You are on page 1of 18

Mathematical Medicine and Biology Advance Access published July 17, 2009

Mathematical Medicine and Biology Page 1 of 18


doi:10.1093/imammb/dqp010

Analytical solution of the Pennes equation for burn-depth determination


from infrared thermographs

R ICARDO ROMERO -M ÉNDEZ†


Facultad de Ingenierı́a, Universidad Autónoma de San Luis Potosı́, San Luis Potosı́,
SLP, Mexico
J OEL N. J IM ÉNEZ -L OZANO AND M IHIR S EN
Department of Aerospace and Mechanical Engineering, University of Notre Dame,
Notre Dame, IN 46556, USA
AND

F. JAVIER G ONZ ÁLEZ


Instituto de Investigación en Comunicación Óptica, Universidad Autónoma de San Luis
Potosı́, San Luis Potosı́, SLP, Mexico
[Received on 8 October 2008; revised on 25 February 2009; accepted on 24 March 2009]

A serious problem in emergency medicine is the correct evaluation of skin burn depth to make the
appropriate choice of treatment. In clinical practice, there is no difficulty in classifying first- and third-
degree burns correctly. However, differentiation between the IIa (superficial dermal) and IIb (deep
dermal) wounds is problematic even for experienced practitioners. In this work, the use of surface
skin temperature for the determination of the depth of second-degree burns is explored. An analytical
solution of the 3D Pennes steady-state equation is obtained assuming that the ratio between burn depth
and the burn size is small. The inverse problem is posed in a search space consisting of geometrical
parameters associated with the burned region. This space is searched to minimize the error between the
analytical and experimental skin surface temperatures. The technique is greatly improved by using local
one-dimensionality to provide the shape of the burned region. The feasibility of using this technique and
thermography to determine skin burn depth is discussed.

Keywords: thermographs; Pennes equation; burns.

1. Introduction
Temperature is commonly used as an indication and as a means of monitoring the progression of sickness
(Clark & Goff, 1990). The temperature distribution of skin has been used for diagnosis (Park et al.,
2007; Ya’ish et al., 2007), for follow-up treatments (Weibel et al., 2006), and for the study of the
physiological functions of healthy individuals (Shusterman et al., 1997). In recent years, the sensitivity
of infrared sensors has improved considerably and is now close to 0.025◦ C, so that it is now possible
to measure small changes in surface temperature (Head & Elliott, 2002). The development of focal
plane technologies has enabled the acquisition of high-quality images for computational processing.
The emissivity of human skin has an almost constant value of 0.97 ± 0.02 between wavelengths of

† Email: rromerom@uaslp.mx

c The author 2009. Published by Oxford University Press on behalf of the Institute of Mathematics and its Applications. All rights reserved.
2 of 18 R. ROMERO-MÉNDEZ ET AL.

2 and 14 µm (Jones, 1998), behaving almost as a black body, and making it an ideal material for surface
temperature measurements with an infrared camera (Gore & Xu, 2003; Teich, 1996).
One outstanding problem in burn therapy is the correct evaluation of skin burn depth that can have
repercussions on the appropriate choice of treatment (Shakespeare, 1991). Current approaches assume
that shallower burn wounds that heal spontaneously within 3 weeks of the trauma should be treated con-
servatively, while deeper wounds that fail to heal during this period, need surgical intervention. However,
the clinical assessment of burn-wound depth is mainly based on visual inspection. The traditional ap-
proach distinguishes the following grades of burn wound: I (superficial), IIa (superficial dermal), IIb
(deep dermal) and III (full thickness of the skin). In clinical practice, even an inexperienced physician
has no difficulty in classifying first- and third-degree burns correctly. However, differentiation between
the IIa and IIb wounds is problematic even for the most experienced practitioners (Rumin et al., 2007;
McGill et al., 2007). The objective of the present research was to develop an inverse heat transfer
algorithm that will take an infrared thermograph of a burned area, measure the temperature difference
between the burned skin and the unaffected reference area (Gardner & Martin, 1994), and from that
determine burn depth. In previous investigations related to the treatment of burns, good correlation has
been shown between the difference in temperatures between burned and healthy skin and the depth of
the burn. Zhu & Xin (1999) determined that the correlation was better if the temperature difference was
between a burned area and another symmetrical area of the body, but this measurement may not always
be possible.
Although visual evaluation is the most commonly used clinical method of estimation of the depth of
burns, the difference in depth that determines whether or not a burn will heal within 3 weeks may be very
small. Several studies have found that visual evaluation has an accuracy of 50–70% when the diagnosis is
performed by a well-experienced surgeon (Shakespeare, 1991). It would thus be very helpful to develop
precise, non-invasive diagnostic tools for this purpose. Thermography presents such an opportunity; it
enables the surface temperature field to be determined from which some of the characteristics of the
burned tissue can be inferred.
In the present work, a bioheat transfer model for skin with known material properties is used to
develop a procedure to determine burn depth from thermographs. After some simplification, the mathe-
matical model for bioheat transfer in the skin is solved. This is used as the basis for calculation of burn
depth by determining the geometrical parameters of a given burn shape that minimize the mean squared
error between the analytical and experimental skin surface temperatures.

2. Governing equations
2.1 Mathematical model
The mathematical model for bioheat transfer developed here is based on the Pennes equation (Pennes,
1948). This was derived for the situation where blood vessels cross a region. Some of the assumptions of
the model are that the material is homogeneous and has isotropic thermal properties, the blood vessels
are isotropic and that the blood enters at arterial temperature but reaches tissue temperature before
leaving the arterial system. This model represents heat transfer in living tissue and considers metabolic
heat generation and convective heat transfer due to the circulation of blood. It is important to mention
that there are alternative models for bioheat transfer (Charny, 1992; Arkin et al., 1994), such as the
continuum model of Chen & Holmes (1980), that of Weinbaum et al. (1984) that takes into account the
existence of large and small blood vessels and other models that look at the tissue as a porous medium
(Khaled & Vafai, 2003). Procedures similar to the one described here can be adopted for these other
models.
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 3 of 18

FIG. 1. Geometry of burned tissue.

The steady-state Pennes equation for healthy tissue is


!
∂2T ∂2T ∂2T
k + + + Wb Cb (Ta − T ) + Q m = 0, (1)
∂x2 ∂ y2 ∂z 2

where T (x, y, z) is the tissue temperature, k is the tissue thermal conductivity, Wb is the blood perfusion
rate, Cb is the blood specific heat, Ta is the arterial temperature and Q m is the tissue metabolic heat
generation rate. x and y are the coordinates parallel to the surface of the skin, and z is positive measured
into the skin. For burned tissue, blood perfusion and metabolic heat generation cease to operate because
the tissue is dead.

2.2 Problem description


Consider a region formed of two layers as shown in Fig. 1. Layer 1 is a shallow, superficial burned
region, and layer 2 is a deep substrate that is healthy tissue. The burn has an arbitrary shape; its depth
is H = H (x, y) and it has a surface delimited by G. The governing equations for each region can be
written as

∂ 2 T1 ∂ 2 T1 ∂ 2 T1
+ + =0 for 0 6 z < H (x, y), (2)
∂x2 ∂ y2 ∂z 2
!
∂ 2 T2 ∂ 2 T2 ∂ 2 T2
k2 + + + Wb Cb (Ta − T2 ) + Q m = 0 for H (x, y) 6 z < ∞, (3)
∂x2 ∂ y2 ∂z 2
4 of 18 R. ROMERO-MÉNDEZ ET AL.

where subscripts 1 and 2 are used for the respective regions. The boundary conditions are
∂ T1
k1 = h(T1 − T∞ ) at z = 0, (4)
∂z
∂ T1 ∂ T2
k1 = k2 at z = H (x, y), (5)
∂z ∂z
T1 = T2 at z = H (x, y), (6)
T2 = finite as z → ∞, (7)

and with zero heat flux at all other boundaries. It is assumed that there is no thermal resistance between
the two layers. Here T∞ is the ambient temperature and h is the convective heat transfer coefficient
between the surface of the skin and the surrounding air.

2.3 Non-dimensionalization
We can scale the coordinates parallel and normal to the surface using the characteristic size, Rc , and
the characteristic depth, Hc , of the burned region, respectively. For example, we can take Rc and Hc to
be the maximum surface coordinate (i.e. either in x or y) and depth coordinate (i.e. in z) of the burn,
respectively. Rc is measurable from a thermograph. The temperature can be scaled by using the arterial
and ambient temperatures. Thus, the non-dimensional variables are
x y z T1,2 − Ta H
ξ= , η= , ζ = , θ1,2 = , φ(ξ, η) = . (8)
Rc Rc Hc T∞ − Ta Hc
The governing equations and boundary conditions become

∂ 2 θ1 ∂ 2 θ1 ∂ 2 θ1
 +  + =0 for 0 6 ζ < φ, (9)
∂ξ 2 ∂η2 ∂ζ 2
2
∂ θ2 2
∂ θ2 2
∂ θ2
 2 + 2 + − m 2 θ2 − Q =0 for φ 6 ζ < ∞, (10)
∂ξ ∂η ∂ζ 2
∂θ1
= Bi(θ1 − 1) at ζ = 0, (11)
∂ζ
∂θ1 ∂θ2
=κ at ζ = φ, (12)
∂ζ ∂ζ
θ1 = θ2 at ζ = φ, (13)
θ2 = finite as ζ → ∞, (14)

with zero normal temperature derivatives at all other boundaries. Here φ(ξ, η) defines the non-
dimensional shape of the burned region, and the non-dimensional parameters are

Wb Cb Hc2 Q m Hc2 Hc2 k2 h Hc


m2 = , Q=− , = , κ= , Bi = (15)
k2 k2 (T∞ − Ta ) Rc2 k1 k1

where  is the aspect ratio squared, κ is the thermal conductivity ratio and Bi is the Biot number. Physi-
cally, the parameter m is the ratio of the depth scale Hc and the thermal penetration depth (k2 /Wb Cb )1/2 .
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 5 of 18

TABLE 1 Standard properties of homogeneous, well-


irrigated tissue (Liu et al., 1999; Rabin & Satahovic,
2003) and convective conditions

Property Magnitude (units)


Cb 4200 (J/kg ◦ C)
k 0.2 (W/m ◦ C )
Wb 0.5 (kg/m3 s)
Qm 200 (W/m3 )
h 10 (W/m2 ◦ C)
Ta 36.5 (◦ C)
T∞ 22.5 (◦ C)
H 0.0025 (m)

2.4 Parameter values


Table 1 shows the typical thermal properties of homogeneous tissue (Liu et al., 1999; Rabin & Satahovic,
2003), and standard arterial and ambient temperatures. Taking Hc = 2.5 mm and Rc = 25 mm as
being typical of second-degree burns, the calculated values of , m 2 , Q and Bi are  = 0.01, m 2 =
6.5625 × 10−2 , Q = 4.46428 ×10−4 and Bi = 0.25. We have used κ = 2.

3. Temperature field for small burn depth approximation


The aspect ratio parameter  is small; furthermore, the perturbation problem is regular. An asymptotic
solution can thus be obtained in terms of a power series, but if we restrict ourselves to the leading order
the first two terms in (9) can be neglected to give

∂ 2 θ1
= 0. (16)
∂ζ 2
This essentially means that, due to the shallow nature of the burn, conduction of heat in the burned
region in a direction parallel to the surface of the skin is negligible compared to that normal to it. This
gives a linear temperature profile in the ζ -direction in the burned region
θ1 (ξ, η, φ) − θ1 (ξ, η, 0)
θ1 (ξ, η, ζ ) = θ1 (ξ, η, 0) + ζ. (17)
φ
Substituting the derivative of this in (11) and solving for the surface temperature gives
θ1 (ξ, η, φ) + φBi
θ1 (ξ, η, 0) = . (18)
1 + φBi
Because of (13), (17) and (18), (12) can be written as
∂θ2 θ2 − 1
= at ζ = φ. (19)
∂ζ κ(Bi−1 + φ)
6 of 18 R. ROMERO-MÉNDEZ ET AL.

Thus the burned layer, having no perfusion or metabolic heat generation, simply offers a conductive
thermal resistance to the flow of heat from the substrate to the surface of the skin. There is another
thermal resistance due to convection from the surface to the surroundings. Equation (19) combines
these two to provide a boundary condition for the healthy substrate. Since we have assumed that the
burned region is thin compared to its other dimensions, we can apply the condition at ζ = φ but assume
that the curved boundary between burned and healthy tissue can be approximated as a plane.
The temperature field in the healthy tissue layer can be determined from (10) to be

∂ 2 θ2 ∂ 2 θ2 ∂ 2 θ2
 +  + − m 2 θ2 − Q = 0, (20)
∂ξ 2 ∂η2 ∂ζ 2
with (14) and (19) as boundary conditions and zero normal derivatives at all other boundaries. There is
no approximation possible in the healthy skin (as was done for the burned skin) since the coordinates
parallel to and normal to the skin extend to infinity. In some regions of the domain, the term ∂ 2 θ2 /∂ζ 2
could be of the same order of magnitude as the others; the terms that appear to be of O() are nec-
essary to satisfy (14). In any case, there is no need to remove any of the terms of (10) to obtain a
solution.
A particular solution is

p Q
θ2 = − . (21)
m2
The homogeneous part of the solution is obtained from

∂ 2 θ2h ∂ 2 θ2h ∂ 2 θ2h


 +  + − m 2 θ2h = 0, (22)
∂ξ 2 ∂η2 ∂ζ 2
with
∂θ2h θ h − Qm −2 − 1
= 2 at ζ = φ, (23)
∂ζ κ(Bi−1 + φ)

where θ2h is finite as ζ → ∞, and the heat flux is 0 at all other boundaries. This can be solved by
separation of variables to give
∞ X
X ∞
 
θ2h = Apq cos( pπξ ) cos(qπη) exp {m 2 + p 2 π 2 + q 2 π 2 }1/2 (−ζ + φ) . (24)
p=0 q=0

The constant Apq can be obtained from


R1R1
(−Qm −2 − 1) 0
cos( pπξ ) cos(qπη)
0 κ(Bi−1 +φ){m 2 +p 2 π 2 +q 2 π 2 }1/2 −1 dξ dη
Apq = R1 R1 . (25)
2 2
0 cos ( pπξ ) dξ 0 cos (qπη) dη

The complete solution is then


XX∞ ∞
Q  1/2 
θ2 (ξ, η, ζ ) = − 2
+ Apq cos( pπξ ) cos(qπ η) exp m 2 + p 2 π 2 + q 2 π 2 (−ζ + φ) .
m
p=0 q=0
(26)
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 7 of 18

4. 1D approximation
In this section, we determine the temperature field using a 1D approximation for use later. Such
approximations are commonly made for the Pennes equation (Jiang et al., 2002; Zhao et al., 2005)
but rarely fully justified. In (9) and (10), we have terms of different orders of magnitude: there are terms
of order 1, m 2 and . If   1, the derivatives with respect to ξ and η can be eliminated from (9) and
(10), and the set of equations become 1D. Numerical values of the parameters in Section 2.4 indicate
that m 2 is of the same order and only slightly larger than , so that m 2  1 also. However, we choose
to keep the order m 2 term in (10) so that there is an exponential decay of the temperature with ζ .
Under this approximation, (9) and (10) can be written as
∂ 2 θ1
=0 for 0 6 ζ < φ, (27)
∂ζ 2
∂ 2 θ2
− m 2 θ2 − Q = 0 for φ 6 ζ < ∞. (28)
∂ζ 2
The boundary conditions are
∂θ1
= Bi(θ1 − 1) at ζ = 0, (29)
∂ζ
∂θ1 ∂θ2
=κ at ζ = φ, (30)
∂ζ ∂ζ

θ 1 = θ2 at ζ = φ, (31)

θ2 = finite as ζ → ∞. (32)
The solution to these equations is
 
mκ(1 + Qm −2 ) 1
θ1 = −ζ − + 1, (33)
1 + mφκ + mκ Bi−1 Bi

(1 + Qm −2 ) exp(−mζ ) Q
θ2 = − . (34)
1 + mφκ + mκBi−1 exp(−mφ) m 2
From (33), it is possible to evaluate the temperature at the surface at ζ = 0 of a layer of thickness
H = Hc for which φ = 1 everywhere, to get
mκ(1 + Qm −2 )
θs = 1 − . (35)
Bi + mκBi + mκ
Using the definitions given in (15), substituting into (35) and solving for Hc , we get
( )
k1 1 − Q̃ h
Hc = −1− , (36)
h 1 − θs (k2 Wb Cb )1/2
or
( )
k1 (θs − Q̃)(k2 Wb Cb )1/2 − h(1 − θs )
Hc = , (37)
h (1 − θs )(k2 Wb Cb )1/2
8 of 18 R. ROMERO-MÉNDEZ ET AL.

where Q̃ = −Q m /{Wb Cb (T∞ − Ta )}. The last expression is very useful since it provides a 1D approxi-
mation of the thickness of a burn given the dimensionless surface temperature.

5. Direct problem for given burn shape


5.1 Variable thickness burn
Given the shape of the burn H = H (x, y), the determination of the surface temperature field Ts =
Ts (x, y) is the direct problem. We will generate a surface temperature field from (26) and from the
following ancillary equation

T2 (x, y, H )k1 H −1 + hT∞


Ts (x, y) = , (38)
h + k1 H −1
which states that the heat that crosses the burned tissue by conduction must be equal to the heat that
leaves the burned surface by convection. The relation between the dimensionless and dimensional forms
of the temperature is given by (8) as

T2 (x, y, H ) = θ2 (ξ, η, φ)(T∞ − Ta ) + Ta . (39)

The determination of θ2 from (26) requires calculation of constants Apq from (25) for which the
integrals have to be evaluated. These integrals were obtained numerically using a trapezoidal rule with
1ξ = 1η = 0.02. The use of a finer mesh in the calculation of Apq did not make a significant difference.
The infinite summations of terms of the solution were truncated at 80 terms, since inclusion of more
terms did not affect the results.
We assume a Gaussian-shaped burn
!
x 2 + y2
H = Hc exp − , (40)
b Rc2

with axisymmetry. b is a burn spread parameter which is assumed to be known. The characteristic size
of the burn, Rc , is assumed to be externally measurable and therefore known. The shape and depth of
the burn then depend on two parameters: the aspect ratio  (since Hc =  1/2 Rc ) and the spread of the
burn determined by parameter b.
Figure 2 shows the surface temperature field, Ts (x, y), that is produced from a 3D Gaussian burn
described by (40) with Rc = 0.025 and Hc = 0.0025 (for which  = 0.01) and b = 0.1 using
the properties of Table 1. The axis are the x and y coordinates and the contours are isotherms. It is
possible to observe from the figure that the surface temperature is smallest at x = y = 0, where
the burn is thickest, and the maximum surface temperature is obtained for large values of x and y,
where the thickness of the burn is negligible; at (x 2 + y 2 )/Rc2 = 1 the thickness of the burn is
4.54 × 10−5 times smaller than the maximum thickness at x = y = 0. The temperature difference
between these two regions is nearly 1.5◦ C which is easily detectable in thermographs. This is in qual-
itative agreement with the evidence of reduction of surface temperature of burned skin compared to
healthy skin that has been reported in the literature (Newman, 1980; Cole et al., 1990; Zhu & Xin,
1999).
Figure 3 shows the temperature, T2 (x, 0, z − H ), that is produced at a plane y = 0. The axes are the
x and z coordinates and the contours are isotherms. The deeper tissue for large values of z exponentially
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 9 of 18

FIG. 2. Temperature contours at surface of skin, T (x, y, 0).

FIG. 3. Temperature contours in plane normal to skin, T (x, 0, z − H ).

approaches the body temperature Ta − (T∞ − Ta )Q/m 2 = 36.59◦ C, while the healthy skin closer to
the burned tissue is colder. The temperature at z − H = 0 represents the temperature at the interface
between healthy and burned tissue. The interface temperature at x = Rc is 31.97◦ C, approaching that
of the surface of healthy, unburned tissue. Even though the temperature at x = z = 0 is a maximum,
10 of 18 R. ROMERO-MÉNDEZ ET AL.

FIG. 4. Temperature profile inside a uniform burn, T (z); − · −, H = 1.25 mm; —, H = 2.5 mm; −−, H = 5 mm.

because of the high thermal resistance due to the burn thickness, the surface temperature directly above
is a minimum.

5.2 Constant burn thickness


It is important to know whether a simpler analysis based on 1D heat transfer is sufficiently accurate.
By simplifying the problem as described in Section 4, it is possible to obtain approximate results for
temperatures inside the tissue. Strictly speaking, these results are only exact when the burn is of constant
thickness.
Figure 4 shows the temperature inside a tissue composed of a burned layer and a healthy tissue sub-
strate. The characteristic burn size is Rc = 0.025 m. Three different burn thicknesses, 0.00125, 0.0025
and 0.005 m, are plotted. For each, the burn temperature is linear and the healthy tissue temperature ex-
ponentially approaches the core tissue temperature of Ta − (T∞ − Ta )Q/m 2 = 36.59◦ C for large values
of z. The temperature of the interface between healthy and burned tissue grows as the burn thickness is
increased, and the burned surface temperature decreases as the burn thickness is increased, a behaviour
that is similar to that described for Fig. 3.
Figure 5 shows the variation of the burned surface temperature as a function of the burn thick-
ness. For the tissue properties, ambient temperature and convective heat transfer coefficient detailed in
Table 1, the surface temperature approaches a value of 31.97◦ C as the thickness approaches 0; as the
burn thickness is increased the surface temperature decreases.

6. Inverse problem for given surface temperature field


The determination of the shape of a burn H = H (x, y) from an experimental surface temperature field
TsE (x, y) that is obtained from a thermograph requires the solution of an inverse heat transfer problem.
In practice, the surface temperature will be provided by a thermograph, but in this analysis we will use
the surface temperature field produced by the reference Gaussian shape described in Section 5.1 as if it
were the result of experimental measurement.
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 11 of 18

FIG. 5. Skin surface temperature as a function of burn thickness for a uniform burn.

6.1 Procedure for burn shape determination


The procedure that follows consists of proposing a burn shape with some parameters to be determined.
Since this shape is not a priori known, the result of the calculations is somewhat dependent upon the
choice of that function. However, it must be remembered that, for the purpose for which it is being
proposed, extreme precision is not needed. In this section, different shapes of burn are proposed, and
the resulting H = H (x, y) is analysed. For example, one can propose a shape H = H (x, y; , a)
that has two free parameters,  and a, which have to be determined by a regression procedure. The
constraints on the function are that its first derivative with respect to x and y approaches 0 at x =
y = 0, x = Rc or y = Rc , so that the boundary conditions described in the problem description are
satisfied. The function should also be smooth and have a small  = Hc 2 /Rc 2 ratio. Once a function
is proposed, the difference between an experimentally measured surface temperature, TsE (x, y) and
the surface temperature Ts (x, y) obtained with the aid of (26), (38) and (39) can be determined. The
summation of the square of the errors on the surface temperatures determines the mean-squared error
Z Rc Z Rc
1  E 2
E(, a) = 2 Ts (x, y) − Ts (x, y) dx dy. (41)
Rc 0 0

The burn shape is obtained by determining the combination of  and a that globally minimizes E.
We will use several different burn shapes to test the algorithm. The experimental surface temperature
data will be generated as in Section 5.1, with the burn shape given in (40) with  = 0.01 and b = 0.1.
(a) Proposed shape: Gaussian
The first check will be with a function of the shape
!
x 2 + y2
H = Hc exp − , (42)
a Rc2
where parameters  and a are to be determined. Note that b in (40) is for the profile from which the
data are generated, while a in (42) is to be determined by the inverse method. Here  does not appear
12 of 18 R. ROMERO-MÉNDEZ ET AL.

explicitly in (42) but Hc relates to  through Hc =  1/2 Rc . For each value of  and a, we calculate the
error, E. Figure 6 shows E versus  and a. It is observed that there is only one minimum in the range
of parameters tested, and that, as expected, the minimum coincides with the values that were used to
produce the surface temperature information.
Thermographs, however, have a finite resolution in the temperature measurements, usually of the
order of 0.1◦ C that may produce errors in the results, so that the effect of finite round-off must be
studied. Table 2 shows E,  and a that are recovered when the temperatures are truncated downward
at 0.1◦ C, 0.25◦ C and 0.5◦ C. Figure 7 shows the shape of the burn obtained from different truncated
temperatures. The effect of downward truncation is to overestimate the thickness of the burn since the
truncation produces smaller readings than real temperatures and smaller temperatures are related to
thicker burns. Although the error E and the differences in shape increase, the inverse heat procedure
seems to be very robust, and the error in the shape of the burn is small. The resolutions of standard
thermographic cameras are thus suitable for this application.
(b) Proposed shape: inverse quadratic
A burn of inverse quadratic shape
!−1
x 2 + y2
H = Hc (43)
a Rc2

is proposed, which will also produce an axisymmetric burn. It is assumed that Rc is known, and the shape
and depth of the burn depends on two parameters: the aspect ratio , remembering that Hc =  1/2 Rc ,

FIG. 6. Error E as a function of parameters a and .

TABLE 2 Parameters  and a and error E recovered from thermographs with different round-off errors

Round-off error (◦ C)  a E
0.5 1.1 × 10−2 9.1 × 10−2 58.27 × 10−4
0.25 1.02 × 10−2 9.8 × 10−2 20.17 × 10−4
0.1 1.005 × 10−2 9.95 × 10−2 6.28 × 10−4
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 13 of 18

FIG. 7. Effect of round-off error on temperature field or shape of burned region; —, reference burn shape; −−, 0.1◦ C; − · −,
0.25◦ C; · · ··, 0.5◦ C.

FIG. 8. Error E as a function of  and a for quadratic shape burn.

and the spread of the burn determined by parameter a. Figure 8 shows the error versus the burned region
geometry. It is observed that there is again only one minimum in the range of parameters tested. The
minimum error E occurs at  = 0.0158 and a = 0.03.
(c) Proposed shape: constant thickness
A burn of constant thickness H = Hc can also be proposed. The thickness is only a function of
the aspect ratio , again with Hc =  1/2 Rc . It was found that the minimum error E occurs for  =
5.6234 × 10−5 .
14 of 18 R. ROMERO-MÉNDEZ ET AL.

FIG. 9. Determination of burn profile by optimal determination of shape parameters for given functions; −−, reference Gaussian
burn; —, quadratic profile; − · −, flat profile.

6.2 Comparison between proposed shapes


We now compare the values of the depths determined by assuming different burn shapes. Figure 9 shows
a comparison of the shape of the burn determined by the inverse procedure for the three
different functions proposed. Of course, the best results are obtained with the Gaussian since the surface
temperature field was generated with that burn shape. The inverse quadratic, however, overpredicts
the maximum thickness of the burn by only 25.69%. This may be acceptable for burn thickness mea-
surements since the goal is to distinguish between IIa and IIb burns and is much better than visual
inspection. The constant burn thickness function underpredicts the maximum burn depth by 92.5% and
is of little use.

7. Inverse problem using 1D approximations


The optimization procedure described above depends upon the shape of the burn that is proposed, which
of course is impossible to know beforehand. Here we will describe another approach that may help
circumvent this inconvenience.

7.1 Locally 1D analysis


We can use the 1D solution to the governing equations, (37), to relate the local burn thickness at any
location (x, y) to the burn depth at that point H (x, y) = H1D (x, y). Once again we generate surface
temperatures, TsE (x, y), using the direct procedure on (40). Figure 10 shows the results. The dotted line
corresponds to the Gaussian profile, (40), with  = 0.01 and b = 0.1, while the solid line corresponds
to the burn thickness that is determined by inverse calculation from (37). The locally 1D procedure
overpredicts the thickness by 15.6% and also produces an unphysical negative burn thickness at the
periphery. This is because this model does not take into account the heat flow in the x and y directions,
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 15 of 18

FIG. 10. Determination of burn profile using 1D solution; −−, reference Gaussian burn; —, predicted profile.

and thus the heat flow does not avoid the zones of high thermal resistance near the centre. As a conse-
quence, it overpredicts the heat flow near x = y = 0, and also overpredicts the burn thickness there.
The opposite happens in the zones of small thermal resistance, where the model underpredicts both the
heat flow and burn thickness.
The error produced by this locally 1D procedure is less than that of the more elaborate 3D method
using the inverse quadratic function. The unphysical negative burn depth is not a significant problem
since the maximum burn depth, important for the diagnosis, is usually what is sought.

7.2 Using locally 1D in a 3D analysis


Though the locally 1D approach is useful, it can be improved greatly by using its results to provide a
shape that the more accurate procedure of Section 6.1 can build upon. For instance, we can propose a
function of the following form
H = cH1D (x, y) + d, (44)

where H1D is determined by the locally 1D procedure, and the parameters c and d are to be obtained
by the optimization procedure of Section 6.1. Parameter c represents a scaling of the 1D burn depth
and parameter d a vertical translation of the curve that together minimize the error E. Parameter d in
particular may help correct the non-physical negative thickness of the burn at the periphery that was
found in Section 7.1.
It is seen that there is only one minimum for E, and that this occurs at c = 0.878 and d = 2.25 ×
10−5 . Figure 11 compares the shape of the burn determined by this procedure and the Gaussian shape,
(40), with which the surface temperature TsE (x, y) was produced. We can see that this procedure is
able to accurately reproduce both qualitatively and quantitatively the original Gaussian shape, with an
overprediction of the maximum thickness of the burn of only 2.41%.
16 of 18 R. ROMERO-MÉNDEZ ET AL.

FIG. 11. Determination of burn profile using optimized 1D solution shape; −− reference Gaussian burn, — predicted profile.

8. Conclusions
The objective of this work has been to develop an algorithm that will allow thermographic data to be used
to determine the depth of burned skin. For this purpose, an analytical solution of the bioheat transfer
problem in burned skin has been obtained and used to determine burn thickness from thermographic
data. The advantage of analytical versus numerical solutions lies in the speed with which the results
can be obtained. The present work assumes known properties of both burned and healthy tissues: the
thermal conductivity is assumed known for both, and the arterial temperature and metabolic heat rate
for the latter.
An asymptotic approximation, valid for shallow burns, is used to simplify the governing equations.
This enables the steady-state temperature field to be calculated analytically. Assuming a specific form for
the shape of the burned region, the characteristic depth can then be obtained from the surface temperature
field. The error in the calculated depth depends upon the proposed shape, which is not known a priori.
An inverse quadratic shape gives an overprediction of 25.69%. A locally 1D burn depth approximation
is shown to produce a good estimation of the burn shape. By itself it gives an error of around 15.6%
in the estimation of the maximum burn depth. However, if it is used to provide an initial shape, a 3D
analysis then reduces the error to just 2.41%.
As explained before, it would be very useful to have simple experimental techniques to determine the
depth of a burn. The procedure described in this paper is a step in the direction of using thermography
for this purpose. It has been shown that thermographic data may be used as a non-invasive tool for
estimating the depth of burns. Many other issues need to be addressed, however, before this procedure
can be successfully applied in practice. In this work, we have assumed that the material and heat transfer
properties were known, but in reality these parameters are available only within wide ranges and, even
for the same person, depend on many conditions. Future work will extend the procedures developed
here to situations in which these properties are also to be simultaneously determined using, for example,
additional information available from time-dependent video thermographs.
BURN DEPTH DETERMINATION USING THE PENNES EQUATION 17 of 18

Acknowledgement
The authors thank CONACyT (Mexico) for its support of RR-M for projects CB-2008-84618 and
FMSLP-C01-2007-62604 that enabled a research visit to the University of Notre Dame, JNJ-L for a fel-
lowship, and FJG for support through FMSLP-2005-C01-28, CB-2006-60349 and FMSLP-C01-87127.

Funding
Consejo Nacional de Ciencia y Tecnologı́a (CB-2008-84618 and CB-2006-60349); Fondos Mixtos—
San Luis Potosı́ (FMSLP-C01-2007-62604, FMSLP-C01-2005-28 and FMSLP-C01-87127).

R EFERENCES
A RKIN , H., X U , L. X. & H OLMES , K. R. (1994) Recent developments in modeling heat-transfer in blood-perfused
tissues. IEEE Trans. Biom. Eng., 41, 97–107.
C HARNY, C. K. (1992) Mathematical models of bioheat transfer. Adv. Heat Transf., 22, 19–155.
C HEN , M. M. & H OLMES , K. R. (1980) Microvascular contributions in tissue heat transfer. Ann. N. Y. Acad. Sci.,
335, 137–150.
C LARK , R. P. & G OFF , M. R. (1990) Medical thermography: current status. Proc. SPIE, 1320, 242–250.
C OLE , R. P., J ONES , S. G. & S HAKESPEARE , P. G. (1990) Thermographic assessment of hand burns. Burns, 16,
60–63.
G ARDNER , G. G. & M ARTIN , C. J. (1994) The mathematical modelling of thermal responses of normal subjects
and burned patients. Physiol. Meas., 15, 381–400.
G ORE , J. P. & X U , L. X. (2003) Thermal imaging for biological and medical diagnostics. Biomedical Photonics
Handbook, Chapter 17 (T. Vo-Dinh ed.). Boca Raton, FL: CRC Press, pp. 446–457.
H EAD , J. F. & E LLIOTT, R. L. (2002) Infrared imaging: making progress in fulfilling its medical promise. IEEE
Eng. Med. Biol. Mag., 21, 80–85.
J IANG , S. C., M A , N., L I , H. J. & Z HANG , X. X. (2002) Effects of thermal properties and geometrical dimensions
on skin burn injuries. Burns, 28, 713–717.
J ONES , B. F. (1998) A reappraisal of the use of infrared thermal image analysis in medicine. IEEE Trans. Med.
Imaging, 17, 1019–1027.
K HALED , A. R. A. & VAFAI , K. (2003) The role of porous media in modeling flow and heat transfer in biological
tissues. Int. J. Heat Mass Transf., 46, 4989–5003.
L IU , J., C HEN , X. & X U , X. (1999) New thermal wave aspects on burn evaluation of skin subjected to instanta-
neous heating. IEEE Trans. Biomed. Eng., 46, 420–428.
M C G ILL , D. J., S RENSEN , K., M AC K AY, I. R., TAGGART, I. & WATSON , S. B. (2007) Assessment of burn
depth: a prospective, blinded comparison of laser doppler imaging and videomicroscopy. Burns, 33, 833–842.
N EWMAN , P. (1980) A practical technique for the thermographic estimation of burn depth: a preliminary report.
Burns, 8, 59–63.
PARK , D., S EUNG , N. H., K IM , J., J EONG , H. & L EE , S. (2007) Infrared thermography in the diagnosis of
unilateral carpal tunnel syndrome. Muscle Nerve, 36, 575–576.
P ENNES , H. H. (1948) Analysis of tissue and arterial blood temperatures in the resting human forearm. J. Appl.
Physiol., 1, 93–122.
R ABIN , Y. & S ATAHOVIC , T. F. (2003) Cryoheater as a means of cryosurgery control. Phys. Med. Biol., 48,
619–632.
RUMIN , J., K ACZMAREK , M., R ENKIELSKA , A. & N OWAKOWSKI , A. (2007) Thermal parametric imaging in
the evaluation of skin burn depth. IEEE Trans. Biomed. Eng., 54, 303–312.
18 of 18 R. ROMERO-MÉNDEZ ET AL.

S HAKESPEARE , P. G. (1991) Looking at burn wounds: the A.B. Wallace Memorial Lecture. Burns, 18,
287–295.
S HUSTERMAN , V., A NDERSON , K. P. & BARNEA , O. (1997) Spontaneous skin temperature oscillations in normal
human subjects. Am. J. Physiol. Regul. Integr. Comp. Physiol., 273, R1173–R1181.
T EICH , J. S. (1996) Digital infrared imaging for medicine. Recent advances in I.R. focal plane array imaging
technology. Engineering in Medicine and Biology Society, 1996. Proceedings of the 18th Annual International
Conference of the IEEE, 31 October–3 November 1996. Piscataway, NJ: IEEE, pp. 2079–2080.
W EIBEL , L., S AMPAIO , M. C., V ISENTIN , M. T., H OWELL , K. J., W OO , P. & H ARPER , J. I. (2006) Evaluation
of methotrexate and corticosteroids for the treatment of localized scleroderma (morphoea) in children. Br. J.
Dermatol., 155, 1013–1020.
W EINBAUM , S., J IJI , L. M. & L EMONS , D. E. (1984) Theory and experiment for the effect of vascular microstruc-
ture on surface tissue heat-transfer. Part 1. Anatomical foundation and model conceptualization. Part 2. Model
formulation and solution. ASME J. Biomech. Eng., 106, 321–330 and 331–340.
YA’ ISH , F. M. M., C OOPER , J. P. & C RAIGEN , M. A. C. (2007) Thermometric diagnosis of peripheral nerve
injuries - Assessment of the diagnostic accuracy of a new practical technique. J. Bone Joint Surg., 89B,
933–939.
Z HAO , J. J., Z HANG , J., K ANG , N. & YANG , F. Q. (2005) A two level finite difference scheme for one
dimensional Pennes’ bioheat equation. Appl. Math. Comput., 171, 320–331.
Z HU , W. P. & X IN , X. R. (1999) Study on the distribution pattern of skin temperature in normal Chinese and
detection of the depth of early burn wound by infrared thermography. Ann. N. Y. Acad. Sci., 888, 300–313.

You might also like