You are on page 1of 9

Open Physics 2020; 18: 1084–1092

Research Article

Hongyun Wang*, Wesley A. Burgei, and Hong Zhou

Analytical solution of one-dimensional Pennes’


bioheat equation
https://doi.org/10.1515/phys-2020-0197 therein). The bioheat equation attributed to Pennes is
received August 11, 2020; accepted October 15, 2020 the most commonly used formulation of heat transfer in
Abstract: Pennes’ bioheat equation is the most widely biological systems such as human skin [2].
used thermal model for studying heat transfer in biolo- In 1978, Foster et al. applied the Fourier transform to
gical systems exposed to radiofrequency energy. In their Pennes’ bioheat equation and obtained the temperature
article, “Effect of Surface Cooling and Blood Flow on the increase in a semi-infinite domain of tissue irradiated by
Microwave Heating of Tissue,” Foster et al. published an an incident electromagnetic beam perpendicular to the
analytical solution to the one-dimensional (1-D) problem, surface of tissue domain [3]. While an analytical expres-
obtained using the Fourier transform. However, their article sion of the temperature increase was given as a function of
did not offer any details of the derivation. In this work, we depth and time in ref. [3], no mathematical details regarding
revisit the 1-D problem and provide a comprehensive math- derivation were provided. Later, Nelson et al. used the
ematical derivation of an analytical solution. Our result cor- Laplace transform to obtain a formula for temperature at
rects an error in Foster’s solution which might be a typo in the tissue surface only [4]; they did not give an explicit for-
their article. Unlike Foster et al., we integrate the partial dif- mula for the temperature profile at locations other than the
ferential equation directly. The expression of solution has skin surface. The Laplace transform has also been employed
several apparent singularities for certain parameter values in studying anomalous diffusion equations with the decay
where the physical problem is not expected to be singular. exponential kernel [5]. Green’s functions for Pennes’ bioheat
We show that all these singularities are removable, and we equation in various cases were studied in refs. [6–9]. Exact
derive alternative non-singular formulas. Finally, we extend solutions in the form of infinite series were obtained in the
our analysis to write out an analytical solution of the 1-D case of one-dimensional (1-D) multiple planar tissue layers
bioheat equation for the case of multiple electromagnetic or spherical tissue layers, based on the method of separation
heating pulses. of variables and Green’s function method [10,11]. However,
none of these efforts addressed the fundamental question of
Keywords: Pennes’ bioheat equation, blood flow cooling, how to explicitly obtain the analytical solution of Pennes’
surface cooling, analytical solution, pulsed heating bioheat equation. Therefore, in this work we revisit the 1-D
bioheat transfer problem governed by Pennes’ equation. We
present a detailed derivation of the analytical solution for RF
irradiation of semi-infinite media. The mathematical formu-
1 Introduction lation contains (i) a heating-source term (electromagnetic
beam), (ii) a temperature-restoring term (blood flow), and
Heating of biological systems by radiofrequency (RF) (iii) a boundary condition reflecting surface cooling (radia-
energy has many applications, and it has been studied tion and convection). Our approach to the solution relies on
intensively in the literature (see ref. [1] and references transforming this boundary value formulation of non-homo-
geneous heat equation to an initial value problem of the
standard heat equation.

* Corresponding author: Hongyun Wang, Department of Applied
Mathematics, University of California, Santa Cruz, CA 95064,
United States of America, e-mail: hongwang@soe.ucsc.edu
Wesley A. Burgei: U.S. Department of Defense, Joint Intermediate
Force Capabilities Office, Quantico, VA 22134, United States of
2 Mathematical formulation
America
Hong Zhou: Department of Applied Mathematics, Naval Following ref. [3], we consider the problem of microwave
Postgraduate School, Monterey, CA 93943, United States of America heating of skin tissue coupled with the effects of surface

Open Access. © 2020 Hongyun Wang et al., published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0
International License.
Analytical solution of one-dimensional Pennes’ bioheat equation  1085

cooling and blood flow. Heating the skin by other RF cooling. Mathematically, Tp(x′) can be obtained through
energy such as millimeter waves can be studied in the solving a boundary value problem (BVP)
same formulation with different parameter values. We
 d2Tp
treat the living tissue as semi-infinite, homogeneous, 0 = − λTp
 dx ′ 2
and isotropic media and thereby only focus on a 1-D  dT (x ′) (2)
case with constant thermal parameters. We first introduce  p = α(Tp(0) + T0 − Te).
 dx ′
notations for independent variables, functions, and para-  x ′= 0
meters, similar to those used in ref. [3]:
Solving (2) gives us
• (x′ , t′) ≡ (depth, time): original independent variables
before scaling α(Te − T0)
Tp(x′) = e−x ′ λ
. (3)
• (x , t ) ≡ (depth, time): new independent variables after α+ λ
scaling
To separate out the effect of initial temperature profile
• T0: temperature of the arterial blood entering the tissue,
Tp(x′) in (1), we introduce
independent of (x′ , t′)
• Te : temperature of the environment outside the skin, Th(x′ , t′) ≡ T (x′ , t′) − Tp(x′). (4)
independent of (x′ , t′)
Th(x′ , t′) is the temperature increase attributed to micro-
• T ′(x′ , t′): temperature of the tissue as a function of
wave heating. It satisfies
(x′ , t′)
• T (x′ , t′) ≡ T ′(x′ , t′) − T0: temperature of the tissue rela-  ∂Th ∂ 2Th −x ′/ L
tive to T0.  μ ∂t ′ = ∂x ′2 − λTh + q0 e

 ∂Th(x ′ , t ′) (5)
Here, the skin surface is at x′ = 0 (zero depth). The = αTh(0, t′)
 ∂x ′
 x ′= 0
relative temperature T (x′ , t′) is governed by an initial Th(x′ , 0) = 0.

boundary value problem (IBVP) of Pennes’ bioheat
equation: IBVP (5) constitutes the mathematical formulation for our
analysis.
 ∂T ∂ 2T −x ′/ L
 μ ∂t ′ = ∂x ′2 − λT + q0 e

 ∂T (x ′ , t ′) (1)
= α(T (0, t′) + T0 − Te)
 ∂x ′

x ′= 0 3 Transforming the formulation to
T (x′ , 0) = Tp(x′),

an initial value problem of the
where μ ≡ ρCp/ k , λ ≡ Vs/ k , q0 ≡ P0/(kL), and
• ρ : mass density of the tissue
standard heat equation
• Cp : specific heat of the tissue
IBVP (5) contains a non-insulated boundary condition
• k: heat conductivity of the tissue
simulating the surface cooling, a source of microwave
• Vs : product of flow rate, mass density, and specific heat
heating, and a temperature restoring term caused by
of the blood flow
blood flow. The objective of this section is to transform
• P0: microwave power absorbed into the tissue per unit
(5) to an initial value problem (IVP) of the standard heat
skin surface area
equation, which allows us to write out an analytical solu-
• L: reciprocal of absorption coefficient of the micro-
tion. We start by scaling (x′ , t′ , Th) to simplify the coeffi-
wave frequency used, describing the depth at which
cients in (5). We define new variables, temperature func-
the microwave power deposition is reduced by a
tion, and parameters as follows:
factor e.
The effect of surface cooling on temperature evolu-  x ≡ x′/ L
tion is described by coefficient α ≡ hc / k , where hc is the t ≡ t′/(μL2)

heat transfer coefficient of radiation and convection Ts(x , t ) ≡ Th(x′ , t′)/(q0 L2) (6)
cooling.  λ ≡ λL2
 s
The initial temperature profile of the tissue prior to  α s ≡ αL.

the start of microwave heating, denoted by Tp(x′), is the
steady state of the skin tissue, corresponding to the bal- The scaled temperature Ts(x , t ) as a function of scaled
ance between the blood flow warming and the surface independent variables satisfies
1086  Hongyun Wang et al.

+∞
 ∂Ts ∂ 2Ts −x 1 αs
 ∂t = ∂x 2 − λ s Ts + e v(x , t ) = u(x , t ) + ∫ u(ξ , t ) dξ . (13)
 1 + αs 1 + αs
 ∂Ts(x , t ) (7) x

 ∂x = αs Ts(0, t )
 x=0 It is straightforward to verify that v(x , t ) is governed by
Ts(x , 0) = 0.

 ∂v ∂ 2v
Once Ts(x , t ) is solved, T (x′ , t′) can be expressed in the  ∂t = ∂x 2

form of  ∂v(x , t ) (14)
 ∂x =0
 x′ t′  α(Te − T0) −x ′  x=0
T (x′ , t′) = (q0 L2) Ts  , 2  + e λ
. (8)  v(x , 0) = e−x .

 L μL  α+ λ
The insulated boundary condition for v(x , t ) allows us to
Our goal is to transform (7) to an IVP of the standard heat
extend it as an even function of x. After the even extension,
equation. On the road to that destination, our first step
(14) is converted to an IVP of the standard heat equation:
is to remove the inhomogeneous term in the partial dif-
ferential equation (PDE). We apply Duhamel’s principle  ∂v ∂ 2v
 = 2
to move the effect of the heating into the initial value.  ∂t ∂x (15)
Duhamel’s principle provides a convenient approach for  v(x , 0) = e−| x|.

obtaining a solution to the inhomogeneous heat equation
by utilizing the solution to the auxiliary homogeneous The solution of (15) can be written in terms of the
problem [12]. We consider the auxiliary IBVP of the Gaussian heat kernel as [12]
homogeneous PDE corresponding to (7): +∞ 2
1 −(ξ − x )

 ∂G ∂ 2G
v(x , t ) = ∫ 4πt
e 4t v(ξ , 0) dξ
 ∂t = ∂x 2 − λ s G −∞
 +∞
 ∂G(x , t ) (9) 1 e −(ξ4+tx)2 + e −(ξ4−tx)2  e−ξ dξ
 ∂x = αs G(0, t ) =
4πt
∫  
 x=0
0
G(x , 0) = e−x .
 +∞
1 e −(ξ 2+2(24t +t x) ξ +x 2)
Duhamel’s principle states that solutions of (7) and (9)
=
4πt
∫ 
0
are related by the equation [12]
+e
−(ξ 2 + 2(2t − x ) ξ + x 2)  dξ .

t 4t

Ts(x , t ) = ∫G(x, t − s) ds. (10)


0 Completing the square in the exponent of each inte-
gral, we write v(x , t ) as
Thus, the task becomes solving (9). Our next step is to get
rid of the temperature restoring term (−λ s G). We write  +∞
1  t+x −(ξ +(2t + x ))2
function G(x , t ) as v(x , t ) =
4πt 
e ∫e 4t dξ
 0
G(x , t ) ≡ e−λs t u(x , t ). (11)
+∞
 −(ξ +(2t − x ))2 (16)
It follows from (9) that u(x , t ) satisfies + et−x ∫edξ  4t

0 
 ∂u ∂ 2u
 ∂t = ∂x 2 et+x 2t + x et−x 2t − x
 = erfc + erfc ,
2 4t 2 4t
 ∂u(x , t ) (12)
 ∂x = αs u(0, t ) 2 +∞ 2

 x=0 where erfc(x ) ≡ π ∫ e−t dt is the complementary error


 u(x , 0) = e−x . x
function. In the first integral of (16), we have used change

(ξ + (2t + x ))
of variables s = :
The remaining obstacle in (12) is the surface cool- 4t
+∞ +∞
ing boundary condition. We write it mathematically −(ξ +(2t + x ))2 2 π 2t + x
as an insulated boundary condition on a transformed ∫ e 4t dξ = 4t ∫ e−s ds = 4t
2
erfc
4t
.
0 2t + x
function. We map function u(x , t ) to v(x , t ) as 4t
follows:
The second integral in (16) is treated in a similar way.
Analytical solution of one-dimensional Pennes’ bioheat equation  1087

4 Analytical solution then complete the square in the exponent to bring the
integral into the form:
Starting from the analytical solution v(x , t ) given in (16), +∞
we go step by step, to u(x , t ), to G(x , t ), to Ts(x , t ), and
finally write out T (x′ , t′) in original variables (x′ , t′).
A ≡ ∫ e−c ξ erfc(c1(ξ + c0)) dξ
2

x
+∞
−1
=
c2
∫ erfc(c1(ξ + c0)) d(e−c ξ ) 2

x
4.1 Solution of u(x , t ) e−c2 x 2c1
+∞
2 2
=
c2
erfc(c1(x + c0)) −
c2 π
∫ e−c (ξ +c ) e−c ξ dξ
1 0 2

We invert mapping (13) to express u(x , t ) in terms of x


v(x , t ). Let e−c2 x
= erfc(c1(x + c0))
+∞ c2
+∞
w(x , t ) ≡ ∫ u(ξ , t ) dξ . (17)

2c1 c22 /(4c12)+ c0 c2
e
2 2
∫ e−c (ξ +c +c /(2c )) dξ .
1 0 2 1
2

x c2 π
x
From equation (13), we get a linear differential equation
Expressing the second part in terms of erfc() and noting
on w(x , t ):
c12 = |c1 |, we obtain
∂w(x , t )
− + αs w(x , t ) = (1 + αs) v(x , t ).
∂x e−c2 x
A = erfc(c1(x + c0))
c2
We multiply it by the integrating factor e−αs x and then (19)
  c 
2 2
integrate it to solve for w(x , t ). This process yields e c2 /(4c1 )+ c0 c2
− sign(c1) erfc |c1 | x + c0 + 22  .
c2   2c1  
∂(e−αs xw(x , t ))
− = (1 + αs) e−αs xv(x , t )
∂x The condition of c1 > 0 or c2 > 0 is necessary for the con-
+∞
vergence of the improper integral A. When both c1 and c2
e−αs xw(x , t ) = (1 + αs) ∫ e−α ξ v(ξ , t ) dξ
s
are negative, the integrand does not decrease to zero
x
+∞
as ξ → +∞.
Formula (19) appears to have a singularity as c2 → 0.
w(x , t ) = (1 + αs ) e αs x ∫ e−α ξ v(ξ , t ) dξ .
s
However, for c1 > 0, the singularity is removable. When
x
c1 > 0, we rewrite (19) as
Differentiating w(x , t ) with respect to x gives u(x , t ) in
1 −zx z = c2
terms of v(x , t ): A = e ∣z = 0 erfc(c1(x + c0))
c2
∂w(x , t ) z = c2 (20)
u(x , t ) = − 1 2 2   z 
∂x − e z /(4c1 )+ c0 z erfc c1 x + c0 + 2   .
c2   2c1  
 +∞
 z=0
= (1 + αs) v(x , t ) − αs e αs x ∫ e−αs ξ v(ξ , t ) dξ  .
  In the limit of c2 → 0, both terms in (20) converge to
 x  derivatives of the corresponding functions with respect
(18) to z at z = 0. Thus, we have
Before we use (18) to derive an expression of u(x , t ), 1 2 2
lim A = −(x + c0)⋅ erfc(c1(x + c0)) + e−c1 (x + c0) .
we work out a general integration formula for c2 → 0 c1 π
+∞
A≡∫ e−c2 ξ erfc(c1(ξ + c0)) dξ in the case of c1 > 0 or With the help of formula (19), we evaluate the integral in (18)
x
c2 > 0 or both. We carry out integration by parts and by substituting expression (16) of v(x , t ) into the integral:
1088  Hongyun Wang et al.

+∞ +∞ t t
 et+ξ 2t + ξ
∫ e−αs ξ v(ξ , t ) dξ = ∫ e−αs ξ  2 erfc 4t Ts(x , t ) = ∫ e−λs(t − s) u(x , ∫
t − s) ds = e−λs τu(x , τ) dτ. (23)

x x 0 0

et−ξ 2t − ξ  We now use u(x , t ) obtained in (22) to calculate Ts(x , t ).


+ erfc dξ
2 4t 
t
+∞
et  −(αs − 1) ξ 2t + ξ Ts(x , t ) = ∫e−λ τu(x, τ) dτ
= ∫ e erfc
s

2  4t 0
x
t
2t − ξ  (1 + αs) x
+ e−(αs + 1) ξ erfc
4t 
dξ . =
2(1 − αs)
e ∫e(1−λ ) τerfc 2τ 4+τx dτ
s

0
We apply formula (19) on each part inside the brackets. t
1 −x (24)
The first part has c1 = 1/ 4t , c0 = 2t , and c2 = (αs − 1); the +
2
e ∫e(1−λ ) τerfc 2τ 4−τx dτ
s

second has c1 = −1/ 4t , c0 = −2t , and c2 = (αs + 1). The 0


integral becomes t
αs
+∞

1 − αs
e αs x ∫e(α −λ ) τerfc 2αs τ4τ+ x dτ
2
s s

et  e−(αs − 1) x2t + x 0
∫ e−α ξ v(ξ , t ) dξ
s = 
2  αs − 1
erfc
4t (1 + αs) 1 αs
x ≡ I1 + I2 − I3.
e−(αs + 1) x 2t − x 2(1 − αs) 2 1 − αs
+ erfc (21)
αs + 1 4t We perform integration by parts on each term in (24). For
2
2e−(1 − αs ) t 2αs t + x  I1, we have
+ erfc .
1 − αs2 4t  t
(1 − λ s) τ 
Putting expression (21) into equation (18), we find the I1 = e x ∫erfc 2τ 4+τx d  e1 − λs 
 
expression of u(x , t ): 0
e(1 − λs) t 2t + x
1 + αs t + x 2t + x 1 2t − x = ex erfc
u(x , t ) = e erfc + e t − x erfc 1 − λs 4t
(25)
2(1 − αs) 4t 2 4t t
(22) e(1 − λs) τ 2 −(2τ + x )  2τ + x 
2


αs 2 2
e αs x + αs t erfc s
α t + x
. + ex ∫ e 4τ d  
1 − αs 4t 1 − λs π  4τ 
0
I1B
≡ I1A + .
1 − λs
4.2 Several terms in the calculation of Ts(x , t ) In I1B , we combine all exponents and then complete the
square, which gives
Equations (10) and (11) give Ts(x , t ) in terms of u(x , t ):

∫e−λ τ− 4τ d  2τ 4+τx 
2 x2
I1B = s

π  
0
t
 1 + λs  2 λs τ + x  1 − λs  2 λs τ − x 
2
λs τ + x 
−
2
2 
=e λs x
∫ e  2 τ
  d   + d   
π  2 λ s  2 τ  2 λs  2 τ  
0 (26)
2 2
t  x + 2 λs τ  t  x − 2 λs τ 
1+ λs 2 −   x + 2 λs τ  1 − λs − 2 −   x − 2 λs τ 
= e λs x
π
∫ e  2 τ  d   − e λs x
∫ e  2 τ  d  
2 λs
0  2 τ  2 λs π
0  2 τ 
1+ λs x + 2 λs t 1 − λs − x − 2 λs t
= − e λs x erfc + e λ s x erfc .
2 λs 2 t 2 λs 2 t
Analytical solution of one-dimensional Pennes’ bioheat equation  1089

In a similar manner, I2 is calculated. The noticeable dif- In I3B , we gather all exponents and complete the square,
ference is that in the integration by parts of I2 , at t = 0+, which produces
we have erfc(−∞) = 2 , instead of erfc(+∞) = 0. t t 2
2 λs τ + x 
2 x 2  2α τ + x  2 − 2 τ 
t I3B = ∫ e−λs τ − 4τ d  s  =e
λs x
∫ e 

∫erfc 2τ 4−τx d  e1 − λs  
(1 − λ s) τ π 4τ  π
I2 = e−x 0 0
   αs + λ s  2 λ s τ +
0 x αs − λ s  2 λ s τ − x 
× d   + d   
e(1 − λs) t 2t − x 2  2 λ s  2 τ  2 λs  2 τ  
= e−x erfc − e−x
1 − λs 4t 1 − λs t
(27) αs + λs 2  x+2
−
λs τ 
2
 x + 2 λs τ 
t τ 
e(1 − λs) τ 2 −(2τ − x )  2τ − x 
2 = e λs x
π
∫ e 2  d  
+ e−x ∫ e 4τ d  
2 λs
0  2 τ  (30)
1 − λs π  4τ  2
0 t  x − 2 λs τ 
αs − λs 2 −   x − 2 λs τ 
I2B − e− λs x
∫ e 2 τ  d  
≡ I2A + . π
1 − λs 2 λs
0  2 τ 

In I2B , we collect all exponents, complete the square and αs + λs λs x erfc


x + 2 λs t
= − e
rearrange the terms, which yields 2 λs 2 t

t t αs − λs λs x erfc
x − 2 λs t
2 λs τ + x 
2
+ e− .
2 x  2τ
2
− x 2 − 2 τ  2 t
I2B = ∫ e−λs τ − 4τ d  =e
λs x
∫ e  2 λs
π  4τ  π
0 0

 1 + λs  2 λs τ − x
× d  
 2 λ s  2 τ 
4.3 Solution of Ts(x , t )
1− λs  2 λs τ + x 
+ d   
2 λs  2 τ   Incorporating results (25)–(30) into equation (24), we
2
t  x + 2 λs τ  obtain Ts(x , t ) in the form:
1− λs 2 −   x + 2 λs τ  (28)
= e λs x
π
∫ e 2 τ  d  
(1 + αs) e x +(1 − λs) t 2t + x
2 λs
0  2 τ  Ts(x , t ) = erfc
2 2(1 − αs) 1 − λ s 2 t
t  x − 2 λs τ 
1+ λs 2 −   x − 2 λs τ 
− e− λs x ∫ e 2 τ  d   (1 + αs) (1 + λ s ) e λs x x + 2 λs t
π − erfc
2 λs
0  2 τ  2(1 − αs) 2(1 − λ s) λ s 2 t
1− λs λs x erfc
x + 2 λs t (1 + αs) (1 − λ s ) e− λs x x − 2 λs t
= − e + erfc
2 λs 2 t 2(1 − αs) 2(1 − λ s) λ s 2 t
1+ λs λs x erfc
x − 2 λs t 1 e−x +(1 − λs) t 2t − x e−x
+ e− . + ⋅ erfc −
2 λs 2 t 2 1 − λs 2 t 1 − λs
For I3 , the calculation is again similar to that of I1. 1 (1 − λ s ) e λs x x + 2 λs t
− ⋅ erfc (31)
t
2 2(1 − λ s) λ s 2 t
 (αs2 − λ s) τ 
I3 = e αs x ∫erfc 2αs τ4τ+ x d  eα 2 − λ  +
1 (1 + λ s ) e− λs x
⋅ erfc
x − 2 λs t
0  s s 2 2(1 − λ s) λ s 2 t
2 2
e(αs − λs) t 2αs t + x αs e αs x +(αs − λs) t 2α t + x
= e αs x erfc − erfc s
αs2 − λs 4t 2
1 − αs αs − λ s 2 t
(29)
t
αs (αs + λ s ) e λs x
2
e(αs − λs) τ 2 −(2αs τ + x )  2αs τ + x  x + 2 λs t
2

+ e αs x ∫ αs2 − λ s π
e 4τ d   + 2
1 − αs 2(αs − λ s) λ s
erfc
2 t
 4τ 
0
I3B αs (αs − λ s ) e− λs x x − 2 λs t
≡ I3A + . − erfc .
2
αs − λ s 1 − αs 2(αs2 − λ s) λ s 2 t
1090  Hongyun Wang et al.

Combining like terms in (31), we express the final expres- 5.1 Case 1: λs = 1 and αs ≠ 1
sion for Ts(x , t ) as
We calculate lim Ts(x , t ). Let ε ≡ λ s − 1. We write
1  −x 1 + αs λs → 1
Ts(x , t ) = e +
λs − 1 
2 ( λ s − αs ) λ s = 1 + ε and write equation (32) in terms of ε
x + 2 λs t 1 + αs
⋅e λs x erfc − 1  −x 1 + αs
2 t 2( λ s + αs) Ts(x , t ) = e +
ε(2 + ε)  2(1 − αs + ε)
x − 2 λs t
λs x erfc x + 2t + 2εt 1 + αs
⋅ e− ⋅ e x + εx erfc −
2 t (32) 2 t 2(1 + αs + ε)
1 −x +(1 − λs) t 2t − x
− e erfc x − 2t − 2εt
2 2 t ⋅ e−x − εx erfc
2 t
(1 + αs) x +(1 − λs) t 2t + x
− e erfc 1 2t − x
2(1 − αs) 2 t − e−x − ε(2 + ε) t erfc (34)
2 2 t
αs(λ s − 1) 2 2α t + x 
+ e αs x +(αs − λs) t erfc s . (1 + αs) x − ε(2 + ε) t 2t + x
(1 − αs)(λ s − αs2) 2 t  − e erfc
2(1 − αs) 2 t
αs ε(2 + ε)
+
(1 − αs)(1 − αs2 + ε(2 + ε))
4.4 Analytical solution in original variables 2 2αs t + x 
× e αs x +(αs − 1) t − ε(2 + ε) t erfc .
(x ′, t ′) and comparison with Foster’s result 2 t 

Based on the expression of Ts(x , t ) we just derived in (32) In the limit as ε → 0, we have
and the scaling relations given in (6) and (8), we arrive at Ts(x , t )∣λs = 1 ≡ lim Ts(x , t )
the following formula for T (x′ , t′) in terms of the original λs → 1

variables (x′ , t′): 1 + αs  1  x x + 2t


= x + 2t −  e erfc
4(1 − αs)  1 − αs  2 t
q0  −x ′/ L 1/L + α
T (x′, t′) = e + 1 1  −x x − 2t
λ − 1/L2  2( λ − α) + x − 2t +  e erfc
4 1 + αs  2 t (35)
λ x′  x′ μ λt ′ 
⋅e erfc  +    2

2 t′ μ  + te−x −
1 − x + t 
t e  4t 
(1 − αs) π
1/L + α λ x′  x′ μ λt ′ 
− ⋅ e− erfc  −  αs 2α t + x
μ 
2
2( λ + α) 2 t′ + e αs x +(αs − 1) t erfc s .
(1 − αs)(1 − αs2) 2 t
1 − xL′ +  12 − λ  tμ′ 1 t′ x′ μ
− e L  erfc 
L −  (33)
2  μ 2 t ′ 

(1/L + α) xL′ +  12 − λ  tμ′ 1 t′ x′ μ 


− e  L  erfc  + 
2(1/L − α) L μ 2 t′  5.2 Case 2: λs = 1 and αs = 1
α(λ − 1/L2 ) 2 t′  t′
+ e αx ′+(α − λ) μ erfc α
(1/L − α)(λ − α2)  μ  
In order to evaluate lim lim Ts(x , t ) , we write equation
αs → 1 λ s → 1 
x′ μ   α(Te − T0) −x ′ λ
+   + e . (35) in terms of η ≡ αs − 1
2 t′  α+ λ
2 + η 1 x + 2t
lim Ts(x , t ) = x + 2t +  e x erfc
−4η 
The only difference between Foster’s result and ours is the
λs → 1 η 2 t
factor (λ − 1/L2 ) in the second to last term in (33). In ref. [3],
the corresponding factor was stated as (λ + 1/L2 ). 1 1  −x x − 2t
+ x − 2t +  e erfc
4 2 + η 2 t
(36)
 2 
1 − x + t 
+ te−x + t e  4t 
5 Removable singularities in (32) 1+η
η π
2t + 2ηt + x
+ 2 e x + ηx + η(2 + η) t erfc .
η (2 + η) 2 t
Expression (32) appears to have singularities when αs = 1
or λ s = 1 or λ s = αs2 . These singularities are removable.
We discuss these cases one by one.
Analytical solution of one-dimensional Pennes’ bioheat equation  1091

Taking the limit as η → 0 yields 5.4 Case 4: λs = 0 and αs = 0


 
Ts(x , t )| λs = 1, αs = 1 ≡ lim  lim Ts(x , t )  
αs → 1  λ s → 1  In computing lim  lim Ts(x , t ) , we calculate lim Ts(x , t )
λ s → 0αs → 0  αs → 0
1
= (x + 2t )2 + 2t and write it in terms of ε ≡ λ s
4
−  e x erfc 1  −x
1 x + 2t 1 εx x + 2εt
2
lim Ts(x , t ) = e + e erfc
2 t (37) αs → 0 ε2 − 1  2ε 2 t
1 1  −x x − 2t 1 −εx x − 2εt
+ x − 2t +  e erfc − e erfc
4 2 2 t 2ε 2 t
(40)
t 1 t−x 2t − x
+ te−x − (x + 2t − e erfc
2 π 2 2 t
 2 
− x + t  1 x+t 2t + x 
+ 1) e  4t  . − e erfc + O(ε 2).
2 2 t 

Letting ε → 0, we see that

 
Ts(x , t )| λs = 0, αs = 0 ≡ lim  lim Ts(x , t )
5.3 Case 3: λs = αs2 with αs ≠ 1 and αs > 0 λ s → 0 αs → 0 
1 t−x 2t − x 1 2t + x
= e erfc + e x + t erfc (41)
To determine lim2 Ts(x , t ), we introduce ε ≡ λ s − αs and 2 2 t 2 2 t
λ s → αs
x 2 t −x 2
substitute λ s = αs + ε into equation (32) to obtain − e−x − x ⋅ erfc + e 4t .
2 t π
1
Ts(x , t ) =
αs2 − 1 + O(ε)
 1 + αs x + 2αs t + 2εt
× e−x + ⋅ e αs x + εx erfc
 2ε 2 t
6 Solution for a sequence of
1 + αs −αs x x − 2αs t 2
− e erfc − e(1 − αs ) t
4αs 2 t heating pulses
1 2t − x (1 + αs) x 2t + x 
×  e−x erfc + e erfc  In formulation (5), the microwave power is on and
 2 2 t 2 (1 − α s ) 2 t 
remains constant after t > 0. Mathematically, after
αs(αs2 − 1 + ε(2αs + ε))
+ scaling transformation (6), the heating term q(x , t ) in
(1 − αs) ε(2αs + ε)
IBVP (7) is expressed in the form:
2αs t + x 
× e αs x − ε(2αs + ε) t erfc + O(ε). (38) q(x , t ) = H (t ) e−x ,
2 t 
where H (t ) is the Heaviside step function (H (t ) = 1 for
As ε approaches zero, we find that t > 0). In the case of constant power heating, the analy-
Ts(x , t )∣λs = αs2 tical solution of Ts(x , t ) for t > 0 is derived in Section 4
≡ lim 2 Ts(x , t ) and is given explicitly in (32). To facilitate the discussion
λ s → αs below, we extend function Ts(x , t ) to t ∈ (−∞ , +∞) as
1  e−x
 t − 4x t + αs2 t 
2
follows:
=  − e  
αs −1  αs + 1 π  0, for t ≤ 0
Ts(x , t ) =  (42)
1 −αs x x− 2αs t 2
(39)  given by (32 ) , for t > 0.
− e erfc − e(1 − αs ) t
4αs 2 t
We consider the case where the microwave power is a
 e−x 2t − x ex 2t + x 
× erfc + erfc  sequence of rectangular pulses. Each heating pulse is
 2(1 + αs) 2 t 2(1 − αs) 2 t 
specified by a starting time t j(s) and an ending time t j(e).
1 3αs2 + 1  αs x x + 2αs t  
For a sequence of M pulses, the heating term q(x , t ) in
+ x + 2α s t − 2  e erfc .
2 2αs(αs − 1)  2 t  

IBVP (7) takes the form:


1092  Hongyun Wang et al.

M
porting this work. The views expressed in this document
q(x , t ) = ∑(H (tj(s) ) − H (tj(e) )) e−x . (43)
are those of the authors and do not reflect the official policy
j=1
or position of the Department of Defense or the US
Let Ts(x , t ; {t j(s), t j(s), 1 ≤ j ≤ M }) denote the tempera- Government.
ture increase caused by the sequence of microwave
pulses. The analytical solution of Ts(x , t ; {t j(s), t j(s) }) is given Conflict of interest: The authors declare that there is no
by a superposition of Ts(x , t ) conflict of interest.

Ts(x , t ; {t j(s), t j(s), 1 ≤ j ≤ M })


M
(44)
= ∑(Ts(x, t − tj(s) ) − Ts(x, t − tj(e) )).
j=1 References
[1] Foster KR, Ziskin MC, Balzano Q. Thermal response of human
skin to microwave energy: a critical review. Health Phys.
2016;111(6):528–41.
7 Conclusion [2] Pennes H. Analysis of tissue and arterial blood temperatures
in the resting human forearm. J Appl Physiol. 1948;1:
93–122.
The bioheat equation attributed to Pennes [2] is applied [3] Foster KR, Kritikos HN, Schwan HP. Effect of surface cooling
most often to study RF irradiation heating of biological and blood flow on the microwave heating of tissue. IEEE Trans
systems such as human skin. In ref. [3], Foster et al. Biomed Eng. 1978;25(3):313–6.
analyzed the 1-D Pennes equation describing the heating [4] Nelson DA, Walters TJ, Ryan KL, Emerton KB, Hurt WD,
Ziriax JM, et al. Inter-species extrapolation of skin heating
of a semi-infinite domain of skin tissue irradiated by an
resulting from millimeter wave irradiation: modeling and
incident microwave beam. They used the Fourier transform experimental results. Health Phys. 2003;84:608–15.
to obtain an analytical expression of the resulting tempera- [5] Yang X-J, Feng Y-Y, Cattani C, Inc M. Fundamental solutions of
ture increase as a function of time and depth from the skin anomalous diffusion equations with the decay exponential
surface. However, no detail regarding the mathematical kernel. Math Meth Appl Sci. 2019;42(11):4054–60.
[6] Gao B, Langer S, Correy PM. Application of the time-
derivation was given in their paper. Here, we revisited the
dependent Green’s function and Fourier transforms to the
1-D bioheat transfer problem to present a detailed mathe-
solution of the bioheat equation. Int J Hypertherm.
matical derivation of the analytical solution based on inte- 1995;11:267–85.
grating the PDE directly. Our analytical expression cor- [7] Yeung C, Atalar E. A Green’s function approach to local RF
rected an error in Foster’s result which might be a typo heating in interventional MRI. Med Phys. 2001;28:826–32.
[3]. The analytical expression has apparent singularities [8] Deng Z, Liu J. Analytical study on bioheat transfer problems
with spatial or transient heating on skin surface or inside
for several combinations of parameter values where the
biological bodies. J Biomech Eng. 2002;124:638–49.
physical problem is not expected to have any singularity.
[9] Wang H, Burgei W, Zhou H. A concise model and analysis for
We found that all these apparent singularities are remo- heat-induced withdrawal reflex caused by millimeter wave
vable and derived an alternative non-singular formula for radiation. Am J Oper Res. 2020;10:31–81.
each case. Finally, we extended the analytical solution to [10] Durkee JW, Antich PP, Lee CE. Exact-solutions to the multire-
the case where the skin tissue is heated by a sequence of RF gion time-dependent bioheat equation 1. Solution develop-
pulses each specified by a starting time and an end time. ment. Phys Med Biol. 1990;35:847867.
[11] Durkee JW, Antich PP. Exact-solution to the multiregion time-
dependent bioheat equation with transient heat-sources and
Acknowledgments: The authors acknowledge the Joint boundary-conditions. Phys Med Biol. 1991;36:345368.
Intermediate Force Capabilities Office of U.S. Department [12] Fritz J. Partial differential equations. 4th edn. New York:
of Defense and the Naval Postgraduate School for sup- Springer-Verlag; 1982.

You might also like