Author’s Accepted Manuscript

Thermoregulation in premature infants: a

mathematical model

Carina Barbosa Pereira, Konrad Heimann, Michael

Czaplik, Vladimir Blazek, Boudewijn Venema,
Steffen Leonhardt
www.elsevier.com/locate/jtherbio

PII: S0306-4565(16)30028-6
DOI: http://dx.doi.org/10.1016/j.jtherbio.2016.06.021
Reference: TB1778
To appear in: Journal of Thermal Biology
Received date: 31 January 2016
Revised date: 29 April 2016
Accepted date: 29 June 2016
Cite this article as: Carina Barbosa Pereira, Konrad Heimann, Michael Czaplik,
Vladimir Blazek, Boudewijn Venema and Steffen Leonhardt, Thermoregulation
in premature infants: a mathematical model, Journal of Thermal Biology,
http://dx.doi.org/10.1016/j.jtherbio.2016.06.021
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
 Thermoregulation in premature infants: a
mathematical model
Carina Barbosa Pereira1*, Konrad Heimann2, Michael Czaplik3, Vladimir Blazek1,4, Boudewijn Venema1,
Steffen Leonhardt1

1
Chair for Medical Information Technology, Helmholtz-Institute for Biomedical Engineering, RWTH
Aachen University, Aachen, Germany

2
Department of Neonatology, University Children's Hospital, University Hospital RWTH Aachen,
Germany

3
Department of Anesthesiology, University Hospital RWTH Aachen, Aachen, Germany

4
Czech Institute of Informatics, Robotics and Cybernetics (CIIRC), CTU Prague, Prague, Czech Republic

Corresponding author. Chair for Medical Information Technology, Helmholtz-Institute for Biomedical
Engineering, RWTH, Aachen University, Pauwelsstrasse 20, D-52074 Aachen, Germany, Phone: +49
241 80 – 23202. pereira@hia.rwth-aachen.de

Abstract
PURPOSE:
In 2010, approximately 14.9 million babies (11.1%) were born preterm. Because preterm infants
suffer from an immature thermoregulatory system they have difficulty maintaining their core body
temperature at a constant level. Therefore, it is essential to maintain their temperature at, ideally,
around 37 °C. For this, mathematical models can provide detailed insight into heat transfer processes
and body-environment interactions for clinical applications.
METHODS:
A new multi-node mathematical model of the thermoregulatory system of newborn infants is
presented. It comprises seven compartments, one spherical and six cylindrical, which represent the
head, thorax, abdomen, arms and legs, respectively. The model is customizable, i.e. it meets
individual characteristics of the neonate (e.g. gestational age, postnatal age, weight and length)
which play an important role in heat transfer mechanisms. The model was validated during thermal
neutrality and in a transient thermal environment.
RESULTS:
During thermal neutrality the model accurately predicted skin and core temperatures. The difference
in mean core temperature between measurements and simulations averaged 0.25 ± 0.21 °C and that
of skin temperature averaged 0.36 ± 0.36 °C. During transient thermal conditions, our approach
simulated the thermoregulatory dynamics/responses. Here, for all infants, the mean absolute error
between core temperatures averaged 0.12 ± 0.11 °C and that of skin temperatures hovered around
0.30 °C.
CONCLUSIONS:
The mathematical model appears able to predict core and skin temperatures during thermal
neutrality and in case of a transient thermal conditions.

Keywords: Thermoregulation;
Physiological processes; Premature
infants; Bioheat model; Computer
simulation
1. Introduction
In 2010, approximately 11.1% (uncertainty range 9.1-13.4%) of all live births were preterm (≤ 37
weeks of gestation). This estimate corresponds to 14.9 (range 12.3-18.1) million babies worldwide
(Blencowe et al., 2012). According to Blencowe et al. (2012), 10.4% [95% confidence interval (CI)
10.3-10.5%] were very preterm (28 to < 32 weeks), 5.2% (95% CI 5.1-5.3) extremely preterm (≤ 28
weeks) and the remaining 84.3% (95% CI 84.1-84.5) were moderate or late preterm (32 to < 37
weeks). In contrast to full-term babies, premature and very low birth weight (VLBW) babies (< 1,500
g) have an incompletely developed thermoregulatory system, making them highly vulnerable to
changes in environmental temperature. Therefore, it is important to maintain their body
temperature within a small target range (Abbas et al., 2011; Bissinger and Annibale, 2010; Lemons et
al., 2001; Wrobel et al., 2010). According to clinical standards, the skin temperature of these babies
should be maintained at 35.5-37.5 °C, with a core temperature of around 37 °C (36.8-37.5 ˚C) (Chen
et al., 2012; Heimann et al., 2013; Meyer and Bold, 2007). However, the maintenance of body
temperature is a complex process that involves lipolysis and gluconeogenesis: the more energy used
to keep a constant body temperature, the less energy is available for other processes such as growth,
brain development and lung maturation (Abbas et al., 2011; Bissinger and Annibale, 2010). In
addition, very immature infants are extremely vulnerable to heat loss since they have: a thin
subcutaneous fat layer and reduced insulation capacity; a reduced amount of brown adipose tissue;
high levels of evaporative losses; increased surface area to body weight ratio; and an
underdeveloped autonomic control of the skin vasculature in the first days of life (Cramer et al.,
2005; Elzouki et al., 2011; Knobel et al., 2011).
Therefore, to reduce the mortality rate of premature/VLBW infants and improve clinical
outcome, it is essential to avoid hypo- and hyperthermia by keeping an optimum environmental
temperature (Asakura, 2004; Bissinger and Annibale, 2010; Chen et al., 2012; Elzouki et al., 2011;
Ginalski et al., 2008; Wrobel et al., 2010).

Detailed knowledge on heat transfer processes and body-environment interactions [e.g. related
to room temperature, humidity and airflow in neonatal intensive care units) can be achieved by using
adapted distributed models. These systematic tools can be used to plan and optimize warming
therapies and hyperthermia processes, and also provide more detailed knowledge on
pathophysiology related to heat transfer and influences of external factors. The models allow the
simulation and prediction of clinical situations without involving real patients in experimental trials.
Mathematical modeling of the human thermoregulatory system dates from the 1930s (Burton,
1934). From simple human heat balance equations (Aschoff and Wever, 1958; Burton, 1934), several
two-node models were developed (Azer and Hsu, 1977; Gagge and Fobelets, 1985), followed later by
more sophisticated models composed of several/multi segments (Fiala et al., 1999; Huizenga et al.,
2001; Stolwijk and Hardy, 1966; Stolwijk et al., 1971; Werner and Webb, 1993; Wissler, 1964; Yi et
al., 2004). In contrast to the two-node models, the multi-node models simulate a more detailed
human body (thermoregulatory system) and more accurately predict global and local physiological
responses. Stolwijk et al. (Stolwijk and Hardy, 1966; Stolwijk et al., 1971) developed a pioneering
thermoregulation model (multi-node model) based on a man with a body weight of 74.1 kg and a
body surface area of 1.89 m2: the model is composed of six compartments representing head, trunk,
arms, hands, legs and feet. This novel approach inspired others who further refined, improved and
developed it (Fiala et al., 1999; Huizenga et al., 2001; Wissler, 1964). More recently, Fiala et al. (1999)
proposed a sophisticated multi-layer model consisting of 15 spherical or cylindrical body elements
(head, face, neck, shoulders, arms, hands, thorax, abdomen, legs and feet). Each body element is
composed of several annular concentric layers which represent different tissue types (e.g. brain,
lung, bone, viscera, muscle, fat and skin). Despite their accuracy, these models can only be used to
study heat transfer processes in human adults. However, infants (particularly preterm/VLBW infants)
present with significant physiological and anatomical differences. To date, only a few groups have
explored this field, including: Simbruner (1983a), Bussmann et al. (1998), Breuß et al. (2002), Ying et
al. (2004) and, more recently Fraguela et al. (2015).

We present a new multi-node model of the thermoregulatory system of newborn infants. The
whole body (composed of seven compartments) aims to model the heat-transfer processes occurring
in the tissues and at the body surface. Section 2 describes the mathematical model. For validation,
the measurements of Hammarlund et al. (1983) were used. Furthermore, a small clinical study in
moderate preterm infants was performed. Section 3 describes the experimental protocol. The results
are presented in Section 4 , discussed in Section 5, and Section 6 presents the conclusions and offers
some future perspectives.

2. Mathematical Model
The model proposed is schematically presented in Figure 1; it was inspired by Stolwijk et al.
(1971), Fiala et al. (1999), Bussmann et al. (1998) and Werner and Buse (1988). The model consists of
two main systems: 1) the controlled system (also called the plant or passive system) and 2) the
controller (controlling system or active system).

The first (passive) system models the physical body of a neonate and the heat-transfer processes
occurring in the tissues and at the body surface. It is composed of one spherical and six cylindrical
segments (compartments) which correspond to head, thorax, abdomen, arms and legs, respectively.
These body compartments are composed of annular concentric layers that represent various tissues
such as brain, bone, fat, lung, muscle, skin and viscera. The head consists of four layers: brain, bone,
fat and skin; the thorax and abdomen are composed of five elements (thorax: lung, bone, muscle, fat
and skin; abdomen: viscera, bone, muscle, fat and skin); and the upper and lower limbs are
composed of four tissues (bone, muscle, fat and skin). Figure 2 shows the body segments and the
respective layers. In addition, the model comprises a central blood compartment, which represents
large arteries and veins; it exchanges heat by convection (blood flow) with all other compartments.

The second (active) system aims to describe/predict the regulatory mechanisms. Adults have
several regulatory responses to thermal stress, including vasomotion, sweating, shivering
thermogenesis (involuntary muscle movement/contraction), nonshivering thermogenesis, and
voluntary muscle movement. Neonates have only a few of these responses. Nonshivering
thermogenesis [oxidation of brown adipose tissue (BAT)] is the main, rapid mechanism of heat
production in newborn infants in response to cold stress. This is triggered by the hypothalamus via
the sympathetic nervous system (SNS). Vasomotion is another important regulatory response in
neonates that allows to control peripheral blood flow through vasodilation (widening of blood
vessels) or vasoconstriction (narrowing of blood vessels). Thus, in case of cold stress, peripheral
vasoconstriction may occur to reduce heat loss.
The model was implemented in the modeling and simulation environment Dymola® (Dymola
2015 FD01, Dassault Systèmes AB, Lund, Sweden) which is based on the open Modelica® modeling
language.

2.1 Heat transport within the tissue

Mathematical modeling of the human thermoregulatory system started with efforts to formulate
the heat balance. The widely known bioheat differential equation was formulated by Pennes (1948)
and mathematically describes one-dimensional, steady-state heat transfer mechanisms taking place
in living tissues according to

c  = k ,!!!!"!
ρ!"!# + !!!# 0+ M ρ9: c9: w9:,> (T9:,& − T ),
+ K !!!!!!!"!!!!!!!#
 -  / 
7
 
 *- *
*
(1)
$%& ()*&+% 1)23456)2 8%&9):6(; B:))3 C%&62+

where x represents the tissue type (e.g. brain, bone, fat, lung, muscle, skin or viscera) and bl blood, ρ
is the density (kg m-3), c stands for heat capacitance (J kg-1 K-1), T denotes the tissue temperature (K)
and t is the time (s). In addition, k represents tissue conductivity (W m-1 K-1), r denotes the radius (m),
ω is a geometry factor (dimensionless) [ω=1 corresponds to polar coordinates (compartments:
abdomen, thorax, arms and legs) and ω=2 stands for the spherical coordinates (compartment:
head)]. Finally, M defines the heat production by metabolism (W m-3), K is a countercurrent factor
(dimensionless), wbl,0x is the blood perfusion in thermal neutrality (s-1) and Tbl.a denotes the

temperature of the arterial blood (K). In brief, this formulation proposed by Pennes (1948) considers
a radial and conductive heat flow transport from warmer to colder tissues (e.g. bone to skin) (Figure
1b), as well as the contributions of the convective heat transported through the cardiovascular
system and metabolic heat production. Note that whereas basal metabolism is the major source of
heat in premature newborns, blood circulation is the main process of heat distribution. The
combination of those effects influences and determines heat storage in the tissue layers (Figure 3).
The bio-heat balance equation was applied to the tissue layers of all segments by using the
corresponding material constants (ρ, c, k), basal blood perfusion rate (wbl,0) and basal heat
production (M). Table 1 presents the thermophysical and physiological properties/parameters used
in the passive system of our model.

2.1.1 Heat conduction

This model considers temperature variations in the radial direction only, as given by Eq. (1); the total
amount of heat transferred in the angular and axial direction is insignificant compared with the heat
transported radially. Therefore, the angular and axial heat flows are neglected and not incorporated
in the heat conduction term. The local heat flow density qDD (rE, t) (heat flow per unit area, per unit
time in the direction of decreasing temperature gradient – W m-2) in a homogeneous, isotropic solid
is given by Fourier’s law:
qDD (rE, t) = −k∇T(rE, t), (2)

where ∇T(rE, t) stands for the temperature gradient in K and the negative sign refers to the heat
transfer from higher to lower temperatures. As emphasized, the heat transport was considered one-
dimensional (r-direction - radial). Thus, Fourier’s law can be simplified as given by Eq. (3):

Q̇ = −kA .
3
3*
(3)

Here, Q̇ (W) represents the heat flow rate in the r direction and A (m2) the area of the tissue.
In cylinders, heat conduction is dependent on the (a) internal ri (m) and external re (m) radius of the
segments, (b) its length l (m), and the temperature difference between the inner Ti (K) and outer Te
(K) wall as given by

Q̇ = 2kπl :2(*L ⁄*O ),

 N
(4)
O L

:2(*O ⁄*L )
RSU:
where, represents the thermal resistance.

In spherical structures, the heat transport is dependent only on the internal and external radius of
the segment,

Q̇ = 4kπ L O L O,
( N )* *
*O N*L
(5)
*O N*L
WSU*L *O
and stands for the thermal resistance.

2.1.2 Metabolism

The brain is the major contributor to heat production in preterm neonates. In the first year of life the
metabolic rate of this organ corresponds to more than half of the basal metabolic rate (BMR)
(approximately 0.6) (Holliday, 1986; Okken and Koch, 2012). Other organs (such as heart, kidneys,
liver, lung and muscles) also play an important role in heat production. Therefore, in this model, we
assumed that 23%, 11% and 6% of the BMR are produced in the viscera, lung and muscles,
respectively. This assumption is based on the work of Bussmann et al. (1998), Holliday (1986), and
Simbruner (1983a).
The total BMR (W) is governed by the equation
(1.7 + 0.1 ∙ PA) ∙ W, PA < 10
BMR = Z
d2.7 + 0.01 ∙ (PA − 9)f ∙ W, PA ≥ 10
. (6)

This is dependent on both the postnatal age (PA) in days and on the weight W of the neonate (kg).
The BMR of each individual tissue (BMRx,0) can be extrapolated from the total BMR. For example,
BMR B*&62,> = 0.6 ∙ BMR.
Control of body temperature is achieved via negative feedback; this is a complex system that aims for
an equilibrium between heat loss and gain. The temperature-regulating center is located in the
preoptic area, i.e. the anterior portion of the hypothalamus. In case of thermal stress, the
hypothalamic structures, which trigger the thermoregulatory response, are activated by neuronal
thermal inputs coming from the core and cutaneous thermoreceptors (via afferent pathway). The
afferent signals are further weighted and integrated in the hypothalamus, which responds by, e.g.,
increasing the oxidation of BAT (via efferent pathway). These thermoregulatory processes are
illustrated in Figure 4.
To model the afferent inputs, the approach proposed by Bussmann et al. (1998) was applied. The
afferent signal (also afferent temperature) Taff (K) is governed by the following weight function
ijmm = ∑so∑udps,u ∙ is,u fv, (7)

where, Ts,x (K) is the temperature in a layer x of a determined segment/compartment s, and gs,x is a
weighting factor. A deviation ΔTaff (K) of the afferent temperature Taff (K) from the setpoint (target
value) Tsetpoint (K) can be calculated according to
xijmm = ijmm − isyz{|}~z . (8)

Hence, a ΔTaff < 0 corresponds to a decrease in body and/or ambient temperature(s), whereas ΔTaff >
0 corresponds to an increase of this/these parameter(s).
BAT usually accumulates in regions such as neck, axillae, back, mediastinum, and abdomen (Carter
and Schucany, 2008). As a result, the thermoregulatory heat production by nonshivering
thermogenesis was limited to the core layers of the torso. The response to a cold stress ΔM
(additional heat produced by nonshivering thermogenesis) in the thorax and abdomen is given by
0, xijmm ≥ 0
x€ = ‚ƒ„ ∙ ‡ˆ , xi‘juŽ < xijmm < 0,
‡ˆ‰ŠŠ
†
‹‰ŒŽ
(9)
ƒ„ † , xijmm ≤ xi‘juŽ

where, maxM and ΔTmaxM are parameters of the model. Thus, if the afferent temperature is higher

than the Tsetpoint (ΔTaff ≥ 0) due to an increase of ambient or internal temperatures, there is no heat
production. If ΔTaff is negative and decreases successively, a linear increase in heat production by
nonshivering thermogenesis can be seen as a response to cold temperatures. Saturation (maxM = 1) is
reached at xi‘juŽ = -1. In contrast to the other core layers, heat production in the core of the
thorax and abdomen (Mx) is given by the term
€u = “€”u,> (1 + x€) ∙ 2‡ˆ‰ŠŠ ⁄•>. (10)

This is dependent on the basal value BMRx,0 and on the additional heat produced ΔM (e.g. in the
BAT). The factor 2‡ˆ‰ŠŠ ⁄•> defines the Van’t Hoff Q10 effect with a sensitivity coefficient of 2; this
reflects the dependence of biochemical reaction on temperature. This Van’t Hoff Q10 effect is also
noticeable in other layers (brain, arms and legs). Their metabolic rate is defined according to,
€u = “€”u,> ∙ 2‡ˆ‰ŠŠ ⁄•>. (11)

2.1.3 Blood perfusion

Vasomotion is another basic regulatory response in neonates, responsible for adjustment of blood
flow in the skin (Δwbl,skin). Thus, wblskin (blood perfusion in the skin) is governed by the equation

–—˜™š›œ = –—˜,>™š›œ ∙ d1 + x–—˜,sž}~ f, (12)

and wbl,0skin is the skin’s basal blood perfusion. In thermal neutrality, body tissues are supplied with

blood at basal perfusion rates. However, in non-neutral conditions (cold and hot stress) the skin’s
blood flow varies according to Δwbl,skin, which is given by
ƒ„ £¤ , xijmm ≥ xi‘ju,£¤
⎧
⎪ƒ„ ∙ , xi‘ju,£¤ > xijmm ≥0
‡ˆ‰ŠŠ
£¤ ‡ˆ
x–—˜,sž}~ =
‹‰Œ,¥¦

⎨ ƒ¨©£¤ ∙ , 0 > xijmm > xi‘}~,£¤

‡ˆ‰ŠŠ . (13)
⎪ ‡ˆ‹›œ,¥¦
⎩ ƒ¨©£¤ , xijmm ≤ xi‘}~,£¤

Here, minBF and maxBF stand for the minimal and maximal value for the decrease and increase of the
blood flow, respectively. ΔTmin,BF and ΔTmax,BF are parameters that represent deviations from the
setpoint (negative and positive deviations). On the one hand, if ΔTaff is ≥ 0 and increases successively
in response to heat, a proportional increase in blood perfusion is observed. Saturation (upper
limit/plateau of the effector response) is reached at ΔTmax,BF (0.3 K). This mechanism corresponds to
an effector response to an increase in body temperature. On the other hand, if ΔTaff is ≤ 0 and
diminishes in response to cold, a proportional decrease in blood perfusion is seen. The maximal
response activity (saturation) is reached at ΔTmin,BF (-0.3 K). The previous mechanism corresponds to
an effector response to a decrease in body temperature. During cold stress (ΔTaff < 0),
vasoconstriction occurs, reducing heat loss through the skin. In hot environments (heat stress, ΔTaff >
0), vasodilation occurs and the increased blood perfusion promotes heat loss through the skin.

2.1.4 Geometric and Anatomic model

This model aims to describe the thermoregulation processes in neonates. Since the total body
surface area (BSA) plays an important role in the heat transference processes, it was used as an input
parameter for our mathematical model. Various studies have aimed to formulate a suitable equation
that can accurately estimate BSA for high-risk infants (Ahn, 2010; Boyd, 1935; Du Bois and Du Bois,
1916; Meban, 1983; Mosteller, 1987; Sharkey et al., 2001). In the present model, the approach
proposed by Meban (1983)
“ª« = (6.4954 ∙ (­ ∙ 1000)>.®¯R ∙ °>.±R> )⁄10000, (14)
was used. Here, BSA (m2) stands for body surface area, W corresponds to the neonate’s weight in kg

Body Layer ¸ ¹ º BMRx,0 »¼½,¾¿ À

Compartments

and L (cm) is the crown-heel length. Meban analyzed 79 dead human fetuses (11-42 weeks) weighing
8-4080 g. According to him the expression showed an excellent relationship between the body
surface and the variables body weight and crown-heel length (correlation coefficient reached 0.979).
To calculate the area of each body segment (head Ahead, arm Aarm, leg Aleg, thorax Athorax and abdomen
Aabdomen) the work of Simbruner (1983) was taken into consideration. This is based on both the BSA
and neonate’s weight according to
«²yj³ = (0.248 − 0.0153 ∙ ­) ∙ “ª«, (15)

«j¶‘ = (0.062 + 0.00497 ∙ ­) ∙ “ª«, (16)

«:y· = (0.134 + 0.00165 ∙ ­) ∙ “ª«, (17)

«z²|¶ju = (0.129 + 0.00072 ∙ ­) ∙ “ª«, (18)

«j—³|‘y~ = (0.230 + 0.00128 ∙ ­) ∙ “ª«. (19)

As defined in Eq. (16) to Eq. (19), this method considers that the surface of the arms, legs and trunk
(thorax and abdomen) increases with the total BSA. In contrast, the surface of the head decreases
with increasing body size [Eq. (15)]. Those areas are used to calculate the radii used to compute the
thermal resistances [Eq. (4) and Eq. (5)].

2.2 Heat exchange with the Environment

At the body surface, heat is exchanged by convection, evaporation, radiation, and conduction (e.g.
with a mattress). Therefore, the bioheat equation of the skin includes four additional terms to
represent those processes. A detailed description of each heat transport mechanism is presented
and the convective and radiative heat transfer coefficients are given in
 W m-1 K-1 kg m-3 J kg-1 K-1 m3 s-1 m-3
Brain 0.49 1080 3850 0.60 3.3 x 10-3 1
Bone 0.40 1500 1591 0.00 0 1
Head
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9
Lung 0.28 550 3718 0.11 7.0 x 10-3 1
Bone 0.40 1357 1700 0.00 0 1
Thorax Muscle 0.42 1085 3768 0.00 4.0 x 10-3 1
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9
Viscera 0.53 1000 3697 0.23 7.0 x 10-3 1
Bone 0.40 1357 1700 0.00 0 1
-3
Abdomen Muscle 0.42 1085 3768 0.00 4.0 x 10 1
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9
Bone 0.40 1357 1700 0.00 0 1
Muscle 0.42 1085 3768 0.03 1.25 x 10-3 1
Arms
Fat 0.16 850 2300 0.00 0 1
-3
Skin 0.47 1085 3680 0.00 1.25 x 10 0.9
Bone 0.40 1357 1700 0.00 0 1
Muscle 0.42 1085 3768 0.03 1.25 x 10-3 1
Legs
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9

Á is the thermal conductivity,  defines the tissue density, à stands for the specific heat, “€”u,> corresponds
Blood 1069 3650

to the basal metabolism, –—˜,>Œ is basal blood perfusion rate or blood perfusion in thermal neutrality, and Ä is
the countercurrent factor.

Table 2.
Convection
The convective heat exchange qc between the skin of the neonate (with temperature Tskin) and the
incubator air (with temperature Tair) was modeled according to
ÅÆ = ℎÆ ∙ «sž}~,j ∙ (isž}~ − ij}¶ ). (20)

Here, hc (W m-2 K-1) and Askin,a denote the convection coefficient and area of the skin of a determined
segment in contact with the air.

2.2.1 Evaporation

Evaporation of water is one of the main modes of heat loss. Insensible water loss is a sum of two
elements: evaporation from the skin qev and water loss from the respiratory system. The former can
be described by the following equation
ÅyÈ = ÉyÈ ∙ «sž}~,j ∙ iÊ­°, (21)

where, eev is the specific enthalpy of the vaporization of water (0.65 W h g-1) and TEWL (g m-2 h-1)
represents the transepidermal water loss. Hammarlund et al. determined TEWL in newborns under
controlled environmental conditions. They found an exponential relationship between TEWL and
gestational age (GA) at different PAs (from birth to 4 weeks after birth) (Figure 5) (Hammarlund et
al., 1983).

According to Ultman (1987), this mathematical relationship can be described by

iÊ­° = Ë ∙ ,Ìs∗ − •>> ∙ Ìj∗ 0,

¶Î
(22)

where, d represents a diffusion coefficient (g h-1 m-2 kPa-1), rH is the relative humidity, Ìs∗ and Ìj∗
stand for the partial pressure of water vapor at body surface temperature and at air temperature,
respectively. The diffusion coefficient is dependent on the skin temperature of the newborn as given
by
-ÐÑÐ
Ï.••ÏN
Ë = Ë> ∙ É Ò™š›œ
. (23)

The coefficient d0 is governed by the following equation

d> = 1.5 ∙ ,1 + 0,
R∙³Ô N±
RÕÖ×
(24)

where d1 (Bussmann et al., 1998; Ultman, 1987) is given by

1.3, Ù« > 35
˕ = Ø
24000 ∙ É N>.Rڕ∙Û× , Ù« ≤ 35
. (25)

The respiratory water loss (RWL) can be computed according to the equation proposed by Ultman
(1987)

”­° = − ,0.411 + 0.000968 ∙ (ij}¶ − 273.15) − 0∙ ­.

Ü.W∙¶Î∙{‰∗
•>•±±N¶Î∙{‰∗
(26)

This is dependent on the air temperature Tair, relative humidity rH, partial pressure of water vapor at
air temperature Ìj∗ and on the weight of the neonates W.
2.2.2 Radiation

Ŷ = ℎ¶ ∙ «sž}~,j ∙ (isž}~ − i†Ýˆ ),

Radiation corresponds to the heat transfer by long-wave radiation qr as defined by
(27)

where, TMRT (K) is mean radiant temperature (MRT) and hr (W m-2 K-1) stands for the radiative heat-
exchange coefficient.
2.2.3 Conduction

The conductive heat exchange between the neonate and the mattress (qcond) also needs to be
considered. This can be described by
ÅÆ|~³ = Á‘ ∙ «sž}~,‘ ∙ (isž}~ − i‘ ), (28)

where, km (W m-2 K) stands for the heat transfer coefficient, Tm is the temperature of the mattress,
and Askin,m denotes the area of the skin of a determined segment in contact with the mattress.
2.3 Validation of the mathematical model
To validate the mathematical model, the work of Hammarlund et al. (1983) was used. In their study,
measurements in 19 newborns [10 males and 9 females, GA 39.58 ± 0.61 weeks (mean ± SD), weight
3.56 ± 0.51 kg, length 0.51 ± 0.02 cm] were performed directly after birth. Their aim was to examine
the relationship between ambient humidity and insensible water loss. The infants were placed in an
incubator and the relative humidity was varied from 20-65%, with increments of 5%. Moreover, the
ambient temperature (incubator temperature) was maintained as constant as possible. During the
measurements (lasting 2.4 ± 0.8 h) rectal and skin temperature at the trunk were determined. To
validate our mathematical model, we compared the temperatures experimentally obtained by
Hammarlund et al. (1983) with those estimated in our simulations.
In addition, an experimental feasibility study was conducted in the department of Neonatology of the
University Hospital RWTH Aachen. The aim was to analyze the performance of our model during
transient thermal conditions using the following experimental protocol.

3. Experimental Protocol
To validate the mathematical model during transient thermal conditions, a study was performed at the RWTH Aachen Unive
Aachen (EK032/09) and parental consent was obtained. Both infants had a stable condition;

Table 3 presents their characteristics.

The experiments had three phases. In the first phase, core and skin temperatures of the newborns
were measured while they were inside the incubator (circa 1 h); skin temperature measurements
were performed every 15 min. Core temperature was measured at time points 0 and 60 min. Then,
infants were removed from the incubator for Kangaroo care (also known as skin-to-skin care) for
about 1 h and skin temperature measurements were performed every 20 min. Rectal temperature
was assessed at time point 120 min. In the final phase, the infants were replaced in the incubator.
Skin and core temperature measurements were again performed (every 15 min for 1 h, and at time
point 180 min, respectively). Figure 6 presents all measurement time points. The core/rectal
temperature was measured with a standard clinical thermometer. Skin temperature was measured
using a long wave infrared camera VarioCAM® HD head 820S/30 mm (InfraTec GmbH, Dresden,
Germany). Room temperature and relative humidity were acquired with the HOBO® U12-012 data
logger (Onset Computer Corporation, Bourne, Massachusetts, USA). As described in section 4, our
model is also dependent on the temperature and relative humidity of the incubator. Standardly,
incubators present sensors that automatically assess these properties without the need for
additional measurement devices.

4. Results
The work of Hammarlund et al. (1983) was used to validate our model during stable environmental
conditions. Table 4 compares the empirical data (rectal/core temperature and skin temperature) for
all 19 subjects obtained by the latter group with the model simulations. The average difference in
mean core temperature between Hammarlund et al. (1983) and the mathematical model was 0.25 ±
0.21°C, and that of skin temperature was 0.36 ± 0.36 °C.
Furthermore, to verify the outcome of the model during transient thermal conditions, a study was
performed with two preterm babies as previously mentioned. Figure 7 presents a comparison
between core temperatures for model predictions (dashed and solid lines) and experimental
measurements (circles and squares): the dashed green line/circles represent infant S1, the solid blue
line/squares represent the newborn preterm S2. For infant S1, the mean absolute error between
core temperatures [rectal temperature (measurements) and temperature of the internal layer of the
abdomen (viscera) (simulation)] was 0.04 ± 0.04 °C and that for infant S2 was 0.19 ± 0.19 °C. Figure 8
compares the skin temperatures (measured using a thermal camera and simulated using the current
mathematical model) for each body segment (head, thorax, abdomen, arms and legs) and the
infants: green lines are the results for infant S1 and blue lines for infant S2. The temperatures
measured with the thermal camera are illustrated with markers (circles for S1, squares for S2). Table
5 complements Figure 8 in presenting the mean absolute errors between the measurements and
simulations. As shown, similar results were obtained for both infants and compartments (hovering
around 0.30 °C). The only exception was the skin temperature of the legs: the mean absolute error
was 0.18 ± 0.14 °C and 0.67 ± 0.43 °C for infants S1 and S2, respectively.

5. Discussion
Perinatal and neonatal research remains controversial with regard to ethical considerations, as
children are extremely vulnerable research subjects (Behrman et al., 2007; Laventhal et al., 2012).
Therefore, pediatric research has to consider ethical challenges, including: balancing risks and
benefits, parental informed consent and assent, and clinical equipoise (Laventhal et al., 2012). Few
studies have examined heat transfer processes in infants and preterm neonates and studies on the
influence of thermal transients are often lacking due to ethical reasons. Nevertheless, such studies
are needed since thermoregulation in preterm neonates is not as developed as that in full-term
newborns.
Currently available thermoregulatory models of adults are related to a ‘standard’ person, with
predefined body characteristics including, e.g., body weight, fat-to-body mass ratio, and skin surface
area. In contrast, our mathematical approach is customizable, since it incorporates individual
characteristic of neonates such as GA, PA, weight and length. These features are essential because
evaporation losses are highly dependent on the area of the skin, and on the GA and PA (Figure 5). In
addition, the remaining heat processes (in tissues and at the body surface) are influenced by the area
of the tissue layers which can be derived from the BSA and, hence, from the body weight and length
of the newborn.
To validate our model during thermal neutrality, the work of Hammarlund et al. (1983) was used.
This group performed measurements in a relatively homogeneous group of newborns, with a GA
averaging 39.58 ± 0.61 weeks (section 2.3). Table 4 compares these latter empirical results with
those obtained in our simulations. Similar environmental conditions were reproduced, i.e. the same
ambient temperature and relative humidity, including its variations (from 20-65%, with increments of
5%). The two parameters measured, i.e. rectal temperature (core temperature) and skin temperature
of the trunk, were compared with the calculated core temperature of the abdomen (layer viscera)
and its skin temperature. Table 4 shows that similar results were obtained with the real
measurements and simulations. The mean core temperature difference averaged 0.25 ± 0.21 °C and
the mean skin temperature difference averaged 0.36 ± 0.36 °C. The standard deviation (SD) of the
results is (in nearly all cases) identical in magnitude to the absolute magnitude of the value. As
mentioned, the study population had a similar weight, length, GA and PA [PA 1 ± 0.00 days (mean ±
SD), GA 39.58 ± 0.61 weeks, weight 3.56 ± 0.51 kg, length 0.51 ± 0.02 cm]. The environmental
conditions (ambient temperature and relative humidity) were also similar. In the measurements

performed by Hammarlund et al. (1983), the SDs were also similar to ours ( SD Tcore = 0.13 and

SDTskin = 0.19 ); however, small fluctuations were visible, possibly due to varying environmental
temperatures. In contrast to our simulations in which the environmental temperature was kept
constant, small temperature variations were found during the measurements. Our model
demonstrated its ability to correctly estimate core temperature and skin temperature in thermal
neutrality in full-term newborns. Unfortunately, we were unable to find studies similar to that of
Hammarlund et al. (1983) dealing with preterm infants; therefore, our validation was performed on
full-term infants with regard to body temperature in thermal neutrality.
In addition to thermal neutrality, it is important to verify the outcome of our model in case of
transient thermal conditions, e.g. a drop in environmental temperature. For this, a small pilot study
was performed at the department of Neonatology of the University Hospital RWTH Aachen with two
preterm newborns. Figure 7 shows the variation in rectal temperature and the core temperature
dynamics for both infants estimated with our approach. As expected, there was a decrease in
temperature followed by an increase; also, a high level of agreement was achieved between the
measurements and simulations. The absolute error between core temperatures averaged 0.04 ± 0.04
°C and 0.19 ± 0.19 °C for infant 1 and infant 2, respectively. Figure 8 presents the measured and
estimated skin temperatures. Once more a small decrease, followed by an increase in temperature
was perceived. For all body compartments there was excellent agreement between the measured
and the simulated temperatures. This is further corroborated by the results presented in Table 5. The
mean absolute error of almost all compartments hovered around 0.30 °C. The only exception was the
skin temperature of the legs, with a mean absolute error of 0.18 ± 0.14 °C and 0. 67 ± 0.43 °C for
infant S1 and S2, respectively. In this context, it should be noted that during Kangarooing care the
skin-to-skin contact between mother and infant was taken into account. Equally important is the fact
that, in both validations, only healthy infants were included.
Other research groups also studied heat exchange processes and energy balance in newborns.
Bussmann et al. (1998) and Breuß et al. (2002) presented very similar approaches. The former group
developed a compartment model to describe the thermoregulatory system, and the latter
introduced a finite volume approach. Both models are composed of six compartments: head, trunk,
arms, and legs. The head consists of 4 layers (core, bone, fat and skin) and the other compartments
have only 3 layers (core, fat and skin). As far as we know, the above-mentioned authors did not
compare their simulations with clinical data. Ying et al. (2004) proposed a simple model, composed
of 3 body segments (head, trunk and legs) to study the influence of clothing on the thermoregulation
processes. To validate their approach, the experimental study of Bolton et al. (1996) was taken into
account. The results showed a good agreement between skin temperatures for trunk and legs. In
2015, Fraguela et al. presented an algorithm to control the temperature of the incubator based on
the infant’s temperature. The heat exchange and the energy balance in the newborn was described
with a single compartment model consisting of two layers, core (representing tissues, bones and
internal organs) and surface (representing fat and skin). However, the authors did not compare their
simulations with clinical data (Fraguela et al., 2015). In contrast to the previous approaches, our
model includes more compartments and layers, allowing to simulate a more detailed human body
(section 2). Therefore, more accurate global and local physiological responses can be predicted. In
addition, the model was successfully validated during both thermal neutrality and transient thermal
conditions.

6. Conclusion
As premature babies suffer from an incompletely developed thermoregulatory system, maintenance
of an optimal thermal environment (ideally between 36.8 and 37.5 °C) is indispensable for their
survival and growth. Mathematical models can help to provide detailed knowledge on heat transfer
processes and body-environment interactions.
We have proposed a multi-node model of the thermoregulatory system of newborn infants. A first
successful validation of our approach in thermal neutrality was made by comparing our simulations
with the empirical results obtained by Hammarlund et al. (1983). Their study was performed in a
homogenous group of newborn infants, with a GA of approximately 39 weeks (full-term infants),
directly after birth. Since this approach can also be extended to premature infants (moderate
preterm, very preterm and extremely preterm infants), we aim to validate our model with data of
other experimental groups.
Besides thermal neutrality, thermal transient thermal conditions were also taken into consideration.
Our small pilot study showed a good agreement between model and measured data, as well as the
model’s ability to simulate thermoregulatory dynamics/responses. Because only two infants were
involved, future studies should include more infants.
In conclusion, it appears that our mathematical approach is able to predict body temperatures (core
and skin temperature) during thermal neutrality and even in a transient thermal conditions. It can be
a useful tool for physicians in the aim to achieve a better understanding of the heat transfer
processes and body-environment interactions that occur in preterm infants.

Acknowledgments
C. B. Pereira wishes to acknowledge FCT (Foundation for Science and Technology in Portugal) for her
PhD grant SFRH / BD / 84357 / 2012.

Referencess
Abbas, A.K., Heimann, K., Jergus, K., Orlikowsky, T., Leonhardt, S., 2011. Neonatal non-contact respiratory
monitoring based on real-time infrared thermography. Biomed. Eng. Online 10, 93.
Ahn, Y., 2010. Formula and scale for body surface area estimation in high-risk infants. Coll. Antropol. 34, 1273–
1280.
Asakura, H., 2004. Fetal and neonatal thermoregulation. J. Nippon Med. Sch. 71, 360–370.
Aschoff, J., Wever, R., 1958. Kern und Schale im Wärmehaushalt des Menschen. Naturwissenschaften 45, 477–
485.
Azer, N., Hsu, S., 1977. The prediction of thermal sensation from a simple model of human physiological
regulatory response. ASHRAE Trans 83, 88–102.
Behrman, R.E., Butler, A.S., Outcomes, I. of M. (US) C. on U.P.B. and A.H., 2007. A Review of Ethical Issues
Involved in Premature Birth.
Bissinger, R.L., Annibale, D.J., 2010. Thermoregulation in very low-birth-weight infants during the golden hour:
results and implications. Adv. neonatal care Off. J. Natl. Assoc. Neonatal Nurses 10, 230–238.
Blencowe, H., Cousens, S., Oestergaard, M.Z., Chou, D., Moller, A.-B., Narwal, R., Adler, A., Vera Garcia, C.,
Rohde, S., Say, L., Lawn, J.E., 2012. National, regional, and worldwide estimates of preterm birth rates in
the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications.
Lancet 379, 2162–72.
Bolton, D.P., Nelson, E. a, Taylor, B.J., Weatherall, I.L., 1996. Thermal balance in infants. J. Appl. Physiol. 80,
2234–2242.
Boyd, E., 1935. The Growth of the Surface Area of the Human Body. University of Minnesota Press.
Breuß, M., Fischer, B., Meister, A., 2002. The numerical simulation of unsteady heat conduction in a premature
infant. Int. J. Numer. Methods Fluids 40, 253–261.
Burton, A.C., 1934. The Application of the Theory of Heat Flow to the Study of Energy Metabolism Five Figures.
J. Nutr. 7, 497–533.
Bussmann, O., Nahm, W., Konecny, E., 1998. A model for simulating heat transfer and thermoregulation of
premature infants. Biomed. Tech. (Berl). 43 Suppl, 300–301.
Carter, B.W., Schucany, W.G., 2008. Brown adipose tissue in a newborn. Proc. (Bayl. Univ. Med. Cent). 21, 328–
330.
Chen, W., Oetomo, S.B., Feijs, L. (Eds.), 2012. Neonatal Monitoring Technologies: Design for Integrated
Solutions. IGI Global.
Cramer, K., Wiebe, N., Hartling, L., Crumley, E., Vohra, S., 2005. Heat loss prevention: a systematic review of
occlusive skin wrap for premature neonates. J. Perinatol. Off. J. Calif. Perinat. Assoc. 25, 763–769.
Du Bois, D., Du Bois, E.F., 1916. A formula to estimate the approximate surface area if height and weight be
known. Arch Intern Med 17, 863–871.
Elzouki, A.Y., Stapleton, F.B., Harfi, H.A., Oh, W., Whitley, R.J., Nazer, H., 2011. Textbook of Clinical Pediatrics.
Fiala, D., Lomas, K.J., Stohrer, M., 1999. A computer model of human thermoregulation for a wide range of
environmental conditions: the passive system. J. Appl. Physiol. (Bethesda, Md. 1985) 87, 1957–1972.
Fraguela, A., Matlalcuatzi, F.D., Ramos, Á.M., 2015. Mathematical modelling of thermoregulation processes for
premature infants in closed convectively heated incubators. Comput. Biol. Med. 57, 159–172.
Gagge, A.R., Fobelets, A.P., 1985. A standard predictive index of human response to the thermal environment.
ASHRAE Trans 92, 709–731.
Ginalski, M.K., Nowak, A.J., Wrobel, L.C., 2008. Modelling of heat and mass transfer processes in neonatology.
Biomed. Mater. 3, 034113.
Hammarlund, K., Sedin, G., Strömberg, B., 1983. Transepidermal Water Loss in Newborn Infants. Acta
Pædiatrica 72, 721–728.
Heimann, K., Jergus, K., Abbas, A.K., Heussen, N., Leonhardt, S., Orlikowsky, T., 2013. Infrared thermography for
detailed registration of thermoregulation in premature infants. J. Perinat. Med. 41, 613–620.
Holliday, M.A., 1986. Body Composition and Energy Needs during Growth, in: Falkner, F., Tanner, J.M. (Eds.),
 Postnatal Growth Neurobiology. Springer US, pp. 101–117.
Huizenga, C., Hui, Z., Arens, E., 2001. A model of human physiology and comfort for assessing complex thermal
environments. Build. Environ. 36, 691–699.
Knobel, R.B., Guenther, B.D., Rice, H.E., 2011. Thermoregulation and thermography in neonatal physiology and
disease. Biol. Res. Nurs. 13, 274–282.
Laventhal, N., Tarini, B.A., Lantos, J., 2012. Ethical issues in neonatal and pediatric clinical trials. Pediatr. Clin.
North Am. 59, 1205–20. doi:10.1016/j.pcl.2012.07.007
Lemons, J.A., Bauer, C.R., Oh, W., Korones, S.B., Papile, L.-A., Stoll, B.J., Verter, J., Temprosa, M., Wright, L.L.,
Ehrenkranz, R.A., Fanaroff, A.A., Stark, A., Carlo, W., Tyson, J.E., Donovan, E.F., Shankaran, S., Stevenson,
D.K., Network, for the N.N.R., 2001. Very Low Birth Weight Outcomes of the National Institute of Child
Health and Human Development Neonatal Research Network, January 1995 Through December 1996.
Pediatrics 107, e1–e1.
Meban, C., 1983. The surface area and volume of the human fetus. J. Anat. 137 (Pt 2), 271–278.
Meyer, M.P., Bold, G.T., 2007. Admission temperatures following radiant warmer or incubator transport for
preterm infants <28 weeks: a randomised study. Arch. Dis. Child. Fetal Neonatal Ed. 92, F295–7.
Mosteller, R.D., 1987. Simplified calculation of body-surface area. N. Engl. J. Med. 317, 1098.
Okken, A., Koch, J., 2012. Thermoregulation of Sick and Low Birth Weight Neonates: Temperature Control.
Temperature Monitoring. Thermal Environment. Springer Science & Business Media.
Pennes, H.H., 1948. Analysis of Tissue and Arterial Blood Temperatures in the Resting Human Forearm. J. Appl.
Physiol. 1, 93–122.
Sharkey, I., Boddy, A. V, Wallace, H., Mycroft, J., Hollis, R., Picton, S., 2001. Body surface area estimation in
children using weight alone: application in paediatric oncology. Br. J. Cancer 85, 23–28.
Simbruner, G., 1983. Thermodynamic models for diagnostic purposes in the newborn and fetus. Facultas-
Verlag, Wien.
Stolwijk, J. a. J., Hardy, J.D., 1966. Temperature regulation in man — A theoretical study. Pflüger’s Arch. für die
gesamte Physiol. des Menschen und der Tiere 291, 129–162.
Stolwijk, J.A.J., Laboratory, J.B.P.F., Health, Y.U.D. of E. and P., Administration, U.S.N.A. and S., 1971. A
mathematical model of physiological temperature regulation in man. National Aeronautics and Space
Administration, Washington DC.
Ultman, J.S., 1987. Computational Model for Insensible Water Loss From the Newborn. Pediatrics 79, 760–765.
Werner, J., Buse, M., 1988. Temperature profiles with respect to inhomogeneity and geometry of the human
body. J. Appl. Physiol. 65, 1110–1118.
Werner, J., Webb, P., 1993. A six-cylinder model of human thermoregulation for general use on personal
computers. Ann. Physiol. Anthropol. 12, 123–134.
Wissler, E.H., 1964. A mathematical model of the human thermal system. Bull. Math. Biophys. 26, 147–166.
Wrobel, L.C., Ginalski, M.K., Nowak, A.J., Ingham, D.B., Fic, A.M., 2010. An overview of recent applications of
computational modelling in neonatology. Philos. Trans. A. Math. Phys. Eng. Sci. 368, 2817–2834.
Yi, L., Fengzhi, L., Yingxi, L., Zhongxuan, L., 2004. An integrated model for simulating interactive thermal
processes in human–clothing system. J. Therm. Biol., International Thermal Physiology Symposium:
 Physiology and Pharmacology of Temperature Regulation 29, 567–575.
Ying, B.A., Kwok, Y.L., Li, Y., Yeung, C.Y., Li, F.Z., Li, S., 2004. Mathematical modeling of thermal physiological
responses of clothed infants. J. Therm. Biol. 29, 559–565.

Biographies

Carina Pereira was born in Braga, Portugal, on November 6, 1988. She

received the M.Sc. degree in Biomedical Engineering from the University of Minho, Braga, Portugal,
in 2011. Currently, she is working toward the doctor degree at the Philips Chair for Medical
Information Technology, RWTH Aachen University, Aachen, Germany. Her research interests include
infrared thermography, imaging and biosignal processing as well as feedback control in medicine.

Konrad Heimann was born in Cologne, Germany, in 1969. He received

the M.D. degree at the university of Cologne in 1996. After working for one and a half years at the
Department of Pediatrics of the University Hospital of Cologne, he was educated in pediatrics at the
University Hospital of the RWTH of Aachen. Since 2006, he is a senior physician at the Department of
Neonatology and Pediatric Intensive Care. His main fields of research are non-contact monitoring of
vital parameters and thermoregulation in full and preterm infants.

Associate professor Dr. Michael Czaplik is working at the department of

anesthesiology, university hospital RWTH Aachen as senior consultant in anesthesiology. As leader of
the interdisciplinary section “medical technology” he is in charge of diverse research projects
combining medical and technical knowledge and expertise. His main field of research includes
innovative monitoring modalities and telemedicine in anesthesiology, intensive care medicine and
emergency medicine.

Vladimir Blazek was born in Czechoslovakia, in 1945. He received the

Dipl.-Ing. degree in electrical engineering from the Technical University, Brno, Czech Republic, in
1969, the Dr.-Ing. degree from RWTH Aachen University, Aachen, Germany, in 1979, and the Venia
Legendi degree from the Czech Technical University, Prague, in 1993. In 1971, he was with the
Institute of High Frequency Technology, RWTH Aachen University. Since 2011, he has been with the
Philips Chair of Medical Information Technology, RWTH Aachen University. His current research
interests include optoelectronics in medicine, biomedical sensors, functional optical imaging
techniques, and tissue optics.
 Boudewijn Venema was born in Amsterdam, the Netherlands, in 1983.
He holds the Dr.-Ing. Degree in Electrical Engineering from RWTH Aachen University, Aachen,
Germany. Currently, he is with the Philips Chair of Medical Information Technology, RWTH Aachen
University. His research interests include optoelectronics in medicine, the study of photon-tissue
interaction with Monte-Carlo and contactless long-term measurement of vital signs using
photoplethysmographic and infrared sensor technologies.

Steffen Leonhardt was born in Frankfurt, Germany, in 1961. He received

the M.S. degree in computer engineering from the State University of New York at Buffalo, Buffalo,
NY, USA, the Dipl.-Ing. degree in electrical engineering and the Dr.-Ing. degree in control engineering
from the Technische Universität Darmstadt, Darmstadt, Germany, and the M.D. degree in medicine
from Johann Wolfgang Goethe University, Frankfurt, Germany. After 5 years of industrial work
experience as a R&D manager with Dräger Medical GmbH, Lübeck, in 2013 he was appointed Full
Professor and Head of the Philips Endowed Chair of Medical Information Technology at RWTH
Aachen University, Aachen, Germany.

Glossary
Acronyms
BAT Brown adipose tissue

CI Confidence interval

SD Standard deviation

SNS Sympathetic nervous system

VLBW Very low birth weight

Latin Symbols

 Area (m2)

!" Basal metabolic rate (W)

# Body Surface Area (m2)

$ Heat capacitance (J kg-1 K-1)

% Diffusion coefficient (g h-1 m-2 kPa-1)

& Specific enthalpy of vaporization of water (W h g-1)

' Weighting factor (dimensionless)

( Gestational age (weeks)

ℎ* Convective heat transfer coefficient (W m-2 K-1)

ℎ+ Radiative heat transfer coefficient (W m-2 K-1)

, Thermal conductivity (W m-1 K-1)

- Length (m)

. Crown-heel length (cm)

/ Countercurrent factor (dimensionless)

! Heat production by metabolism (W m-3)

0∗ Partial pressure of water vapor (kPa)

2 Postnatal age (days)

3 44 Heat flow density (W m-2)

5̇ Heat flowrate (W)

7 Radius (m)

78 Relative humidity (%)

"9. Respiratory water loss (g kg h-1)

: Temperature (K)

; Time (s)

:<9. Transepidermal water loss (g m-2 h-1)

= Volume (m3)
> Perfusion rate (s-1)

9 Weight (kg)

Non-Latin Symbols

? Difference

@ Enthalpy of vaporization (W h g-1)

A Density (kg m-3)

B Geometry factor used in the bioheat differential equation (dimensionless)

C Gradient

Subscripts

0 Thermal neutrality

EFF Afferent

G- Blood

G-, E Arterial Blood

$ Convection

$IJ% Conduction

& External

&K Evaporation

L Internal

M Mattress

!": Mean radiant temperature

7 Radiation

N Segment (head, thorax, abdomen, arm, leg)

N&;0ILJ; Setpoint/target value

O Tissue/Layer (brain, bone, fat, lung, muscle, skin or viscera)

Figure 1 – Schematic representation of the plant. (a) Diagram depicting the seven segments (head,
thorax, abdomen, arms and legs) that compose the model. The head is defined as a sphere and the
other segments are modeled as cylinders. (b) Transversal section of the abdomen (section A-A’).
Coordinates r, θ, and z denote the radial, angular, and axial directions, respectively. This model only
considers radial heat transport.

Figure 2 – Illustration of the body segments and respective layers. The head is composed of four
layers, skin, fat, bone and brain; the thorax and abdomen of five elements (thorax: lung, bone,
muscle, fat and skin; abdomen: viscera, bone, muscle, fat and skin); and the legs and arms are
composed of four tissues (bone, muscle, fat and skin).

Figure 3 – Representation of the (1) radial conductive heat flow transport in the abdomen between
adjacent tissue layers (viscera, bone, muscle, fat and skin) (2) convective heat transported through
the cardiovascular system, and (3) heat exchange with the environment (convection, radiation,
evaporation and conduction).
Figure 4 – Graphical representation of the thermoregulatory processes in infants. SNS=sympathetic
nervous system.
Figure 5 – Relationship between transepidermal water loss and gestational age at different postnatal
ages (Hammarlund et al., 1983).
Figure 6 – Chronogram representing all measurement time points. Core/rectal temperature was
measured with a standard clinical thermometer and skin temperature with a thermal camera.

Figure 7 – Comparison between core temperatures for model simulations (solid and dashed lines)
and clinical measurements (circles and squares). The dashed green line/circles represent infant S1,
and the solid blue line/squares represent infant S2. Background: light gray segments indicate the
periods the newborn infants were in the incubator (‘I’ = incubator); dark gray segments indicate the
periods the preterm babies were removed from the incubator for Kangaroo care (K = Kangaroo care).
Figure 8 – Comparison between skin temperatures for the model simulations (solid/dashed lines) and
clinical measurements (circles/squares). The dashed green line/circles represent infant S1 and the
solid blue line/squares represent infant S2. Background: light gray segments represent the periods
the newborn infants were in the incubator (‘I’ = incubator); dark gray segment represents the period
the preterm babies were removed from the incubator for Kangaroo care (‘K’ = Kangaroo care). A
comparison was made between all body compartments (head, thorax, abdomen, arms and legs).
Table 1 – Thermophysical and physiological parameters used in the passive system of the proposed
mathematical model according to Bussmann et al., (1998) and Fiala et al. (1999)

Body U V W XYZ,[\
Layer BMRx,0 ]
Compartments W m-1 K-1 kg m-3 J kg-1 K-1 m3 s-1 m-3
Brain 0.49 1080 3850 0.60 3.3 x 10-3 1
Bone 0.40 1500 1591 0.00 0 1
Head
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9
Lung 0.28 550 3718 0.11 7.0 x 10-3 1
Bone 0.40 1357 1700 0.00 0 1
Thorax Muscle 0.42 1085 3768 0.00 4.0 x 10-3 1
Fat 0.16 850 2300 0.00 0 1
-3
Skin 0.47 1085 3680 0.00 1.25 x 10 0.9
Viscera 0.53 1000 3697 0.23 7.0 x 10-3 1
Bone 0.40 1357 1700 0.00 0 1
Abdomen Muscle 0.42 1085 3768 0.00 4.0 x 10-3 1
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9
Bone 0.40 1357 1700 0.00 0 1
Muscle 0.42 1085 3768 0.03 1.25 x 10-3 1
Arms
Fat 0.16 850 2300 0.00 0 1
Skin 0.47 1085 3680 0.00 1.25 x 10-3 0.9
Bone 0.40 1357 1700 0.00 0 1
Muscle 0.42 1085 3768 0.03 1.25 x 10-3 1
Legs
Fat 0.16 850 2300 0.00 0 1
-3
Skin 0.47 1085 3680 0.00 1.25 x 10 0.9
Blood 1069 3650
, is the thermal conductivity, A defines the tissue density, $ stands for the specific heat, !"P,Q corresponds
to the basal metabolism, >RS,QT is basal blood perfusion rate or blood perfusion in thermal neutrality, and / is
the countercurrent factor.

Table 2 – Convective hc and radiative hr heat transfer coefficients used in the mathematical model
(Bussmann et al., 1998)
Birth Study Study
Birth weight GA PA Feeding
Sub. Gender height weight height
(Kg) (weeks) (days) method
(cm) (Kg) (cm)
S1 M 2.130 44 1.930 44 33 3 Orogastric
S2 M 2.160 46 2.000 46 33 6 Orogastric
Mean ± SD 2.145 ± 0.02 45 ± 1.41 1.965 ± 0.05 45 ± 1.41 33 4.5 ± 2.12
Table 3 – Patient data (Sub. – subjects, SD – standard deviation)

Body hc hR
Layer
Compartments W m-2 K-1 W m-2 K-1
Head Skin 4 5
Thorax Skin 3 4
Abdomen Skin 3 4
Arms Skin 5 4
Legs Skin 5 4

Table 4 – Comparison between the empirical results obtained by Hammarlund et al. (1983) and the
results achieved with our model
Mean
Hammarlund et al. (1983) Mathematical model temperature
difference
GA W L Tamb ± SD Tcore ± SD Tskin ± SD Tamb Tcore ± SD Tskin ± SD Tcore Tskin
I G
weeks kg mm °C °C °C °C °C °C °C °C
I1 M 39 3.970 0.520 34.2 ± 1.5 36.7 ± 0.4 35.4 ± 0.3 34.2 36.8 ± 0.1 35.4 ± 0.1 0.1 0.0
I2 M 38 3.100 0.490 34.6 ± 0.4 36.9 ± 0.1 36.0 ± 0.1 34.6 36.8 ± 0.1 35.7 ± 0.1 0.1 0.3
I3 F 40 2.830 0.495 34.8 ± 0.2 36.7 ± 0.1 35.7 ± 0.2 34.8 36.8 ± 0.1 35.8 ± 0.1 0.1 0.1
I4 F 40 4.010 0.515 34.5 ± 0.3 36.6 ± 0.1 35.5 ± 0.1 34.5 36.9± 0.1 35.8 ± 0.1 0.3 0.3
I5 M 40 4.430 0.555 34.1 ± 0.2 36.6 ± 0.1 34.5 ± 0.4 34.1 36.9 ± 0.1 35.7 ± 0.1 0.3 1.2
I6 M 40 3.610 0.505 34.2 ± 0.4 36.8 ± 0.1 34.6 ± 0.4 34.2 36.8 ± 0.1 35.6 ± 0.1 0.0 1.0
I7 F 40 3.945 0.520 32.6 ± 0.4 36.1 ± 0.0 34.6 ± 0.3 32.6 36.6 ± 0.1 35.2 ± 0.1 0.5 0.6
I8 F 39 3.570 0.500 33.2 ± 0.3 36.4 ± 0.2 35.2 ± 0.1 33.2 36.7 ± 0.1 35.3 ± 0.1 0.3 0.1
I9 M 40 4.400 0.530 32.5 ± 0.2 36.7 ± 0.1 34.9 ± 0.1 32.5 36.7 ± 0.1 35.3 ± 0.3 0.0 0.4
I10 M 40 3.270 0.500 33.6 ± 0.1 36.2 ± 0.2 35.3 ± 0.2 33.6 36.7 ± 0.1 35.4 ± 0.1 0.5 0.1
I11 M 39 3.500 0.510 32.5 ± 0.4 36.9 ± 0.1 36.2 ± 0.1 32.5 36.5 ± 0.1 35.2 ± 0.1 0.4 1.0
I12 M 40 3.680 0.520 34.3 ± 0.2 36.6 ± 0.1 35.9 ± 0.1 34.3 36.9 ± 0.1 35.7 ± 0.1 0.3 0.2
I13 F 40 2.850 0.470 33.7 ± 0.4 36.5 ± 0.1 35.6 ± 0.3 33.7 36.6 ± 0.1 35.4 ± 0.1 0.1 0.2
I14 F 40 3.150 0.490 33.7 ± 0.3 36.6 ± 0.1 36.0 ± 0.3 33.7 36.7 ± 0.1 35.5 ± 0.1 0.1 0.5
I15 M 40 3.580 0.520 33.8 ± 0.2 36.8 ± 0.1 35.2 ± 0.1 33.8 36.8 ± 0.1 35.5 ± 0.1 0.0 0.3
I16 F 39 3.990 0.520 34.1 ± 0.3 36.3 ± 0.2 35.8 ± 0.2 34.1 36.9 ± 0.1 35.8 ± 0.1 0.6 0.0
I17 M 39 3.810 0.520 34.0 ± 0.2 36.2 ± 0.1 35.5 ± 0.2 34.0 36.9 ± 0.1 35.8 ± 0.1 0.7 0.3
I18 F 39 2.700 0.485 32.5 ± 0.5 36.5 ± 0.1 35.3 ± 0.1 32.5 36.3 ± 0.1 35.2 ± 0.1 0.2 0.1
I19 F 40 3.200 0.500 32.1 ± 0.7 36.5 ± 0.1 35.3 ± 0.1 32.1 36.4 ± 0.1 35.2 ± 0.1 0.1 0.1
Mean 39.6 3.558 0.509 33.6 36.6 35.4 33.6 36.7 35.5 0.25 0.36
SD 0.6 0.507 0.019 0.8 0.2 0.5 0.8 0.2 0.2 0.21 0.36
I – infant, G – gender, GA – gestational age, W – weight, L – length, SD – standard deviation

Tamb – ambient temperature, Tcore – core temperature, Tskin – skin temperature

Table 5 – Mean absolute errors between skin temperature measurements (with infrared
thermography) and simulations for all subjects and body compartments

Subject ^_`a,b ± SD °C ^_`a,c ± SD °C ^_`a,de ± SD °C ^_`a,d ± SD °C ^_`a,f ± SD °C

S1 0.39 ± 0.28 °C 0.35 ± 0.42 °C 0.27 ± 0.32 °C 0.30 ± 0.27 °C 0.18 ± 0.14 °C
S2 0.31 ± 0.19 °C 0.35 ± 0.30 °C 0.31 ± 0.32 °C 0.31 ± 0.27 °C 0. 67 ± 0.43 °C
Mean 0.35 ± 0.06 °C 0. 35 ± 0.00 °C 0.29 ± 0.03 °C 0.31 ± 0.01 °C 0.43 ± 0.35 °C
_
^ – mean absolute error, sk – skin, h – head, t – thorax, ab – abdomen, a – arms, l – legs
a b

z
Är

Äz
A è A‘ r
A A‘
 Skin Fat
Bone
Brain
Lung Viscera
Bone
Bone Bone Bone
Muscle
Muscle Muscle Muscle
Fat
Fat Fat Fat
Skin
Skin Skin Skin

Head Thorax Abdomen Leg Arm

 Environment Mattress
Convection,
Radiation, Conduction
and Evaporation

Blood Flow
Skin x=5 Convection

Conduction
Blood Flow
Fat x=4 Convection

Conduction
Blood
Flow Central
Muscle x=3 Convection Blood
Conduction
Blood Flow
Bone x=2 Convection
Conduction
Blood Flow
Viscera x=1 Convection

Segment - Abdomen
 Thermoregulation in Infants

Control Effectors
Efferent pathway
Heat Gain - Nonshivering thermogenesis
Hypothalamus SNS 1. Brown adipose tissue metabolism
Setpoint Vasomotion

Input
Cutaneous Thermoreceptors
Afferent pathway Core Thermoreceptors
 60

50

40

30 0
4
20 8
12
ys)
(da
16
e
10 tal ag
20 tna
24 s
Po

Transepidermal water loss (g/m2h)

0 28
26 28 30 32 34 36 38
Gestational age (weeks)
 Kangarooing Care

0 15 30 45 60 80 100 120 135 150 165 180

Time (minutes)

Core temperature measurements Skin temperature measurements

 38

37.5 S1 - Model
S1 - Rectal temperature
37
S2 - Model
S2 - Rectal temperature
36.5

Temperature (°C)
I K I
36
0 20 40 60 80 100 120 140 160 180
Time (minutes)
 37 38 38

36
36 36

35
34 34

Temperature (°C)
Temperature (°C)
Temperature (°C)
34
I K I I K I I K I
0 50 100 150 0 50 100 150 0 50 100 150
Time (minutes) Time (minutes) Time (minutes)
Head Thorax Abdomen
37 37

S1 - Model
35 35 S1 - IRT
S2 - Model
S2 - IRT
33 33

Temperature (°C)
Temperature (°C)
I K I I K I
0 50 100 150 0 50 100 150
Time (minutes) Time (minutes)
Arms Legs
Highlights
· A new multi-node model of the thermoregulatory system of newborn infants is proposed.
· The model is customizable, i.e. it meets individual characteristics of the neonates.
· The model was able to accurately predict skin and core temperatures during thermal
neutrality.
· This mathematical approach was capable of simulating thermoregulatory dynamics during
thermal transients