Download An introduction to medical imaging with coherent terahertz

INSTITUTE OF PHYSICS PUBLISHING
PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 47 (2002) R67–R84
PII: S0031-9155(02)31211-9
TOPICAL REVIEW
An introduction to medical imaging with coherent
terahertz frequency radiation
A J Fitzgerald1, E Berry1, N N Zinovev2, G C Walker1, M A Smith1
and J M Chamberlain2
1
Academic Unit of Medical Physics and Centre of Medical Imaging Research,
University of Leeds, Leeds, LS1 3EX, UK
2 Institute of Microwaves and Photonics, University of Leeds, Leeds, LS2 9JT, UK
Received 26 November 2001, in final form 23 January 2002
Published 20 March 2002
Online at stacks.iop.org/PMB/47/R67
Abstract
Methods have recently been developed that make use of electromagnetic
radiation at terahertz (THz) frequencies, the region of the spectrum between
millimetre wavelengths and the infrared, for imaging purposes. Radiation at
these wavelengths is non-ionizing and subject to far less Rayleigh scatter than
visible or infrared wavelengths, making it suitable for medical applications.
This paper introduces THz pulsed imaging and discusses its potential for
in vivo medical applications in comparison with existing modalities.
1. Introduction
The region of electromagnetic (EM) radiation occupying the band of frequencies from
100 gigahertz (GHz) to 100 terahertz (THz), bounded by millimetre waves at the lower
end and infrared at the upper, is referred to as the terahertz frequency band. In the past this
band, which corresponds to wavelengths between 3 mm and 3 µm, was sometimes referred
to as the terahertz gap because of the lack of practical coherent sources and detectors. The
technology to generate and detect THz radiation has changed considerably over the past few
years, opening up the THz gap for exploitation. Previously, for techniques such as far infrared
Fourier spectroscopy, THz radiation was generated using thermal sources that produced weak
and incoherent radiation, or by highly complex and bulky equipment such as free electron
lasers or optically pumped gas lasers. Liquid helium-cooled bolometers, which deliver a signal
proportional to radiation intensity, with a relatively poor noise performance, were traditionally
used as the most sensitive detectors. Recent advances in the fields of ultrashort pulsed lasers,
non-linear optics and crystal growth techniques have culminated in practical sources of bright,
coherent, broadband THz pulses and enabled room temperature detection (Auston and Nuss
1988). Coherent detection methods make it possible to record not just intensity, but a timeresolved amplitude of the electric field. A typical broadband THz pulse example is presented
in figure 1.
0031-9155/02/070067+18$30.00
© 2002 IOP Publishing Ltd Printed in the UK
R67
R68
Topical Review
Figure 1. (a) Example of the broadband THz pulse produced by electro-optical techniques, and
(b) frequency spectrum of the pulse, including water vapour absorption lines.
The technique of THz time domain spectroscopy (THz-TDS) makes use of the coherent
detection of these broadband pulses to analyse the THz spectral response of chemicals and
materials more sensitively than previous spectroscopic methods in this range (Grischkowsky
et al 1990, Kindt and Schmuttenmaer 1996, van Exter and Grischkowsky 2001). The technique
can be extended to terahertz pulsed imaging (TPI), which can be used in either transmission
or reflection geometries. The spectroscopic response of an object is mapped by recording
transmitted or reflected THz pulses at many contiguous pixel locations (Hu and Nuss 1995,
Wu et al 1996). The types of systems used for TPI and the practicalities of imaging, such as
spatial resolution and signal to noise ratio, are discussed in section 3.
In comparison with electromagnetic radiation used in other medical imaging modalities,
the use of the THz frequency band has several advantages. Firstly, THz quanta are far less
energetic than those of x-rays and therefore do not pose an ionization hazard for biological
tissue. Secondly, scattering of electromagnetic radiation in heterogeneous biological material,
which may be more complex than a simple Rayleigh treatment, is many orders of magnitude
less for the THz band than for the neighbouring infrared or visible regions of EM spectrum, as
demonstrated in a direct imaging comparison by Han et al (2000). Thirdly, many biological
molecules exhibit vibrational and rotational modes at THz frequencies (Smye et al 2001)
which may provide characteristic fingerprints to differentiate biological tissues in a region
of the spectrum not previously explored for medical use. The interaction of THz radiation
with matter is discussed in section 2 and potential medical applications, which are likely to
be dictated by the strong absorption of THz wavelengths by aqueous media, are outlined
in section 4.
The purpose of this paper is to present an introduction to the generation and detection of
THz signals for imaging. We address theoretical and practical aspects of spatial resolution,
temporal resolution, spectral information, signal-to-noise levels and safety limits before
outlining potential medical applications.
Topical Review
R69
2. The interaction of THz radiation with matter
The interaction of THz radiation of frequency ν with an object can be expressed in terms of
the frequency-dependent complex refractive index N(ν) of the material
cα(ν)
(1)
4πν
where n(ν) represents the real refractive index as a function of the frequency ν, c is the speed
of light in a vacuum and α(ν) is the absorption coefficient. THz-TDS makes it possible to
determine an object’s full complex refractive index by coherent detection of the amplitude of
the electric field of the THz pulse (Grischkowsky et al 1990).
Complex refractive indices are usually determined in THz-TDS by comparison of reflected
or transmitted fields propagated within the object of interest, with the incident reference
pulse E 0 . In the case of coherent detection of transmitted field amplitude E T at location r,
frequency ν propagated through a slab of thickness d, the complex transmittance coefficient,
t, is given by (Born and Wolf 1959)
α(r, ν)
E T (r, ν)
2πν
≈ t01 (r, ν)t02 (r, ν) exp i
n(r, ν)d exp −
d
(2)
t=
E 0 (r, ν)
c
2
N(ν) = n(ν) + i
where we retain in-plane r-dependence of the values describing the differences of the optical
properties in the plane perpendicular to the propagation of light. Transmission fields t01 (r, ν)
and t02 (r, ν) are the Fresnel coefficients for the front and back surfaces of the slab respectively.
Equations (1) and (2) show that with coherent THz pulses one simultaneously obtains the
spectral dependencies for both the coefficient of absorption α(r, ν) and the refractive index
n(r, ν) as well as the local properties of the interfaces. A similar approach can be used to
determine these coefficients from reflection measurements.
THz-TDS has been used to characterize many gases, liquids and solids, revealing sharp
resonances in the THz region for gases, corresponding to rotational and vibrational energy
modes of the molecules (Jacobsen et al 1996), and broader absorption peaks in solids and
liquids. Other substances that have been characterized in the THz region of the spectrum
include semiconductor materials (Greene et al 1992) and dielectrics (Grischkowsky et al
1990).
Spectroscopic methods are beginning to find uses in the biological sciences for
investigating DNA (Brucherseifer et al 2000, Markelz et al 2000), proteins (Walther et al
2000) and tissues (Arnone et al 1999, Han et al 2000). In vivo tissue is highly heterogeneous
and its spectroscopic response will depend on the constituent materials and their physical
arrangement. This means that in vivo measurements on tissues will be different from
spectroscopic measurements on homogeneous samples of molecules or proteins. THz
interactions with biological components of tissue are reviewed by Smye et al (2001).
Many non-polar liquids are reasonably transparent to THz radiation (Pedersen and
Keiding 1992), while polar liquids such as water (Thrane et al 1995) and alcohols (Kindt
and Schmuttenmaer 1996) strongly attenuate THz pulses. Dry materials such as plastics,
paper, cardboard, cloth and wound dressings are readily transparent, (Mittleman et al 1996),
although, as is shown in figure 2, when saturated with water they can become highly absorbing.
Measurements of the absorption coefficient of water show that it strongly attenuates
signals in the THz region, with broad absorption peaks at 6 THz and 19.5 THz
(Kindt and Schmuttenmaer 1996). This has strong implications for the use of TPI for medical
imaging. On one hand the high absorption limits the depth of penetration of signals through
tissues with high moisture content, on the other, it makes THz methods very sensitive to
changes in the water content of tissue. Measurements by Bezant (2000) demonstrated that a
R70
Topical Review
Figure 2. THz pulse transmitted through 3 mm of dry alginate dressing (solid line) and the same
dressing saturated with water (dotted line).
THz pulse could be detected after penetration through at least 1.5 mm of moist skin tissue
when the detection signal-to-noise ratio (SNR) is 500 to 1. This is considerably further than
through water alone. Mittleman et al (1996) addressed the question of sensitivity to water
content and suggested that for an SNR of 100, a minimum concentration given by n × x =
1016 cm−2 can be detected, where n is the density of water molecules and x is the path length.
3. Imaging using THz radiation
The advent of cheaper and more practical sources and detectors of THz radiation has naturally
led to TPI, imaging with THz pulses (Nuss 1996, Wu et al 1996). The dominant imaging
method at this time uses electro-optical (Jiang and Zhang 1999a) or photoconductive antenna
(Mittleman et al 1999) techniques to generate and record broadband pulsed signals. Research
into the other optical method, based on mixing two continuous wave (CW) laser frequencies,
is ongoing (Brown et al 1997, Seibe et al 1999, Verghese et al 1998), as is the development
of all-electronic THz transceivers (Eisle 1999).
3.1. Imaging principles of a TPI system
TPI uses ultra-fast infrared pulsed lasers with pulses typically of the order of 10–100 fs to
generate and detect THz radiation. The approach is based on the ‘pump–probe’ technique of
optical spectroscopy. The laser pulse is split into a pump beam, used to generate THz pulses,
and a probe beam used to detect coherently the amplitude of the THz electric field (figure 3).
Broadband THz pulses are produced by illumination with the pump beam of voltagebiased photoconductive antennae (Auston et al 1984), or in a crystal with high non-linear
susceptibility (optical rectification or/and optical mixing) (Auston and Nuss 1988, Zhang et al
1992). Off-axis parabolic mirrors are used to focus the generated THz pulse onto the
sample and to collect the reflected or transmitted pulse after interaction with the sample,
recombining it with the probe beam at the detector stage. The probe beam gates the detector
and thereby determines the amplitude of the THz electric field at a fixed time point using similar
Topical Review
R71
Figure 3. Schematic of THz pulsed imaging system in transmission mode.
photoconductive or optical rectification principles as in the generation of THz frequency
radiation (van Exter et al 1989, Wu and Zhang 1996). The time point of measurement is
determined by the time delay of the probe beam, set by an optical delay stage. An entire THz
pulse profile is reconstructed by scanning the optical delay line across the selected time-domain
range at a number of discrete points.
In the simplest case a spatial image is mapped pixel by pixel using X–Y translation of the
sample or beam in the focal plane. However, this individual scanning process is very slow and
time consuming. An alternative, faster technique that compromises on SNR, uses a broad beam
illumination scheme with a multi-element array of detectors; for example, a charge coupled
device (CCD) as recently introduced by Wu et al (1996). The trade-off between measurement
time and SNR of THz measurements and the limitations imposed by the techniques used are
discussed in section 3.3
3.2. Frequency range
The frequency band available for image formation and spectroscopic analysis depends on the
source of THz radiation as well as on the detector design. All electronic methods are generally
limited to frequencies below 1 THz. Optical CW systems produce a single frequency, tuneable
within the band 0.1 and 3 THz. Pulsed sources typically give broadband radiation, with a
bandwidth extending from nearly 0 Hz up to several THz as shown in figure 1. Frequencies as
high as 100 THz are attainable, depending on the type of electro-optic conversion crystal and
R72
Topical Review
the duration of the laser pulse. The typical frequency content for photoconductive antenna
sources with laser pulses shorter than 100 fs is a few GHz to 3 THz.
3.3. Imaging considerations, measurement time and pulse signal-to-noise ratio
To exploit the spatial and spectral information available, large amounts of data must be
collected in the spatial and temporal domains. The volume of data depends on the desired
resolutions in the temporal and spatial domains, hardware properties and many accompanying
circumstantial factors. Among them one issue imposes the lowest limit on the data size. It is
the Nyquist criterion that establishes the minimum number of time data points, Mt and pixels,
Mxy to be acquired. According to this criterion the sampling interval must not be less than
double the highest frequency to be detected in time or spatial domains. Assuming the spot in
focus has a diameter approximately equal to the diameter of the Airy disk, then the pixel size
should be at least 4.6 times smaller than the size of the disk. For example, if the Airy disk
diameter, i.e. the diameter of the THz focal spot, is 1 mm then the sampling step must not
exceed 200 µm. If the image square is limited to 20 × 20 mm2 then the minimum number of
points that must be acquired to avoid aliasing and undersampling, and to resolve the required
spatial frequencies, is Mxy = 128 × 128 (rounded to the nearest power of 2).
The above discussion concerns only the X–Y dimensions of the data array. Similar
arguments can be applied to calculate the minimum number of points required in the time
domain to adequately represent spectral responses. This number depends on the THz frequency
bandwidth and the desired spectral resolution. If a conventional ZnTe emitter is used then the
highest frequency to be detected falls in the range 3–5 THz. This highest frequency determines
the sampling interval in the time domain, δt 1/(2fmax ), which for this case is near to 100 fs.
For a spectral resolution of f ∼ 50 GHz the number of points required in the time domain,
Mt = 1/(δtf ) is also close to 128. However, for many cases this spectral resolution may
not be sufficient and the number of points required in the time domain could be as high as
1024. It is important to emphasise that TPI offers an advantage over many existing imaging
techniques because it offers spatial localization, combined with spectroscopic capabilities, in
the unprecedentedly wide band of frequencies, from DC up to several THz.
Combining the spatial and temporal dimensions gives a total number of acquisition points
of the order 106–107. Using mechanical scanning systems, the data collection of a single
spectral image can take many hours. In spite of this, these systems are still widely used in
THz experiments because they offer modulation and gating methods that enhance the SNR of
a measured pulse. Jiang and Zhang (1999b) and Mittleman et al (1996) reported SNR values,
calculated by comparing pulse amplitude to the noise floor when no pulse is present, to be as
high as 106 for these types of systems.
To reduce acquisition time one must trade off some of the properties restricting the
bandwidth (the applications of narrow band THz emitters working at fixed frequencies), or
perform simultaneous acquisition in two of the three dimensions (the applications of multielement detector arrays). These naturally operate at the expense of SNR, the main sources
of noise being fluctuations in the ultrafast femtosecond laser and Johnson noise in detectors
(Mittleman et al 1996, Zhao et al 2002) and result in a subsequent degradation in image
quality. Below we consider several solutions to reduce acquisition times.
Mittleman et al (1996) discuss the implementation of a fast scanning optical delay line
in which the delay stage continuously oscillates at a frequency of 100 Hz. This approach
can reduce acquisition times for individual pixels from several minutes down to 25 µs, with
a subsequent reduction in signal to noise due to the loss of noise filtering capabilities. Jiang
and Zhang (1999a) present a method using broad beam illumination and a CCD system of
Topical Review
R73
measurement that can acquire an entire array of pixels at a single time point, in roughly 25 ms.
The SNR for these systems are of the order 1 × 103 (Wu et al 1996). Jiang and Zhang (2001)
developed an alternative method to alleviate the need for mechanical translation stages, using a
linearly chirped laser pulse. They report that this method requires only milliseconds to record
the full time domain data for a single pulse. However, this method compromises spectral
resolution due to the limitations of measurement duration and temporal resolution and has a
poor SNR of about 60. Of course, averaging can be used to improve the SNR of these faster
acquisition methods, but this negates the gain in speed. To achieve performance comparable
with slower systems the averaging time required is usually comparable to the acquisition time
of the slower systems (Jiang and Zhang 1999a).
3.4. Parametric images
From the wealth of data collected for each pixel a large number of parameters can be used
to form THz images. Pulse height and shape, and delays in the time domain provide
contrast information. Additionally, the spectral content of the pulse can be exploited for
pixel by pixel spectroscopic analysis and material characterization, including transmittance
and absorbance parameters, as commonly used in Fourier transform infrared spectroscopy
(Lasch and Naumann 1998). Combinations of any of these parameters are also possible.
Combinations of parameters are useful for developing thickness-independent parameters that
allow differentiation between time delays due refractive index effects or sample thickness. In
many cases the parameters can be compared to a reference pulse to obtain a relative measure.
Similar parameters can be used for both reflection and transmission images.
Some commonly used image parameters are described below. These are separated into
two main categories: (i) those based on the pulse profile in the time domain and (ii) frequency
domain parameters, determined using information at individual frequencies or combinations
of frequencies in the spectral domain.
Time-domain parameters include:
• Pulse amplitude. Maximum value of pulse in time domain for each pixel. It can be
expressed relative to the reference amplitude and represents the proportion of transmitted
signal for each pixel.
• Pulse time delay. Time difference between the maximum of reference and pixel pulses in
time domain. The thickness and refractive index of a material determine this delay.
• Pulse width. Full width at half maximum of pixel pulse. It is dependent on the ratio of lowand high-frequency components and an increase (pulse broadening) implies preferential
absorption of high frequencies.
• Instantaneous electric field. The electric field amplitude imaged at points in the time
domain. This shows the propagation of the electric field through the material.
Frequency domain parameters include:
• Transmittance. Ratio of transmitted-to-incident intensities at a given frequency,
determined by the sample thickness and absorption coefficient. The use of integrals
over bands in the spectrum can improve noise characteristics for this and other frequency
domain parameters.
• Time delay. Delay time due to the presence of the sample, calculated from the phase
component of Fourier transform. It is proportional to the thickness and refractive index
of the sample.
• Absorbance. Logarithm of the inverse of transmittance.
R74
Topical Review
Figure 4. (a) Reference THz pulse (full line) compared to pulses delayed and attenuated by
0.5 mm (chain curve), 2 mm (broken line) and 4 mm (dotted line) of duraform (nylon 12), (b)
corresponding spectral content of each pulse. Measurements courtesy of Physikalishes Institut der
Johann Wolfgang Goethe-Universitaet Frankfurt.
• Dual frequency or panchromatic. Use of two or more frequency components to image
(for example, ratio of transmittance at ν 1 relative to transmittance at ν 2).
• Phase. Unwrapped phase component from Fourier transform at the given frequency. This
is proportional to the time delay and depends on sample thickness and THz frequency.
• Dispersion. Derivative of transmittance with respect to frequency. This displays the
change of absorption as a function of frequency and highlights characteristic peaks in the
spectral domain.
• α(ν)/(n(ν) − 1). Independent of thickness due to the cancellation of the thickness terms
by the ratio of the expressions of absorption coefficient and refractive index. Will be
material dependent.
Figure 4 shows an example of pixel pulses transmitted through various thicknesses of
a step wedge and figure 5 illustrates various image parameters derived from these pulses
and spectra. The wedge was manufactured from duraform (nylon 12) by the selective laser
sintering process (Berry et al 1997). From figure 4 it can be seen that thicker steps attenuate
and delay the pulse by greater amounts. This provides a contrast in the parametric images
shown in figure 5 for pulse amplitude and pulse delay in the time domain and transmittance
and time delay at 1 THz in frequency domain. The dual frequency transmittance parametric
image shows a contrast between the steps because the higher frequency (1.0 THz) is more
strongly attenuated than the lower (0.8 THz), as is evident in the spectrum shown in figure 4.
Through the instantaneous electric field parameter it is possible to selectively image individual
steps by selection of the appropriate time point.
Figure 6 shows parametric images of a melanoma that was prepared in the standard histopathological way (dehydrated, formalin fixed and wax embedded). Each of the parametric
Topical Review
R75
0.05
(a)
0.55
1.0
0.02
(d)
0.65
0.5
(b)
(e)
14.0 −160
4.3
−0.3
(c)
10
(f)
0.55
Figure 5. THz images of a duraform (nylon 12) step wedge, with step heights of (from right to
left) 0.1 mm, 0.5 mm, 2 mm, 4 mm, 6 mm and 7 mm. The images are: (a) pulse amplitude (relative
to reference), (b) pulse time delay, (c) instantaneous electric field at 5.6 ps, (d) transmittance at
1 THz, (e) time delay at 1 THz and (f) dual frequency image (transmittance at 0.8 THz relative to
1.0 THz). Partial volume effects at step boundaries are responsible for the banding between steps.
Figure 6. Parametric images of the wax embedded melanoma sample (shown in the photograph
on the left) which was prepared in the normal histo-pathological procedure. Images are: (a)
transmittance at 0.5 THz, (b) transmittance at 2 THz, (c) phase angle at 1 THz (from Fourier
transform), (d) α(ν)/(n(ν) − 1), (e) dispersion at 1 THz, (f) dual frequency image (transmittance
at 1 THz relative to 1.5 THz). The images have been interpolated to reduce pixelation effects.
images shows the melanoma in a different way. In the relative transmittance image at
2.0 THz the tumour is more clearly seen relative to the background than in the 0.5 THz relative
transmittance image due to its stronger absorption at higher frequencies. The image based on
the phase angle parameter, which is related to time delay, demonstrates that there is a clear
difference in the refractive index between the tumour and the background. The differences in
absorption coefficient and refractive index between the tumour and the background result in
a contrast when the two are combined as a thickness independent ratio α(ν)/(n(ν) − 1). The
final two images demonstrate that absorption of higher frequencies is stronger in the tumour
than in the background material.
3.5. Image spatial resolution capabilities
According to the Abbe criterion, the minimum size of a recognizable feature in an image
formed using EM radiation is ultimately limited by the wavelength, due to diffraction effects,
R76
Topical Review
Figure 7. MTF curves determined from relative transmittance parametric images of the gold on
TPX test object (inset) at frequencies of 0.5 THz (dotted line), 1.0 THz (broken line), 1.5 THz (chain
line) and 2.0 THz (full line). The improvement of spatial resolution with higher THz frequencies
is clearly shown. Gold bars 100 nm thick were deposited on TPX at the spatial frequencies of 0.25
(top row of inset), 0.50, 0.75 (second row), 1.0, 1.375 (third row), 1.625 and 2.0 line pairs per mm
(bottom row).
although in practice, system properties may prevent this limit from being realized. Since the
THz region of the spectrum covers several decades, spatial resolution in the images strongly
depends on the frequency range used for imaging. For example, at 1 THz, assuming minimal
scattering, the diffraction-limited beam waist corresponds to a size close to 300 µm, compared
to 600 µm at 0.5 THz. An effective method of representing the spatial resolution capabilities
of an imaging system is the modulation transfer function (MTF), which graphs the modulation
of the signal by the imaging system at each spatial frequency. Due to smaller diffraction limits,
higher THz frequency components will reproduce higher spatial frequencies better than lower
THz frequency components and hence exhibit a higher MTF. This is illustrated by the example
MTF in figure 7 for a typical THz imaging system. The MTF curves were determined by
measuring the modulation effect on the relative transmittance of THz signals by a series of
deposited gold bands on TPX (Mitsui, UK), a THz transparent plastic. The arrangement and
spatial frequencies of these gold bands is shown in the inset of figure 7. For THz systems
using standard electro-optical techniques, the spectrum above several THz is of low power,
so that the improved resolution is traded against a reduced image SNR at higher frequencies.
Using the principle
√ of confocal microscopy one can reduce the size limit of feature
recognition to ∼ λ/ 2. Using near-field conditions it can be improved even further
(Hunsche et al 1997). An aperture smaller than the diffraction limited beam waist is placed
in front of the THz source and the sample is placed immediately beyond this, within the
distance comparable to the wavelength so that near-field conditions are maintained. The
improved resolution is again traded against a large decrease in signal to noise because the
aperture strongly attenuates the signal intensity with a radial dependence (Mittleman et al
1999, Mitrofanov et al 2000b). The resulting THz pulse is also high pass filtered so that the
spectral content of the transmitted pulse is blueshifted compared to the pulse incident on the
aperture. Wynne et al (1999) quote improvements in spatial resolution of better than a quarter
of the wavelength for wavelengths around 1 mm and Hunsche et al (1998) quote values close
Topical Review
R77
to 70 µm, determined by using a line spread function analysis across a knife blade. Chen
et al (2000) used a dynamic aperture and achieved values of 50 µm; however, this approach
is limited to semiconductor surfaces.
An alternative approach to near-field imaging is to use the detector to measure the
evanescent field emanating from the sample (Mitrofanov et al 2000a, 2001b). To compensate
for strong signal attenuation the near-field detector is optimized by the integration of the
gallium arsenide photoconducting detector and the aperture in very close proximity so that
the near-field signal can be detected with sufficient sensitivity (Mitrofanov et al 2001a). With
this technique the spatial resolution is independent of THz wavelength and is scaled with
aperture size. Spatial resolutions (based on the 10–90% change in transmittance for a scanned
gold edge) of 39 µm for a 30 µm aperture (Mitrofanov et al 2000a) and 7 µm for a 5 µm
aperture (Mitrofanov et al 2001b) have been demonstrated. The limit of aperture dimensions,
given the present system powers and noise levels, is about 3–4 µm due to the d 3 attenuation
in signal by the aperture of diameter d. It is also possible that multiple reflections can occur
between the gold screen layer of the detector and the back surface of the sample that will affect
the measured time domain signal.
For many in vivo medical applications reflection geometry will be required to form images
of the structure below the surface of the body. The time delays of reflected pulses can be used
to calculate the depth of interfaces, in a manner similar to ultrasonic imaging (Mittleman et al
1997, 1999). An extension of this technique using multiple receivers and a Kirchoff migration
image reconstruction technique, originally developed for geophysical prospecting, enables
extended objects that are not parallel to the surface to be localized (Dorney et al 2001). The
vertical resolution of both these methods is limited by the bandwidth of the incident THz pulse
since it follows from the Rayleigh criterion that thin layers cannot be resolved if the separation
is much less than the coherence length of the light pulse. For typical THz bandwidths the
limit of depth resolution is close to 100 µm. Depth resolution can be improved well beyond
this limit by using interferometric techniques, in analogy to optical coherence tomography
(Johnson et al 2001a). Using a Michelson interferometer and exploiting the Gouy phase shift
arising from the lens in the sample arm, depth resolutions close to 10 µm have been achieved
by Johnson et al (2001b).
3.6. Safety
THz imaging systems used in vivo should comply with existing safety guidelines for exposure
to electromagnetic radiation. The guidelines for pulsed EM radiation are given in terms of
the maximum permissible exposure (MPE). This is a measure of energy relating to the entire
radiation train, calculated using root mean square and peak electric and magnetic fields, or
plane wave power densities, depending on the frequency range (American National Standards
Institute 2000, ICNIRP 1996, Ahlbom et al 1998, Institute of Electrical and Electronic
Engineers 1999). Several experimental factors affect the MPE value and must be considered
in the safety analysis of a system. These include the pulse repetition frequency, F; the
total exposure duration of a region of skin, T; the total number of pulses in time T, N; the
cross-sectional area of the incident beam, A; the time duration of a single pulse t (American
National Standards Institute 2000). The method recommended for pulsed laser safety analysis
(American National Standards Institute 2000) has been applied for a typical compact TPI
system using electro-optical or photoconductive antenna sources, by Berry et al (2002). It
was shown that the limiting average power per pulse, MPEPW is given as
MPEPW =
A × MPECW
Ft
R78
Topical Review
where MPECW is the continuous exposure limit of 1000 W m−2 for an exposure area of under
0.01 m2 (American National Standards Institute 2000). Typical values of the parameters were
used (F = 82 MHz; t = 10−13 s; A = 7.8 × 10−7 m2 (resulting from a circular beam of 1 mm
diameter); T = 600 s) to give a calculated MPEPW = 94 W. Presently, the average powers
of terahertz imaging systems using electro-optical or photoconductive generation are below
1 µW per pulse. If amplification is used this can be increased to 1 mW (Cole et al 2001,
Loffler et al 2000). Hence current systems have a power output per pulse which lies well
below the limit set by the guidelines.
It should be noted, however, that the safety guidelines on which the analysis was
based, were derived from experiments performed at shorter wavelengths (less than 10.6 µm)
and using longer pulse durations (over 1.4 ns) than those used in TPI (ICNIRP 1996,
Sliney and Wolbarsht 1980). There are far stronger absorption peaks for the water molecule in
the THz region than at shorter wavelengths, which could lead to more tissue damage. Similarly,
the ultrashort pulses used in TPI have the potential for thermomechanical effects that would
not have occurred with nanosecond pulses. Damage mechanisms induced by THz radiation
are being investigated by the THz BRIDGE Project (www.frascati.enea.it/THz-BRIDGE/).
4. Potential applications for THz imaging in medicine
Researchers have been guided by three properties of THz radiation when seeking the first
in vivo applications for the imaging technique. Firstly, there is a limited penetration depth of
the radiation which has led to consideration of easily accessible parts of the body, in particular
the skin and teeth (Arnone et al 1999). Secondly, as outlined in section 2, THz attenuation is
sensitive to water content and this has led to interest in the investigation of tissue hydration
(Arnone et al 1999, Mittleman et al 1999). Finally, the prospect of spectroscopic signatures
from tissue in different states of health and disease has encouraged work on wound healing
(Mittleman et al 1996, 1999) and also in the characterization of tissues that are not, at present,
accessible to in vivo imaging (Chen et al 1999). It is hoped that terahertz technology could
replace biopsy sampling with subsequent microscopy, in some areas.
The existing alternative measurement and imaging technologies vary slightly from
application to application, but include high-frequency ultrasound, magnetic resonance imaging
and optical and infra-red techniques.
4.1. Dermatology
The skin constitutes one of the largest organs of the body and is accessible for reflection
imaging. It is composed of several layers, principally the epidermis, dermis and hypodermis
(or sub-cutaneous tissues). This layered structure is attractive for reflection imaging as there
are interfaces between materials with different physical properties that cause reflection of THz
pulses. Measurements on model human dermis (Arnone et al 1999) have shown that the skin
constituents could be differentiated by TPI. Characteristic changes are seen in images of the
skin using ultrasound or MRI in the presence of skin diseases such as psoriasis where there is
thickening of the epidermis (El Gammal et al 1999). Changes induced by treatment may be
monitored. The potential offered by TPI is for a measurement technique that is cheaper and
more compact than MRI and does not require contact with the skin as does ultrasound.
Surface wounds may arise from burns or trauma, whilst chronic, non-healing wounds
can result from venous leg ulcers, pressure sores and diabetic ulcers. Wound healing has
three distinct phases. The first stage is haemostasis and includes the inward movement of
inflammatory cells. The second phase is the formation of granulation tissue and the third
Topical Review
R79
is the remodelling of that granulation tissue causing contraction of the wound. In the later
phases there is a need to test the therapeutic agents designed to promote healing. Wound
healing measurement in the second phase has been performed with estimates made of the
relative surface area of unhealthy necrotic and slough tissue and of healthy granulation tissue
(Mekkes and Westerhof 1995). Sometimes an estimate is made of wound volume, by filling
with a casting material or water, or more recently by non-invasive techniques using range
imaging methods (Plassmann and Jones 1998). THz imaging may be able to provide similar
measures, but with the advantages of being non-invasive and without the need to remove
dressings. The promise of TPI in this field has been demonstrated by work differentiating
healthy and burned chicken skin (Mittleman et al 1996, 1999).
Repeated measurements of wound area in the third phase of healing may give a measure
of the speed of response to treatment, but they do not differentiate between wounds that heal
with and without scar tissue formation. THz imaging offers the prospect of a method that can
differentiate scar tissue from healthy tissue, and thus a means of evaluating drugs designed to
inhibit scar formation.
Surface optical imaging techniques (Claridge et al 1992, Kenet 1995) are increasingly
used to differentiate between benign and malignant lesions from their characteristic
morphology. However, the depth of penetration is also important, because cancers that
have not yet penetrated the dermis will not have had the opportunity to metastasize to other
areas of the body. Reflection TPI is expected to provide a measure of depth, assuming
knowledge of the lesion absorption properties. However, this is an application for which
there are several competing technologies. Optical coherence tomography (OCT) uses light in
the visible and infra-red regions (Fujimoto 1999) and produces cross-sectional images with
a resolution of about 10 µm for a penetration depth of 2–3 mm. The images are analogous
to B-mode ultrasound images. The proposed applications for OCT (Welzel et al 1997)
are very similar to those for TPI, but TPI is expected to perform better because of reduced
scattering and a comparable resolution has been obtained using interferometry (Johnson et al
2001a). Preliminary in vitro and in vivo OCT work has shown promise for the detection of early
neoplastic changes in cancers of the gastrointestinal tract and prostate (D’Amico et al 2000).
Microscopic techniques include epiluminescence microscopy and confocal microscopy.
Confocal microscopy allows cross-sectional imaging of planes parallel to the skin surface,
in vivo (Webb 1996) and three-dimensional reconstruction can be used to generate sections
in the more familiar vertical direction used in histological slides (Masters et al 1997). Nearinfra-red wavelengths are preferred to optical wavelengths because of the lower scattering
and can achieve lateral and axial resolutions of 0.5 µm and 2 µm respectively within a small
sampling volume, but only to a depth of 0.5 mm. In contrast, experiments have shown that TPI
penetrates to at least 1.5 mm in a model dermis system suspended in bovine calf serum (Bezant
2000) and is associated with even less scattering (Han et al 2000). Non-optical techniques
under investigation for skin imaging include MRI (Maurer et al 1998) and ultrasonography
(Cammarota et al 1998). Ultrasonography has proved valuable in assessing skin tumours by
their composition and extent, and early work suggests that THz imaging has similar capabilities
(Woodward et al 2001). The potential of TPI, however, to provide both a structural image and
spectroscopic information means that it may be able to complement the alternative techniques.
For example, ultrasound methods cannot distinguish between benign and malignant lesions
(Fornage et al 1993) because naevi can be very thin and not identifiable with ultrasound
probes and it is sometimes hard to differentiate the tumour itself from inflammatory changes
induced by the tumour. It has been shown (Woodward et al 2001) that TPI can be used
to distinguish between normal skin and basal cell carcinomas in vitro. Initial work on an
optical method involving a model of colour formation within the skin (Cotton et al 1997) has
R80
Topical Review
shown promise for demonstrating skin architecture. These techniques may also be valuable
at terahertz frequencies, alone or in combination with the optical method. The water content
of the stratum corneum can affect the physical characteristics of the skin, including its barrier
function, mechanical properties and drug penetration (Berardesca and Borroni 1995). Skin
hydration is also measured when testing the efficacy of cosmetics (Szayna and Kuhn 1998).
Alternative methods used for measuring hydration were surveyed by Berardesca and Borroni
(1995) and this is a very active area of research. The main imaging contender is magnetic
resonance imaging (MRI) and promising results have been reported (Richard et al 1993,
Song et al 1997). However, MRI is expensive and special coils are needed to achieve the
spatial resolution required for imaging skin, so there is a role for TPI. It is known that Fourier
transform infra-red spectroscopy (FTIR) provides valuable spectroscopic information (Stuart
1997) and reduced scattering at THz frequencies may enhance these results.
4.2. Dentistry
Like the skin, teeth are accessible for terahertz imaging. Rather fewer alternative imaging
techniques are used in dentistry than in dermatology, primarily radiography (both conventional
and digital) and more recently optical coherence tomography. Caries formation begins beneath
the surface of the dental enamel, and lesions may grow into the dentine layer of the tooth with
no visible changes on the enamel surface. Imaging with x-rays is insensitive for the detection
of primary caries (Ciesla et al 2000, Douglass et al 1986) and is unsuited to screening for
early dental caries because of the associated exposure to ionizing radiation. If caries could
be detected while still in the enamel, therapy could be used to reverse the process of decay
before filling was necessary. It has been shown (Ciesla et al 2000) in vitro, that TPI can detect
the presence of early caries in the enamel layers of human teeth, and distinguish these from
naturally occurring regions of lower mineralization. Variation in water content with disease
in periodontal soft tissue might also be measured. This is an extremely promising application
for TPI, particularly as the depth of penetration is significantly larger than that possible with
OCT. It may be possible to extend techniques that have been used successfully in OCT, such as
delivery via fibre optic probes, to THz frequencies (Baumgartner et al 2000, Otis et al 2000).
4.3. Other applications
Only a few studies using TPI have been undertaken to date using tissues other than skin and
teeth. Encouraging images have been produced using 2 mm thick slices of pork and chicken
that (Arnone et al 1999) showed differentiation between areas of muscle, fat and kidney.
Measurements from an excised human breast tumour have been reported (Chen et al 1999).
As interest in the technique grows, further studies are underway. In particular, two projects have
been supported by the European Commission to measure the terahertz properties of samples
ranging from DNA to tissue specimens (www.teravision.org, www.frascati.enea.it/THzBRIDGE/).
5. Developments
TPI is in the very earliest stages of its development, but it is possible to suggest how the
field may mature. Endoscopic imaging by fibre optic delivery would allow access to tissues
away from the surface of the body. The first steps towards this goal have been made with
the development of waveguides for THz radiation (Gallot et al 2000) and with the use of fibre
optics in THz delivery (Johnson et al 2000). There is potential to enhance the performance of
Topical Review
R81
conventional excisional biopsies by using terahertz image and spectroscopic guidance to select
biopsy sites. It may be possible to replace conventional biopsy by in situ spectroscopic analysis
and to guide interventional procedures using terahertz technology. Intravascular imaging offers
the promise of atherosclerotic plaque characterization, which is of great interest in the carotid
artery. However, considerable technological developments are required for THz techniques
to be used in the coronary arteries, where a resolution of 30–100 µm is necessary to resolve
the layered structure of the vessel wall (Foster et al 2000). More rapid data acquisition opens
up possibilities associated with the assessment of function and physiological processes, which
can be of value in monitoring drug efficacy.
Although for many potential applications there are a number of alternative technologies,
none has yet emerged as the ideal tool. It is inevitable that TPI will be added to the
clinician’s armamentarium and be valued for the complementary information it can provide.
Multimodality developments may provide the greatest advances, by adding to existing
techniques this new compact, non-contact imaging tool that is free from concerns about
ionizing radiation hazards and making use of its potential for spectroscopic analysis coupled
with low levels of scattering.
Acknowledgments
Many thanks are due to Karsten Siebert and Torsten Loffler (Physikalisches Institut der
J W Goethe-Universitaet) for THz measurements of the step wedge, to Stephen Russell for
supplying the alginate wound dressing, to Andrew Marsden for manufacture of the sintered
step wedge and to Michael Stone for manufacture of the gold on TPX MTF test object. Support
from the European Commission (IST-1999-10154, Teravision) and Engineering and Physical
Sciences Research Council (GR/N39678) is gratefully acknowledged.
References
Ahlbom A et al 1998 Guidelines for limiting exposure to time-varying electric, magnetic, and electromagnetic fields
(up to 300 GHz) Health Phys. 74 494–522
American National Standards Institute 2000 American National Standard for Safe Use of Lasers (ANSI Z136.1—2000)
(Orlando, FL: Laser Institute of America)
Arnone D D, Ciesla C M, Corchia A, Egusa S, Pepper M, Chamberlain J M, Bezant C, Linfield E H, Clothier R and
Khammo N 1999 Applications of terahertz (THz) technology to medical imaging Proc. SPIE 3828 209–19
Auston D H, Cheung K P and Smith P R 1984 Picosecond photoconducting Herzian dipoles Appl. Phys. Lett. 45
211–2
Auston D H and Nuss M C 1988 Electrooptic generation and detection of femtosecond electrical transients IEEE J.
Quantum Electron. 24 184–97
Baumgartner A, Dichtl S, Hitzenberger C K, Sattmann H, Robl B, Moritz A, Fercher Z F and Sperr W 2000
Polarization-sensitive optical coherence tomography of dental structures Caries Res. 34 59–69
Berardesca E and Borroni G 1995 Instrumental evaluation of cutaneous hydration Clinics Dermatol. 13 323–7
Berry E, Brown J M, Connell M, Craven C M, Efford N D, Radjenovic A and Smith M A 1997 Preliminary experience
with medical applications of rapid prototyping by selective laser sintering Med. Eng. Phys. 19 90–6
Berry E, Walker G C, Fitzgerald A J, Chamberlain J M, Smye S W, Miles R E and Smith M A 2002 Do in vivo
Terahertz imaging systems comply with safety guidelines? J. Laser Safety Appl. submitted
Bezant C D 2000 Application of THz pulses in semiconductor relaxation and biomedical Imaging Studies PhD Thesis
University of Nottingham
Born M and Wolf E 1959 Principles of Optics (London: Pergamon)
Brown E R, Verghese S and McIntosh K A 1997 Terahertz photomixing in low-temperature-grown GaAs LEOS ’97:
Conf. Proc. Lasers and Electro-Optics Society Annual Meeting vol 2, p 363
Brucherseifer M, Nagel M, Haring Bolivar P and Kurz H 2000 Label-free probing of the binding state of DNA by
time-domain terahertz sensing Appl. Phys. Lett. 77 4049–51
R82
Topical Review
Cammarota T, Pinto F, Magliaro A and Sarno A 1998 Current uses of diagnostic high-frequency US in dermatology
Eur. J. Radiol. 27 (suppl) S215–S223
Chen Q, Jiang Z, Xu G X and Zhang X-C 2000 Near-field terahertz imaging with a dynamic aperture Opt. Lett. 25
1122–4
Chen Q, Jiang Z P and Zhang X C 1999 All-optical THz imaging Proc. SPIE 3617 98–105
Ciesla C M, Arnone D D, Corchia A, Crawley D, Longbottom C, Linfield E H and Pepper M 2000 Biomedical
applications of terahertz pulse imaging Proc. SPIE 3934 73–81
Claridge E, Hall P N, Keefe M and Allen J P 1992 Shape analysis for classification of malignant melanoma J. Biomed.
Eng. 14 229–34
Cole B, Woodward R M, Crawley D, Wallace V P, Arnone D D and Pepper M 2001 Terahertz imaging and
spectroscopy of human skin, in-vivo Commercial and Biological Applications of Ultrashort Pulse Lasers:
Laser Plasma Generation and Diagnostics 4276 1–10
Cotton S, Claridge E and Hall P 1997 Noninvasive skin imaging Inf. Processing Med. Imaging 1230 501–6
D’Amico A V, Weinstein M, Li X, Richie J P and Fujimoto J 2000 Optical coherence tomography as a method for
identifying benign and malignant microscopic structures in the prostate gland Urology 55 783–7
Dorney T D, Johnson J L, van Rudd J, Baraniuk R G, Symes W W and Mittleman D M 2001 Terahertz reflection
imaging using Kirchoff migration Opt. Lett. 26 1513–5
Douglass C W, Valachovic R W, Wijesinha A, Chauncey H H and Kapur K K 1986 Clinical efficacy of dental
radiography in the detection of dental caries and periodontal diseases Oral Surg. Oral Med. Oral Pathol. 62
330–9
Eisle H 1999 Two-terminal devices as fundamental solid-state terahertz oscillators Proc. SPIE 3828 70–80
El Gammal S, El Gammal C, Kaspar K, Pieck C, Altmeyer P, Vogt M and Ermert H 1999 Sonography of the skin at
100 MHz enables in vivo visualization of stratum corneum and viable epidermis in palmar skin and psoriatic
plaques J. Invest. Dermatol. 113 821–9
Fornage B D, McGavran M H, Duvic M and Waldron C A 1993 Imaging of the skin with 20-MHz US Radiology 189
69–76
Foster F S, Pavlin C J, Harasiewicz K A, Christopher D A and Turnbull D H 2000 Advances in ultrasound
biomicroscopy Ultrasound Med. Biol. 26 1–27
Fujimoto J G 1999 Biomedical imaging using optical coherence tomography Frontiers of Engineering: Reports on
Leading Edge Engineering from the 1998 NAE Symposium on Frontiers of Engineering (The National Academy
of Engineering) 12–23
Gallot G, Jamison S P, McGowan R W and Grischkowsky D 2000 Terahertz waveguides J. Opt. Soc. Am. 17 851–63
Greene B I, Saeta P N, Dykaar D R, Schmitt-Rink S and Chuang S L 1992 Far-infrared light
generation at semiconductor surfaces and its spectroscopic applications IEEE J. Quantum Electron. 28
2302–12
Grischkowsky D, Keiding S, van Exter M and Fattinger C 1990 Far-infrared time-domain spectroscopy with Terahertz
beams of dielectrics and semiconductors J. Opt. Soc. Am. B Opt. Phys. 7 2006–15
Han P Y, Cho G C and Zhang X-C 2000 Time-domain transillumination of biological tissues with terahertz pulses
Opt. Lett. 25 242–4
Hu B B and Nuss M C 1995 Imaging with terahertz waves Opt. Lett. 20 1716–8
Hunsche S, Cai Y, Brener I, Nuss M C and Wynn J 1998 Near-field time-resolved imaging with far-infrared dipole
sources Lasers and Electro-Optics, 1998. CLEO 98: Technical Digest pp 64–5
Hunsche S, Koch M, Brener I and Nuss M C 1997 Near-field THz optics LEOS ’97: Conf. Proc., Lasers and
Electro-Optics Society Annual Meeting vol 2, pp 368–9
ICNIRP 1996 International Commission on Non-Ionizing Radiation Protection: Guidelines on limits of exposure to
laser radiation of wavelengths between 180 nm and 1000 nm Health Phys. 71 804–19
Institute of Electrical and Electronic Engineers. 1999 Standard for Safety Levels with Respect to Human Exposure
to Radiofrequency Electromagnetic Fields, 3 KHz to 300 GHz (New York: Institute of Electrical and Electronic
Engineers)
Jacobsen R H, Mittleman D M and Nuss M C 1996 Chemical recognition of gases and gas mixtures with terahertz
waves Opt. Lett. 21 2011–3
Jiang Z and Zhang X-C 1999a Terahertz imaging via electrooptic effect IEEE Trans. Microwave Theory Tech.
47 2644–50
Jiang Z and Zhang X-C 1999b THz imaging via electro-optic effect Microwave Symposium Digest, 1999 IEEE MTT-S
International 3 941–4
Jiang Z and Zhang X-C 2001 Electro-optic measurement of THz field pulses with a chirped optical beam Appl. Phys.
Lett. 72 1945–7
Johnson J L, Dorney T D, Rudd J V and Mittleman D M 2000 Terahertz reflectometry: a model system for the inverse
problem LEOS: Proc. Laser and Electro-optics Society, 13th Annual Meeting vol 1, p 242
Topical Review
R83
Johnson J L, Dorney T D and Mittleman D M 2001a Enhanced depth resolution in terahertz imaging using phase-shift
interferometry Appl. Phys. Lett. 78 835–7
Johnson J L, Dorney T D and Mittleman D M 2001b Interferometric imaging with terahertz pulses IEEE J. Selected
Topics Quantum Electron. 7 592–9
Kenet R D 1995 Digital imaging in dermatology Clinics Dermatol. 13 381–92
Kindt J T and Schmuttenmaer C A 1996 Far-infrared dielectric properties of polar liquids probed by femtosecond
terahertz pulse spectroscopy J. Phys. Chem. 100 10373–9
Lasch P and Naumann D 1998 FT-IR microspectroscopic imaging of human carcinoma thin sections based on pattern
recognition techniques Cell. Mol. Biol. 44 189–202
Loffler T, Jacob F and Roskos H G 2000 Generation of terahertz pulses by photoionization of electrically biased air
Appl. Phys. Lett. 77 453–5
Markelz A G, Roitberg A and Heilweil E J 2000 Pulsed terahertz spectroscopy of DNA, bovine serum albumin and
collagen between 0.1 and 2.0 THz Chem. Phys. Lett. 320 42–8
Masters B R, Gonnord G and Corcuff P 1997 Three-dimensional microscopic biopsy of in vivo human skin: a new
technique based on a flexible confocal microscope J. Microsc. 185 329–38
Maurer J, Knollmann F D, Hell B, Radke C and Felix R 1998 Delineation of maxillofacial cutaneous lesions by means
of high-resolution magnetic resonance imaging Cancer Detect. Prev. 22 540–3
Mekkes J R and Westerhof W 1995 Image processing in the study of wound healing Clinics Dermatol. 13 401–7
Mitrofanov O, Brener I, Harel R, Wynn J D, Pfeiffer L N, West K W and Federici J F 2000a Terahertz near-field
microscopy based on a collection mode detector Appl. Phys. Lett. 77 3496–8
Mitrofanov O, Brener I, Wanke M C, Ruel R R, Wynn J D, Bruce A J and Federici J F 2000b Near-field microscope
probe for far infrared time domain measurements Appl. Phys. Lett. 77 591–3
Mitrofanov O, Harel R, Lee M, Pfeiffer L N, West K W, Wynn J D and Federici J F 2001a Study of single-cycle pulse
propagation inside a terahertz near-field probe Appl. Phys. Lett. 78 252–4
Mitrofanov O, Lee M, Hsu J W P, Brener I, Harel R, Federici J F, Wynn J D, Pfeiffer L N and West K W 2001b
Collection-mode near-field imaging with 0.5-THz pulses IEEE J. Selected Topics Quantum Electron. 7 600–7
Mittleman D M, Gupta M, Neelamani R, Baraniuk R G, Rudd J V and Koch M 1999 Recent advances in terahertz
imaging Appl. Phys. B 68 1085–94
Mittleman D M, Hunsche S, Boivin L and Nuss M C 1997 T-ray tomography Opt. Lett. 22 904–6
Mittleman D M, Jacobsen R H and Nuss M C 1996 T-ray imaging IEEE J. Selected Topics Quantum Electron.
2 679–92
Nuss M C 1996 Chemistry is right for T-ray imaging IEEE Circuits and Devices Magazine 12 25–30
Otis L L, Colston B W, Everett M J and Nathel H 2000 Dental optical coherence tomography: a comparison of two
in vitro systems Dentomaxillofac. Radiol. 29 85–9
Pedersen J E and Keiding S 1992 THz time domain spectroscopy of nonpolar liquids IEEE J. Quantum Electron.
28 2518–22
Plassmann P and Jones T D 1998 MAVIS: a non-invasive instrument to measure area and volume of wounds
Med. Eng. Phys. 20 332–8
Richard S, Querleux B, Bittoun J, Jolivet O, Idy-Peretti I, de Lacharriere O and Leveque J L 1993 Characterization
of the skin in vivo by high resolution magnetic resonance imaging: water behavior and age-related effects
J. Invest. Dermatol. 100 705–9
Seibe F, Siebert K, Leonhardt R and Roskos H G 1999 A fully tunable dual-color CW Ti:Al203 Laser IEEE
J. Quantum Electron. 35 1731–6
Sliney D H and Wolbarsht M L 1980 Laser Safety Standards: Evolution and Rationale. Safety with Lasers and Optical
Sources (New York: Plenum)
Smye S W, Chamberlain J M, Fitzgerald A J and Berry E 2001 The interaction between terahertz radiation and
biological tissue Phys. Med. Biol. 46 R101–R112
Song H K, Wehrli F W and Ma J 1997 In vivo MR microscopy of the human skin Magn. Reson. Med. 37 185–91
Stuart B 1997 Biological Applications of Infrared Spectroscopy (Chichester: Wiley)
Szayna M and Kuhn W 1998 In vivo and in vitro investigations of hydration effects of beauty care products by
high-field MRI and NMR microscopy J. Eur. Acad. Dermatol. Venereol. 11 122–8
Thrane L, Jacobsen R H, Jepsen P U and Keiding S R 1995 THz reflection spectroscopy of liquid water Chem. Phys.
Lett. 240 330–3
van Exter M, Fattinger C and Grischkowsky D 1989 High-brightness terahertz beams characterized with an ultrafast
detector Appl. Phys. Lett. 55 337–9
van Exter M and Grischkowsky D 2001 Carrier dynamics of electrons and holes in moderately doped silicon
Phys. Rev. B 41 12140–9
Verghese S, McIntosh K A, Calawa S, Dintale W F, Duerr E K and Molvar K A 1998 Generation and detection of
coherent terahertz waves using two photomixers Appl. Phys. Lett. 73 3824–6
R84
Topical Review
Walther M, Fischer B, Schall M, Helm H and Uhd Jepsen P 2000 Infrared vibrational spectra of all-trans, 9-cis and
13-cis retinal measured by THz time domain spectroscopy Chem. Phys. Lett. 332 389–95
Webb R H 1996 Confocal optical microscopy Reports on Progress in Physics 59 427–71
Welzel J, Lankenau E, Birngruber R and Engelhardt R 1997 Optical coherence tomography of the human skin J. Am.
Acad. Dermatol. 37 958–63
Woodward R M, Cole B, Wallace V P, Arnone D D, Pye R, Linfield E H, Pepper M and Davies A G 2001 Terahertz pulse
imaging of in-vitro basal cell carcinoma samples CLEO, Europe: Conf. Lasers and Electro-Optics, Technical
Digest 2001 pp 329–30
Wu Q, Hewitt T D and Zhang X C 1996 Two-dimensional electro-optic imaging of THz beams Appl. Phys. Lett. 69
1026–8
Wu Q and Zhang X-C 1996 Design and characterisation of travelling-wave electro-optic terahertz sensors IEEE
J. Selected Topics Quantum Electron. 2 693–700
Wynne K, Carey J C, Zawadzka J and Jaroszynski D A 1999 Superluminal propagation of terahertz pulses in
sub-wavelength structures CLEO ’99: Lasers and Electro-Optics p 397
Zhang X C, Jin Y and Ma X F 1992 Coherent measurements of THz optical rectification from electro-optical crystals
Appl. Phys. Lett. 61 2764
Zhao G, Schouten R N, van der Valk N, Wenckebach W T and Planken P C M 2002 Design and performance of a
THz emission and detection setup based on a semi-insulating GaAs emitter Rev. Sci. Instrum. submitted