Professional Documents
Culture Documents
In Vivo Study of Human Skin Using Pulsed Terahertz Radiation
In Vivo Study of Human Skin Using Pulsed Terahertz Radiation
In Vivo Study of Human Skin Using Pulsed Terahertz Radiation
Related content
- Terahertz pulse imaging in reflection
In vivo study of human skin using pulsed terahertz geometry of human skin cancer and skin
tissue
radiation Ruth M Woodward, Bryan E Cole, Vincent
P Wallace et al.
- Topical Review
To cite this article: E Pickwell et al 2004 Phys. Med. Biol. 49 1595 E Pickwell and V P Wallace
Recent citations
- Terahertz phase imaging and biomedical
applications
Min Wan et al
E-mail: emma.pickwell@teraview.com
Abstract
Studies in terahertz (THz) imaging have revealed a significant difference
between skin cancer (basal cell carcinoma) and healthy tissue. Since water
has strong absorptions at THz frequencies and tumours tend to have different
water content from normal tissue, a likely contrast mechanism is variation
in water content. Thus, we have previously devised a finite difference time-
domain (FDTD) model which is able to closely simulate the interaction of
THz radiation with water. In this work we investigate the interaction of
THz radiation with normal human skin on the forearm and palm of the hand
in vivo. We conduct the first ever systematic in vivo study of the response
of THz radiation to normal skin. We take in vivo reflection measurements
of normal skin on the forearm and palm of the hand of 20 volunteers. We
compare individual examples of THz responses with the mean response for
the areas of skin under investigation. Using the in vivo data, we demonstrate
that the FDTD model can be applied to biological tissue. In particular, we
successfully simulate the interaction of THz radiation with the volar forearm.
Understanding the interaction of THz radiation with normal skin will form a
step towards developing improved imaging algorithms for diagnostic detection
of skin cancer and other tissue disorders using THz radiation.
(Some figures in this article are in colour only in the electronic version)
1. Introduction
1.1. Aims
Increasing interest in medical imaging at terahertz (THz) frequencies (Woodward et al 2003,
Fitzgerald et al 2003) necessitates an understanding of the fundamental interaction of THz
radiation with biological tissue. The aim of this work is to investigate the response of THz
0031-9155/04/091595+13$30.00 © 2004 IOP Publishing Ltd Printed in the UK 1595
1596 E Pickwell et al
radiation to normal skin and to demonstrate that our finite difference time-domain (FDTD)
model can be applied to model the interaction of THz radiation with tissue.
1.2. Background
Terahertz Pulsed Imaging (TPI© ) is a novel, non-invasive, imaging modality for use in
medicine. TPI uses pulses of electromagnetic radiation typically with a full width at half
maximum of 0.3 ps and an average power of 100 nW. The pulses are detected coherently using
a photoconductive device and the Fourier transformed pulse gives a usable frequency range of
0.1–3 THz (Cole et al 2001).
It has been shown using TPI that there is a significant difference between the response
of THz radiation to normal skin and basal cell carcinoma (BCC), a type of skin cancer
(Woodward et al 2003). Evidence from PET and MRI studies indicates that tumours have
increased water content (Bruehlmeier et al 2003, Ross and Gordon 1982, Chen et al 1992,
Rofstad et al 1994). Therefore, as water has strong absorption across the entire THz range
(Ronne and Keiding 2002), it is likely that changes in water content are a source of image
contrast. To determine the contrast mechanism, we now try to understand the response of THz
radiation to normal skin.
Normal skin comprises three main layers: the stratum corneum, epidermis and dermis
(outermost to innermost). The water content of healthy skin is around 70% by weight
depending on environmental conditions. The main proteins contributing to the remaining 30%
are collagen and keratin. These proteins are able to occupy a large volume. For example, in
the papillary dermis (upper layer of the dermis) the mean fractional volume of collagen fibrils
is between 66% and 69% (Silver et al 2001).
Formalin fixed, alcohol dehydrated and paraffin mounted biological samples have been
previously imaged at THz frequencies (Loffler et al 2002). Other THz investigations of
fixed biological samples include the larynx of a pig and a human liver with metastasis
(Knobloch et al 2002). THz imaging of pork fat tissue has been compared with an optical
time-gated technique in reference Han et al (2000). In this work we investigate the interaction
of THz radiation with normal skin on the forearm and the palm of the hand in vivo. In doing
this we are able to observe characteristic signals for conditions such as dry skin and also
calculate the thickness of the stratum corneum on the palm.
The high bandwidth of TPI allows spectroscopic information to be obtained across a wide
frequency range. Effects of Rayleigh scattering are less pronounced than in the near-infrared,
and sub-millimetre resolution is achieved. TPI spectra contain both phase and amplitude
information, hence it is possible to determine the absorption and refractive index of a material.
When a wave enters a medium of higher refractive index its phase is shifted by π radians
and the wave is inverted. Thus properties of skin which cause a change in refractive index
are revealed by resulting phase changes. The phase and amplitude information, can therefore
reveal important information about the skin. For example, the dryness of the exposed layer of
skin and the thickness of the stratum corneum on the palm of the hand.
Since THz radiation is also non-ionizing, TPI is likely to be suitable for medical
applications. Being able to non-invasively examine skin is of great medical significance. For
example, skin thickness is an important indicator of changes in the skin due to chronological
ageing and photo-ageing and is also related to bone density (Maheux et al 1996, Rennert et al
2003). At present, the most common method for determining skin thickness and its constituent
layers is through histological examination of a biopsy specimen. This is an invasive, time
consuming and multi-stage process.
In vivo study of human skin using pulsed terahertz radiation 1597
Skin disorders such as eczema and psoriasis cause inflamed, dry and flaky skin. It
would be beneficial if the severity of the condition could be quantitatively measured and thus
healing methods, for example UV light and topical steroids, could be tested for effectiveness
in reducing symptoms.
Another potential application of being able to monitor the surface of the skin, is actinic
keratosis (AK) (Barton et al 2003). AK arises from sun damaged skin and is the first
manifestation in the multi-step process of skin carcinogenesis to invasive squamous cell
carcinoma. Therefore it is an ideal target for chemo-preventative efforts and non-invasive
measures of AK severity would be needed to assess the efficacy of such prevention agents.
Since water plays a key role in both normal tissue and tumour, we have devised a
FDTD model which is able to closely simulate the interaction of THz radiation with water
(Pickwell et al 2004). We expect the response of THz radiation to skin to predominantly
behave like that to water. However, since skin also contains proteins which will affect the
response, we anticipate that there will be slight differences between measurements and our
simulations.
Understanding the interaction of THz radiation with normal skin is a step towards
developing diagnostic detection of skin disorders and their severity using THz radiation.
Our aim is to understand and model the response of THz radiation to tissue. THz radiation is
highly attenuated by skin which means that the current signal-to-noise limits the technique to
measurements in reflection for in vivo studies. If we were to transform the measured electric
field of the reflected THz pulse to the frequency domain, it is unlikely to be useful when
investigating skin, due to multiple reflections off graded interfaces of unknown refractive
indices. Therefore, in contrast to work by Walker et al (2003) on Monte Carlo methods in
transmission, we have devised a simulation which can be validated for reflection data in
the time domain. FDTD techniques exploit the time and position dependence of Maxwell’s
equations to model electromagnetic waves.
Studies in Ronne and Keiding (2002), Kindt and Schmuttenmaer (1996) and Ronne
et al (1997) conclude that in the THz regime, the response of water can be fitted by a double
Debye model characterized by a slow relaxation mode (τ1 ) and a fast relaxation mode (τ2 ).
Debye theory couples the relaxation of the local polarization, P to the local electric field, E
characterized by time constant τ and coupling strength γ :
dP
P +τ = γ E. (1)
dt
By treating the two polarization decay rates independently, we use double Debye theory to
model the fast and slow relaxation processes. Skin is around 70% water (depending on
environmental conditions). We apply double Debye theory to skin and appreciate that the
relaxation times are likely to change to accommodate both the bound and unbound water
present. Additional relaxation modes may potentially be required but we do not investigate
this here. Debye theory has previously been applied to human breast tissue at microwave
frequencies by Li and Hagness (2001). For a material obeying the double Debye model
the frequency-dependent dielectric function ˆ (ω) is given by
εs − ε2 ε2 − ε∞
ˆ (ω) = ε∞ + + (2)
1 + iωτ1 1 + iωτ2
1598 E Pickwell et al
Figure 1. A schematic diagram of the relay optics used in the TPI reflection system. The THz
optics which are scanned in the x–y plane are enclosed in the dashed box.
where εs is the static dielectric constant, ε∞ is the limiting value at high frequency and ε2 is an
intermediate frequency limit. We refer to εs , ε2 , ε∞ , τ1 and τ2 as the double Debye parameters.
Polarization is then given by
dP1 dP2
P1 + P2 + τ1 + τ2 = (γ1 + γ1 )E (3)
dt dt
and from Maxwell’s equations
dH d
− = σ E + 0 (ε∞ E + P ). (4)
dz dt
The key concept in our simulation is to incorporate the time dependence of the polarization
using double Debye theory from equation (3) into equation (4). From these equations we
derived iterative equations to model the local E and H fields and polarization, P where
P = P1 + P2 . We therefore were able to simulate the propagation of a THz pulse through
the discretized region in the time domain. To model electromagnetic problems accurately,
a relatively large number of cells (10–20) per wavelength is required (Taflove 2003). For a
frequency of 3 THz, the wavelength is 100 µm and so we used a cell size of 10 µm.
data. Deconvolution is performed to extract the impulse function (equation (5)). Dividing the
sample by the reference removes the system response but also increases the noise. Several
methods could be used to remove the noise, for example a low pass filter (Mittleman et al
1997). In this work a Gaussian filter function is applied to remove both the low and high
frequency noise components and also to produce a suitable time-domain response. This is
described further in Woodward et al (2002). The same filter is used for the excitation function
input to the FDTD simulation to match the bandwidth of experimental data:
FFT(raw)
FFT(Impulse function) = FFT(filter) ⊗ . (5)
FFT(reference)
For each measurement, the site under investigation was placed on the quartz imaging plate
and an area of 1 cm × 1 cm was scanned in approximately 60 s to collect a grid of 220 (20 ×
11) data points; each data point is a time series containing 512 points over 40 ps. The mean
impulse function for the grid of data points is calculated from all 220 points. For the frequency
range 0.1–3 THz, the coherence length in skin is approximately 100 µm which corresponds to
a depth resolution in reflection of 50 µm (Johnson et al 2001). For the maximum frequency
of 3 THz, the spatial resolution, using the Rayleigh criterion, is approximately 60 µm.
TPI measurements of human subjects have been approved by the Local Research Ethics
Committee. For this study, TPI measurements were performed on 20 volunteers who had
given their informed, signed consent.
In vivo reflection measurements using TPI scan of normal skin on the forearm and palm
of the hand of 20 volunteers have been used to study variations in skin characteristics. We
compare examples of individual responses with the mean response for the three areas under
investigation. All volunteers were white Caucasian. The ages of the volunteers ranged between
24 and 49 years, with mean 33 years. There were four female volunteers. Two sites on the
forearm were scanned—the inner forearm (volar forearm), and the back of the forearm (dorsal
forearm). The palm of the hand was scanned between the thumb and wrist as this region was
thought easiest to position flat on the quartz plate. To be able to use these measurements to
calculate the thickness of the stratum corneum, the palm must make good contact with the
quartz window. If the contact is poor then there is an air gap which gives rise to an air–skin
interface (low refractive index to high refractive index) instead of the quartz–skin interface
(high refractive index to lower refractive index). This would significantly affect the shape of
the impulse function (it drastically increases the peak to peak value), and in most cases would
result in a lower value for the stratum corneum thickness.
To investigate reproducibility, weekly measurements were taken for four weeks. Thus
variation over time as well as person-to-person variation could be considered. Out of the 20
volunteers, 18 were able to attend all four weekly measurements.
(a) (b)
Figure 2. (a) Mean impulse function of the weekly measurements and (b) overall mean impulse
function and standard error for the volar forearm of volunteer 18.
This is explained further in Pickwell et al (2004). This was done for the volar forearm to
obtain parameters for a one-layer simulation.
The skin on the palm of the hand was treated as two layers—the stratum corneum and
the epidermis. By using the Debye parameters of the epidermis, parameters for the stratum
corneum were determined such that the resulting simulation gave a good fit to the measured
impulse function.
3. Results
In this section for the three sites investigated, we analyse the impulse functions plotted against
optical delay in picoseconds.
Of the three sites measured, the volar forearm gave the most consistent results as it has
minimal adnexa and is also less exposed to the sun. Sun exposure may affect skin hydration
and pigmentation as well as causing general damage/irritation. There was minimal variation
(defined here as the standard error divided by the mean) between pixels within a grid (less than
1%) and only slight variation (less than 3%) over time. Figure 2(a) plots the mean impulse
function from each weekly measurement of volunteer 18. The mean and standard error of
these impulse functions are illustrated in figure 2(b). The mean of the impulse functions from
all 18 volunteers is plotted in figure 3. This illustrates the person-to-person variation of the
volar forearm. The dotted lines depict the spread of the measurements. As expected, there are
slight differences between volunteers. The small peak at the beginning of the impulse function
is due to the presence of dry skin on some volunteers. Dry skin has a lower refractive index
than quartz and so there is no phase change in the reflection off this top surface of skin. When
the wave enters more hydrated skin (slightly below the surface and therefore later in time),
where the refractive index is greater, the reflected wave undergoes a phase change. The lack
of phase change at earlier times means that a peak is seen before the main reflection. This is
illustrated schematically in figure 4.
In vivo study of human skin using pulsed terahertz radiation 1601
Figure 3. Mean volar forearm impulse function taken from data of 18 volunteers on four occasions.
Stratum
corneum
Epidermis
nq nsc ne
Input
No Phase Change
Surface reflection
Phase Change
Stratum corneum,
epidermis reflection
Combined reflection
Figure 4. Schematic representation of the reflection off the volar forearm where nq , nsc and ne
represent the refractive indices of the quartz, stratum corneum and epidermis, respectively. In
addition, nsc < nq , ne > ns and nq = 2.1.
The existence of the dry top layer is only seen when it has a refractive index lower than
both quartz and the epidermis. This is because the dry layer is too thin to be resolved by the
system and requires a lack of phase change in the surface reflection to be noticeable.
Figure 6. Mean of dorsal forearm impulse functions from all volunteers. The dotted line indicates
the spread of the measurements.
which gives rise to an air–skin interface (low refractive index to high refractive index) instead
of just the quartz–skin interface (high refractive index to lower refractive index) and this results
in increased peak-to-peak amplitude, as seen for volunteer 9. In addition, depending on the
positioning of the arm, it is often not possible to make perfect contact with the quartz window.
Thus the weekly differences could be due to the positioning of the dorsal forearm, and also
the pressure with which it is pushed against the quartz plate (as this will reduce air gaps).
The mean of dorsal forearm impulse functions from all volunteers is plotted in figure 6.
The dotted line indicates the spread of the measurements. This also illustrates the differences
between subjects.
3.3. Palm
Palm measurements for volunteer 17 differed largely from week to week. This is depicted
in figure 7 where the mean impulse function of the four measurements is plotted with the
In vivo study of human skin using pulsed terahertz radiation 1603
Figure 7. The impulse function over time for the palm of volunteer 17.
standard error between the weekly measurements. The variation for volunteer 17 was around
10% suggesting that measurements are very sensitive to the exact site imaged. The thickness
of the stratum corneum varies considerably over the palm due to the structure of the hand and
general weathering. This contrasts with the volar forearm which is more uniform.
There was also large person-to-person variation between subjects’ palms. In figure 8 we
see the mean of impulse functions from all volunteers and also the spread of the measurements.
Person-to-person variation is expected due to genetic differences and environmental conditions.
These will affect the thickness and hydration of the stratum corneum. The stratum corneum
thickness affects the separation of the two minima and the hydration of the top surface of the
stratum corneum will determine the presence of the small peak before the first minimum. It
is therefore not very meaningful to calculate the average of all impulse functions. As can be
seen in figure 8, artefacts are introduced.
Figure 9. Complex refractive index of the epidermis, calculated using TPI reflection measurements
of the volar forearm.
s 2 ∞ τ1 (ps) τ2 (ps)
Epidermis 58 3.6 3 9.4 0.18
index. Since there is not always a strong peak before the first minimum, the stratum corneum
must be closely index matched to the quartz. Therefore, we assume the refractive index of the
stratum corneum to be 2 (Cole et al 2001, Fitzgerald et al 2003).
The mean stratum corneum thickness of the volunteers was calculated from data to be
170 ± 8 µm.
3.4.2. Two-layer system: palm. The simulated and measured impulse functions for the
palm are illustrated in figure 11. The simulation fits the measured impulse function with a
correlation coefficient of 0.95.
The differences between the measured and simulated impulse functions are likely to be
due to the model oversimplifying the skin structure—the water concentration gradient across
the stratum corneum has not been entered into the simulation—instead two homogeneous
layers have been assumed. It has also been assumed that the double Debye model can
In vivo study of human skin using pulsed terahertz radiation 1605
Figure 10. Comparison of measured and simulated impulse functions of the (volar) forearm.
Figure 11. Comparison of measured and simulated impulse functions of the palm.
be used to model skin—it has been used successfully in other tissues at lower frequencies
(Li and Hagness 2001). However, skin contains a mixture of proteins as well as water. This
introduces more molecular interactions to consider and assumptions within the model may
need to be revised. For instance, extracellular water in distributed around collagen fibres and
water in micelles should have spectroscopic THz signatures which are different from bulk
water.
4. Conclusions
This work has presented new in vivo data of human skin in the THz region of the EM spectrum.
By investigating the three sites, volar forearm, dorsal forearm and palm, we have shown the
impulse function from the volar forearm to be the most consistent throughout all subjects. We
have also illustrated the large amount of variation in measurements of the palm. This is due
to the variation of the stratum corneum thickness on a palm and between subjects.
1606 E Pickwell et al
In short, the impulse functions have shown that we can detect changes in refractive index
including those due to skin dryness and poor contact.
We have also validated the application of our FDTD model to skin. Our simulated
impulse functions for one and two layer skin models agree well with our measured data. The
next stage of the simulation is to model well-defined multi-layered systems. In addition,
there is a water concentration gradient across the stratum corneum (Caspers et al 2000). This
can be simulated by our model. If tumour has a higher water content than healthy tissue,
then incorporating buried layers of increased water content into the model could be used to
investigate the sensitivity of TPI to detect tumours.
The success of applying our FDTD simulation to both water (Pickwell et al 2004) and
skin aids understanding of the interaction of THz radiation with biological systems. This
will potentially benefit research into diseases such as skin cancer and other epithelial cancers
and is a first step towards understanding the origin of contrast in THz pulsed images of skin
cancer. Modelling the propagation of THz radiation in tissue will enable better interpretation
of images and more effective image reconstruction.
In summary, we have applied our FDTD model to in vivo THz measurements by using
double Debye theory to describe the interaction of skin with THz radiation. This model could
also be extended to other emerging technologies in the THz field.
Acknowledgment
One of the authors (EP) gratefully acknowledges financial support from EPSRC and TeraView
Ltd.
References
Loffler T, Siebert K, Czasch S, Bauer T and Roskos H G 2002 Visualization and classification in biomedical terahertz
pulsed imaging Phys. Med. Biol. 47 3847–52
Maheux R, Guy J, Dumont M and Mâsse B 1996 Correlation between skin thickness and bone mass in women
J. North Am. Menopause Soc. 3 197–200
Mittleman D M, Hunsche S, Boivin L and Nuss M C 1997 T-ray tomography Opt. Lett. 22 904–6
Pickwell E, Cole B E, Fitgerald A J, Pepper M and Wallace V P 2004 Simulation of terahertz pulse propagation in
biological systems Appl. Phys. Lett. 84 2190–2
Rennert J, Kees G A and Ruchti T L 2003 Non-invasive method of determining skin thickness, characterizing layers
of skin tissue in vivo US Patent 6671542
Rofstad E K, Steinsland E, Kaalhus O, Chang Y B, Hovik B and Lyng H 1994 Magnetic resonance imaging of human
melonoma xenografts in vivo: proton spin lattice, spin-spin relaxation times verses fractional water content and
fraction of necrotic tissue Int. J. Radiat. Biol. 65 387–401
Ronne C and Keiding S R 2002 Low frquency spectroscopy of liquid water using THz-time domain spectroscopy
J. Mol. Liq. 101 199–218
Ronne C, Thrane L, Astrand P, Wallqvist A, Mikkelsen K V and Keiding S R 1997 Investigation of the temperature
dependence of dielectric relaxation in liquid water by THz reflection spectroscopy and molecular dynamics
simulation J. Chem. Phys. 14 107
Ross K F and Gordon R E 1982 Water in malignant tissue, measured by cell refractometry and nuclear magnetic
resonance J. Microsc. 128 7–21
Silver F H, Freeman J W and DeVore D 2001 Viscoelastic properties of human skin and processed dermis Skin Res.
Technol. 7 18–23
Taflove A 2003 http://www.ece.northwestern.edu/ecefaculty/Allen1.html
Walker G C, Berry E, Smye S W, Zinv’ev N N, Fitzgerald A J, Miles R E, Chamberlain M and Smith M A 2003 Two
methods for modelling the propagation of terahertz radiation in a layered structure J. Biol. Phys. 129 141–8
Woodward R M, Cole B E, Wallace V P, Pye R J, Arnone D D, Linfield E and Pepper M 2002 Terahertz pulse imaging
in reflection geometry of human skin cancer and skin tissue Phys. Med. Biol. 47 3853–63
Woodward R M, Wallace V P, Pye R J, Cole B E, Arnone D D, Linfield E and Pepper M 2003 Terahertz pulsed
imaging for ex vivo basal cell carcinoma J. Invest. Dermatol. 120 72–8