) and (3.9) that if s𝒟β and |s+β| is large enough in order that |O(1/ρs+β)|1/2,  |O(1/vs+β)|1/2, and |β/s|1/2 and, moreover, Re(s)γ0 and Im(s)>0, we have the estimation|I0(ρs+β)K0(vs+β)s|9(|e-(v-ρ)s+β|+|e-(v+ρ)s+β|)8vρ|s||s+β|9(|e-(v-ρ)|s+β|  cos(Arg(s+β)/2)|+|e-(v+ρ)|s+β|cos(Arg(s+β)/2)|)8vρ|s||s+β|92vρ|s|3/2, because cos(Arg(s+β)/2)0 and |1+β/s|1-1/2=1/2. A similar result holds if Im(s)<0 by (3.10). Then we can to apply Bromwich’s formula to find 𝔏-1[f1](ρ,ξ,v).

The residue in the pole s=0 isR(s=0)=lims0sesξI0(ρs+β)K0(vs+β)s=K0(vβ)I0(ρβ). On the other hand, by (3.1) we obtainlims-β,sDβ(s+β)esξ  I0(ρs+β)K0(vs+β)s=0. Hence,L-1[f1]=K0(vβ)I0(ρβ)-12πi--βesξsK0(v-s-βi)I0(ρ-s-βi)ds+12πi--βesξsK0(-v-s-βi)I0(-ρ-s-βi)ds. Having in mind (3.7) and (3.3), we obtainL-1[f1]=K0(vβ)I0(ρβ)-12πi--βπi2eyξyJ0(ρ-y-β)(J0(v-y-β)+iY0(-v-y-β))dy+12πi--βπi2eyξyJ0(ρ-y-β)(J0(v-y-β)+iY0(v-y-β))dy=K0(vβ)I0(ρβ)+12--βeyξyJ0(ρ-y-β)J0(v-y-β)dy, because the imaginary part of log(z) with z𝒟β and Re(z)<β and Im(z)>0 has limit πi if z approach to the real axis. In definitive, after the change -y=x,F1(ρ,ξ,v)=L-1[f1]=K0(vβ)I0(ρβ)-12βe-xξxJ0(ρx-β)J0(vx-β)dx.

4.2. Computation of <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M114"><mml:msup><mml:mrow><mml:mi>𝔏</mml:mi></mml:mrow><mml:mrow><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mrow><mml:mo stretchy="false">[</mml:mo><mml:mrow><mml:msub><mml:mrow><mml:mi>f</mml:mi></mml:mrow><mml:mrow><mml:mn>2</mml:mn></mml:mrow></mml:msub></mml:mrow><mml:mo stretchy="false">]</mml:mo></mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>ρ</mml:mi><mml:mo>,</mml:mo><mml:mi>ξ</mml:mi><mml:mo>,</mml:mo><mml:mi>v</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math></inline-formula>

To find1ρL-1[f2](ρ,ξ,v)dv, we have vρ, and hence, as a consequence of the previous result, we obtainF2(ρ,ξ,v):=L-1[f2](ρ,t,v):=L-1[I0(vs+β)K0(ρs+β)s](ρ,ξ,v)=F1(v,ξ,ρ).

4.3. Computation of <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M118"><mml:msup><mml:mrow><mml:mi>𝔏</mml:mi></mml:mrow><mml:mrow><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mrow><mml:mo stretchy="false">[</mml:mo><mml:mrow><mml:msub><mml:mrow><mml:mi>f</mml:mi></mml:mrow><mml:mrow><mml:mn>3</mml:mn></mml:mrow></mml:msub></mml:mrow><mml:mo stretchy="false">]</mml:mo></mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>ρ</mml:mi><mml:mo>,</mml:mo><mml:mi>ξ</mml:mi><mml:mo>,</mml:mo><mml:mi>v</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math></inline-formula>

Now we computeF3(ρ,ξ):=L-1[f3](ρ,ξ):=L-1[K0(ρs+β)sK0(s+β)](ρ,ξ). By (3.8), if s𝒟β,  Re(s)γ0, and |s+β| is large enough in order that |O(1/ρs+β)|1/2 and |O(1/s+β)|1/2, we have|K0(ρs+β)sK0(s+β)|1ρ1|s||e-(ρ-1)s+β||1+O(1/ρs+β)||1+O(1/s+β)|3ρ|s|  e-(ρ-1)|s+β|cos(Arg(s+β)/2)3ρ|s|, (because cos(Arg(s+β)/2)0) and so Bromwich’s inversion formula cannot be used directly to find F3(ρ,ξ). To circumvent this complication we need to proceed in the wayF3(ρ,ξ)=L-1[K0(ρs+β)sK0(s+β)]=ddξL-1[K0(ρs+β)s2K0(s+β)], finding this last inverse with means of Bromwich’s contour of Figure 2 since|K0(ρs+β)s2K0(s+β)|3ρ|s|2, if |s+β| is large enough.

First, we remark that it follows easily from  (3.3) thatlims-βs+βs2K0(ρs+β)K0(s+β)=0. On the other hand, the function esξf3(s)/s has a pole of order 2 in s=0. As the function esξK0(ρs+β)K0(s+β)   has continuous partial derivatives of any order in some neighbourhood of s=0 and every fixed t, we can writeddξRes[esξK0(ρs+β)s2K0(s+β  )]s=0=ddξ(lims0ddss2esξK0(ρs+β)s2K0(s+β))=lims0sξesξK0(ρs+β)K0(s+β)=K0(ρβ)K0(β).

Finally, since R(z) is an even function, by (3.3), (3.5), and (3.4), we havelimr,ɛ0  12πi  (-L1esξK0(ρs+β)s2K0(s+β)ds-L2esξK0(ρs+β)s2K0(s+β)ds)=-12πi--βesξs2-I0(ρ-s-βi)(log(ρ-s-β/2)+(π/2)i)+Z(ρ-s-βi)-I0(-s-βi)(log(-s-β/2)+(π/2)i)+R(-s-βi)ds+12πi--βesξs2-I0(ρ-s-βi)(log(ρ-s-β/2)-(π/2)i)+Z(ρ-s-βi)-I0(-s-βi)(log(-s-β/2)-(π/2)i)+R(-s-βi)ds=-12πi--βesξs2-J0(ρ-s-β)i-Y0(ρ-s-β)-J0(-s-β  )i-Y0(-s-β)ds+12πi--βesξs2J0(ρ-s-β)i-Y0(ρ-s-β)J0(-s-β)i-Y0(-s-β)ds=-1π--βesξs2J0(ρ-s-β)Y0(-s-β)-Y0(ρ-s-β)J0(-s-β)J02(-s-β)+Y02(-s-β)ds.

As a consequence, putting s=-x, we obtain from (4.26) that F3(ρ,ξ):=L-1[K0(ρs+β)sK0(s+β)](ρ,ξ)=K0(ρβ)K0(β)+1πβe-xξxJ0(ρx-β)Y0(x-β)-Y0(ρx-β)J0(x-β)J02(x-β)+Y02(x-β)dx.

4.4. Computation of <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M141"><mml:msup><mml:mrow><mml:mi>𝔏</mml:mi></mml:mrow><mml:mrow><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mrow><mml:mo stretchy="false">[</mml:mo><mml:mrow><mml:msub><mml:mrow><mml:mi>f</mml:mi></mml:mrow><mml:mrow><mml:mn>4</mml:mn></mml:mrow></mml:msub></mml:mrow><mml:mo stretchy="false">]</mml:mo></mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>ρ</mml:mi><mml:mo>,</mml:mo><mml:mi>ξ</mml:mi><mml:mo>,</mml:mo><mml:mi>v</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math></inline-formula>

To compute1L-1[f4](ρ,ξ,v)vdv, we have v1 and so, in order to findL-1[f4](ρ,ξ,v):=L-1[K0(ρs+β)I0(s+β)K0(vs+β)sK0(s+β)](ρ,ξ,v), using the convolution theorem and our previous results, we can writeF4(ρ,ξ,v):=L-1[f4](ρ,ξ,v)=L-1[K0(ρs+β)sK0(s+β)](ρ,ξ,v)*L-1[I0(s+β)K0(vs+β)](ρ,ξ,v)=F3(ρ,ξ)*ddξL-1[I0(s+β)K0(vs+β)s](ρ,ξ,v)=F3(ρ,ξ)*ddξF1(1,ξ,v)=12  F3(ρ,ξ)*βe-xξJ0(x-β)J0(vx-β)dx=12  0ξF3(ρ,w)(βe-x(ξ-w)J0(x-β)J0(vx-β)dx)dw.

4.5. Complete Solution

Collecting our previous results, by the second translation theorem, Fubini’s theorem, and some natural computations, we obtain finallyV(ρ,ξ)=ρ1v(K0(vβ)I0(ρβ)-12βe-xξxJ0(ρx-β)J0(vx-β)dx)dv+1ρ1v(K0(ρβ)I0(vβ)-12βe-xξxJ0(vx-β)J0(ρx-β)dx)dv-BF3(ρ,ξ)-12  11v(0ξF3(ρ,w)(βe-x(ξ-w)J0(x-β)J0(vx-β)dx)dw)dv, that can be rewritten in the following way which is more suitable for numerical computations:V(ρ,ξ)=11vK0(vβ)I0(ρβ)dv-1ρ1vK0(vβ)I0(ρβ)dv+1ρ1vK0(ρβ)I0(vβ)dv-BF3(ρ,ξ)-1211v(βe-xξxJ0(ρx-β)J0(vx-β)dx)dv-1211v(0ξF3(ρ,w)(βe-x  (ξ-w)J0(x-β)J0(vx-β)dx)dw)dv.

It is interesting to check that, as expected, actually we have V(1,ξ)=-B at every time ξ>0. In fact, as F3(1,ξ)=1 if ξ>0, in the computation of V(1,ξ), we can use Fubini’s theorem and obtain explicitly the integration with respect to w, which is0ξe-x(ξ-w)dw=1x-e-xξx. Finally, after the variable change z=x-β and the application of formula (5) in , Section  13.53, we obtainβJ0(x-β)J0(vx-β)xdx=2  0J0(z)J0(vz)z2+βzdz=2I0(β)K0(vβ), and this gives us V(1,ξ)=-B.

5. Results and Discussion

Once the solution was achieved, we observed that it was hard to make a direct plot of the temperatures with Mathematica 6.0 software (Wolfram Research Inc., Champaign, IL, USA). The computer took around 24 hours for each plot at a fixed time t.

Although we had continuously employed dimensionless variables in the analytical solution of the problem, as here we considered the case of liver RF ablation, we used the following values for the hepatic tissue characteristics: density (η) of 1060 kg/m3, specific heat (c) of 3600 J/kg·K, thermal conductivity (k) of 0.502 W/m·K, and electrical conductivity (σ) of 0.33 S/m . Since this was a case of well-perfused tissue, we considered the following blood characteristics: density (ηb) of 1000 kg/m3, specific heat (cb) of 4148 J/kg·K, and a perfusion rate (ωb) of 6.410-3s-1 . Blood temperature, and hence initial tissue temperature (Tb), was 37°C. For all the simulations, we used a current density (j0) of value 5 mA/mm2.

In order to assess the effect of different coolant temperatures on the temperature profile, we used different boundary temperatures on the electrode surface (TC). Figure 3 shows the temperature profiles at different times for a coolant temperature of 5°C. Likewise Figures 4 and 5 show the same plots for coolant temperatures of 15°C and 25°C, respectively.

Temperatures at 60 s, 180 s, and 360 s and limit temperature (thick line) with a temperature of 5°C in the electrode surface and a current intensity 5 mA/mm2.

Temperatures at 60 s, 180 s, and 360 s and limit temperature (thick line) with a temperature of 15°C in the electrode surface and a current intensity 5 mA/mm2.

Temperatures at 60 s,180 s, and 360 s and limit temperature (thick line) with a temperature of 25°C in the electrode surface and a current intensity 5 mA/mm2.

The results showed that the maximal temperature in the tissue is reached 3 mm from the electrode (see Figures 35), which confirms previous experimental findings . We also observed that the temperature was rising until achieving a steady-state at infinite time (thick line in Figures 35).

We also observed that the temperature distributions were similar for the three values of coolant temperature (5°C, 15°C, and 25°C). The differences were only significant at temperatures very close to the probe. This finding also agrees with previous experimental results in which little difference was observed in lesion size when coolant temperature was varied . In this respect, if we considered the 50°C line as a thermal lesion mark  in Figures 35, the lesion diameter would be around 2 cm, which is exactly that observed experimentally in perfused hepatic tissue  and slightly smaller than that observed in nonperfused (ex vivo) hepatic tissue .

Our results were achieved with a current density of 5 mA/mm2, and it seems obvious that higher current density would cause bigger lesions. However, once the tissue temperature reaches 100°C, the charred tissue around the electrode creates a highly resistive electrical interface, which impedes further power deposition in the tissue . Since our analytical solution is not able to model the nonlinear processes involved in these phenomena, we chose a value of current density in our simulations to keep the tissue temperature always lower than 100°C. This was also used in the previous analytical model by Haemmerich et al. . Temperature dependence of tissue electrical and thermal conductivity was not taken into account. Taking this into account would make the problem nonlinear and most likely impossible to solve analytically. However, in real RF ablation experiments, this effect is present as well as phase transition, charring, and other hard to model effects. Therefore, although comparison of analytical solution to experimental results was favorable, future studies using numerical methods, such as finite element method, should be conducted. However, it is necessary to point out that the solution in (4.37) is exact, while that the solution provided by FEM is an approximation.

Since it is known that the density current pattern around a needle-like electrode is highly heterogeneous [8, 9], the value of 5 mA/mm2 chosen for our simulations cannot be related with the values of current usually employed in RF liver ablation (1-2 A). In spite of this, if we consider a 30 mm long and 0.75 mm radius electrode, the value of 5 mA/mm2 provides a current total of 0.7 A, which is a bit smaller than the values experimentally observed.

6. Conclusion

We have solved analytically the transient time problem of RF ablation with a needle-like internally cooled electrode by using the bioheat equation (i.e., considering the blood perfusion term). The temperature distributions computed from the theoretical model matched the experimental results obtained in previous studies, which suggests the utility of the model and its analytical solution to study the thermal performance of this kind of electrode.

Acknowledgments

This work received financial support from the Spanish “Plan Nacional de I + D + i del Ministerio de Ciencia e Innovación” Grant no. TEC2008-01369/TEC. The translation of this paper was funded by the Universitat Politècnica de València, Spain.

McGahanJ. P.LohS.BoschiniF. J.PaoliE. E.BrockJ. M.MonskyW. L.LiC. S.Maximizing parameters for tissue ablation by using an internally cooled electrodeRadiology201025623974052-s2.0-7795514050410.1148/radiol.09090662BaekJ. H.MoonW. J.KimY. S.LeeJ. H.LeeD.Radiofrequency ablation for the treatment of autonomously functioning thyroid nodulesWorld Journal of Surgery2009339197119772-s2.0-6904911503010.1007/s00268-009-0130-3ThiagalingamA.CampbellC. R.BoydA. C.EipperV. E.RossD. L.KovoorP.Cooled intramural needle catheter ablation creates deeper lesions than irrigated tip catheter ablationPacing and Clinical Electrophysiology20042779659702-s2.0-324277214610.1111/j.1540-8159.2004.00566.xBerjanoE. J.Theoretical modeling for radiofrequency ablation: state-of-the-art and challenges for the futureBioMedical Engineering Online20065242-s2.0-3364655665310.1186/1475-925X-5-24HaemmerichD.ChachatiL.WrightA. S.MahviD. M.LeeF. T.Jr.WebsterJ. G.Hepatic radiofrequency ablation with internally cooled probes: effect of coolant temperature on lesion sizeIEEE Transactions on Biomedical Engineering20035044935002-s2.0-024238033810.1109/TBME.2003.809488YueK.ZhangX.YuF.Analytic solution of one-dimensional steady-state Pennes' bioheat transfer equation in cylindrical coordinatesJournal of Thermal Science20041332552582-s2.0-944427042510.1007/s11630-004-0039-yWatsonG. N.A Treatise on the Theory of Bessel Functions1995Cambridge, UKCambridge University Pressviii+804Cambridge Mathematical Library1349110ChangI.Finite element analysis of hepatic radiofrequency ablation probes using temperature-dependent electrical conductivityBioMedical Engineering Online20032122-s2.0-194253389310.1186/1475-925X-2-12ChangI. A.NguyenU. D.Thermal modeling of lesion growth with radiofrequency ablation devicesBioMedical Engineering Online200431272-s2.0-1314429852410.1186/1475-925X-3-27BurdíoF.BerjanoE. J.NavarroA.BurdíoJ. M.GüemesA.GrandeL.SousaR.SubiróJ.GonzalezA.CruzI.CastiellaT.TejeroE.LozanoR.de GregorioM. A.RF tumor ablation with internally cooled electrodes and saline infusion: what is the optimal location of the saline infusion?BioMedical Engineering Online20076302-s2.0-3454781720410.1186/1475-925X-6-30AhmedM.mahmed@bidmc.harvard.eduBraceC. L.LeeF. T.Jr.GoldbergS. N.Principles of and advances in percutaneous ablationRadiology2011258235136910.1148/radiol.10081634