THz Medical Imaging Using Broadband Direct Detection

THz Medical Imaging Using Broadband Direct Detection
Zachary D. Taylora,*, James Garritanoa, Priyamvada Tewaria, Eric Dieboldc, Jun Sungc, Neha
Bajwaa, Bryan Nowroozib, Alexander Stojadinovicd,e, Nuria Llombartf, Elliott Browng, Warren
Grundfesta,b,c
a
Dept. of Bioengineering, UCLA, 420 Westwood Plaza, Los Angeles, CA, USA 90095
b
Dept. of Surgery, UCLA, 200 UCLA Medical Plaza, Los Angeles, CA, USA 90095
c
Dept. of Electrical Engineering, UCLA, 420 Westwood Plaza, Los Angeles, CA, USA 90095
d
Dept. of Surgery, Walter Reed Army Medical Center, Washington, DC 20307 USA
e
Combat Wound Initiative Program, Washington, DC 20307 USA
f
Delft University of Technology, Mekelweg 4, 2628 CD, Delft, The Netherlands
g
Wright State University, 3640 Colonel Glenn Hwy, Dayton, OH 45435
from Proceedings, Society of Photo-Optical Instrumentation Engineers (SPIE)
SPIE Conference Paper#8624-02
Presented at Photonics West 2013
San Francisco, CA
5 February 2013
THz Medical Imaging Using Broadband Direct Detection
Zachary D. Taylora,*, James Garritanoa, Priyamvada Tewaria, Eric Dieboldc, Jun Sungc, Neha
Bajwaa, Bryan Nowroozib, Alexander Stojadinovicd,e, Nuria Llombartf, Elliott Browng, Warren
Grundfesta,b,c
a
Dept. of Bioengineering, UCLA, 420 Westwood Plaza, Los Angeles, CA, USA 90095
b
Dept. of Surgery, UCLA, 200 UCLA Medical Plaza, Los Angeles, CA, USA 90095
c
Dept. of Electrical Engineering, UCLA, 420 Westwood Plaza, Los Angeles, CA, USA 90095
d
Dept. of Surgery, Walter Reed Army Medical Center, Washington, DC 20307 USA
e
Combat Wound Initiative Program, Washington, DC 20307 USA
f
Delft University of Technology, Mekelweg 4, 2628 CD, Delft, The Netherlands
g
Wright State University, 3640 Colonel Glenn Hwy, Dayton, OH 45435
ABSTRACT
Research in THz imaging is generally focused on three primary application areas: medical, security, and nondestructive
evaluation (NDE). While work in THz security imaging and personnel screening is populated by a number of different
active and passive system architectures, research in medical imaging in is generally performed with THz time-domain
systems. These systems typically employ photoconductive or electrooptic source/detector pairs and can acquire depth
resolved data or spectrally resolved pixels by synchronously sampling the electric field of the transmitted/reflected
waveform. While time-domain is a very powerful scientific technique, results reported in the literature suggest that
desired THz contrast in medical imaging may not require the volume of data accessible from time-resolved
measurements and that simpler direct detection techniques may be sufficient for specific applications. In this talk we
discuss a direct detection system architecture operating at a center frequency of ~ 525 GHz that uses a photoconductive
source and schottky diode detector. This design takes advantage or radar-like pulse rectification and novel reflective
optical design to achieve high target imaging contrast with significant potential for high speed acquisition time. Results
in spatially resolved hydration mapping of burn wounds are presented and future outlooks discussed.
Keywords: THz medical imaging, THz, Burn, Hydration
1. INTRODUCTION
The terahertz (THz) region is located between the millimeter-wave and the far-IR bands and is most broadly defined to
cover the spectrum from 100 gigahertz (GHz) to 10 THz (3 mm to 0.03 mm) ) [1]. Because of its relatively short
wavelength, non-ionizing photon energy, high sensitivity to tissue hydration changes, and good robustness to scattering,
THz imaging has been proposed as a modality for a variety of medical imaging applications including cancer detection
[2, 3], burn imaging [4-6], and corneal hydration sensing [7, 8]. In many of the applications under exploration, factors
such as penetration depth and axial resolution are limited by tissue hydration which constrains the utility of axially or
spectrally resolved features. Furthermore, many of these experimental explorations have been explored with in vitro or
ex vivo samples thus limiting the clinical relevance of the results.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
2. BACKGROUND
A good evaluation of the awareness of the medical community to an emerging diagnostic imaging technology is research
dollars allocated to the technology by the NIH. The first funds allocated by the NIH to THz medical technology was a
K08 grant awarded to Dr. Siegel’s group at JPL in 2002 . This data point is represented by the first bar in the in the bar
graph in Figure 1a and coincides with the establishment of the National Institute of Biomedical Imaging and
Bioengineering (NIBIB, December 29, 2000) Data from the NIH RePORTs system suggest that the increase in funding
over the last decade was aided significantly by funds from this institute. The first THz medical imaging results reported
in the literature were published in 1999 nearly simultaneously by Mittleman et al. [9]. and Arnone et al. [10]. Mittleman
reported on the utility of THz imaging to evaluate the hydration of a burn wound induced by a high powered laser in ex
vivo chicken skin while Arnone reported preliminary results on the feasibility of THz to delineate dental caries based on
differences in absorption.
(a)
(b)
Figure 1: Estimated (a) NIH funding and (b) NSF funding for THz technology
A bar graph representing the estimated total NIH funding to date is displayed in Figure 1a and was generated using the
NIH RePORTS public database [11] and filtering search results with the appropriate keywords (THz, Terahertz, etc...)
Following filter the title and abstract of each proposal was read to determine the relevancy of each project to THz
medical imaging. This methodology is susceptible to error and we emphasize again that Figure 1a is an estimate. In
terms of quantity, the majority of projects are post doc training grants (K08) or subcontracts of center grants whose main
focus appear to be subjects that are compatible with THz sensing. The first R-series appear in the 2008/2009 time frame.
Topics include skin cancer spectroscopy and imaging, liver cancer spectroscopy and imaging, corneal hydration sensing,
burn imaging, and general instrument development. While medical imaging has just finished its first decade of NIH
funding, the field has received decades of funding from the NSF. An estimate of the total NSF awards for THz research
(of any kind) is provided Figure 1b over the same period as Figure 1a. THz imaging technology appears to be following
the same funding trend that many other imaging modalities have followed and is poised to make significant gains in
medical funding and applications development over the next decade.
3. THZ IMAGING SYSTEM METHODOLOGIES AND COMPONENTS
3.1 Time domain Spectroscopy and Imaging
Time-domain spectroscopy (TDS) and Time-domain imaging (TDI) techniques utilize a combination of ultrafast
photocondutors and optoelectronics to generate and detect pulsed THz radiation [12]. In a typical time domain set-up
(Figure 2a), a photoconductive switch is illuminated by femtosecond (fs) laser excitation and generates broadband
current pulses in the underlying ultrafast material. The current is then coupled to a broadband, planar antenna and power
is radiated into free space with > 1 of THz bandwidth [12]. The pulsed THz illumination is re-directed and focused onto
a target of interest using metalic mirrors, and the transmitted or reflected beam is sampled with an electro optic crystal or
a photocoductive switch identical to the transmitter.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
An electro-optic configuration is displayed in Figure 2a. In this set-up a probe pulse, split from the initial fs laser,
effectively samples the THz pulse when passed through a electro-optic crystal via Pockels effect. The THz electric field
induces a birefringence in the material resulting in a polarization rotation in the femtosecond probe beam. By
mechanically tuning the optical delay line, a THz pulse waveform is optically sampled at a different transient point from
each of the repeated pulses, and the pulse shape can be traced. This methodology allows measurement of both amplitude
and the phase information from the reflected/transmitted information and one can extract complex permitivities of the
materials of interest [13].
(a)
(b)
(c)
Figure 2: Time domain phenomenology. (a) Block diagram of typical time domain system employing a
photoconductive source and electro-optic sampling. (b) Sampled time domain signal with pertinent parameters
labeled. (c) Representation of THz spectral data obtained through Fourier transform.
The pertinent parameters related to time domain data processing are displayed in Figure 2b and Figure 2c. The
parameters of spectrum extraction from time domain data are very similar to interferometry. The highest resolvable
frequency in the acquired power spectrum (f max in Figure 2c) is inversely proportional to the step size in the time domain
used to sample the THz pulse (Δt in Figure 2b). Similarly the frequency resolution (Δf in Figure 2c) is inversely
proportional to the total duration sampled in the time domain pulse (T in Figure 2b)
Image contrast can be generated directly from the time domain signal or a Fourier transform of the pulse can be
computed and contrast generated from the spectral content. The most common techniques are pulse maximum (Emax in
Figure 2b), Time Post Peak (TPP) (Emax/E(t1) in Figure 2b), Ratio of maximum to minimum amplitudes (Emax/Emin in
Figure 2b), and band integrated power which can be computed from the transformed spectra (Figure 2c).
While excellent imaging contrast has been generated by all of these methods in a range of applications, each contrast
generation technique requires a fully sampled waveform at each pixel in the field of view (FOV). More specifically, the
delay line driven sampling must be sufficient to resolve the temporal location of the reflected waveform peak and pulse
width as well as densely spaced enough to achieve the desired maximum frequency content. Furthermore a significant
duration of the reflected pulse must also be sampled to obtain sufficient frequency resolution. These requirements on
data acquisition necessitate stationary samples and increased acquisition times and can hinder clinical translation.
3.2 Pulsed Direct Detection
One interesting observation regarding contrast generation reported in the literature is the dearth of hyperspectral or
spectrally resolved imagery. This is likely due to the smooth, resonance free spectra of most tissues in vivo and ex vivo
[14]. Another observation is that three of the four most common generation methodologies are broad band
measurements that do not utilize the spectral resolving capability of Time-domain phenomenology and instead rely on
total integrated power or an inference of pulse dispersion (e.g. TPP). This suggests, that for a number of imaging and
sensing applications involving contrast generation in hydrated tissues, the volume of data available with time domain
measurements may not be necessary. Our group has thus decided to focus on broad band integrated power because it is
less data intensive than extracting reflected/transmitted pulse widths or measuring pulse dispersion. A block diagram of
the system is presented in Figure 3a.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
(a)
(b)
(c)
Figure 3: Direct detection phenomenology. (a) Block diagram of our direct detection phenomenology employing
a photoconductive source and zero-bias Schottky diode detector. (b) Time domain signals detailing the temporal
overlap of the rectified THz pulse with the gating pulse. (c) Representation of THz spectral content sed to
generate a pixel in the direct detection system. The pixel amplitude is proportional to the area under the curve.
The major differences between this system architecture and standard time domain are the detector and post detection
phenomenology. Our direct detector system utilizes a zero-bias Schottky diode detector mounted in a WR1.5
waveguide. The rectangular waveguide transverse mode distributions and detector E-field probe design limit the predetection bandwidth to the fundamental (TE10) mode of the waveguide yielding an operational band spanning from ~
400 GHz to 700 GHz. The output voltage of direct detector is therefore proportional to the total transmitted/reflected
power in the pre-detection band. Since the input signal is pulsed (femtosecond laser pumped photoconductive switch),
the output of the direct detector is also pulsed allowing for gated detection to improve SNR. The rectified THz pulse is
amplified and coupled to the RF port of a broad band mixer. A reference beam is generated from the femtosecond laser
output using a photodetector and coupled to the LO port of the broad band mixer. The reference pulse is amplified to
reach near the IP3 point of the RF and passed through a preset, fixed delay such that it arrives coincident with THz
pulse. The DC voltage generated by the beating of frequency components of the rectified THz pulse and reference pulse
are detected with an analog to digital converter (ADC).
This configuration is effectively a high speed, boxcar integrator where the reference pulse only turns on the mixer at the
arrival of the rectified THz pulse. Plots detailing the temporal and spectral operational principles of the system are
displayed in Figure 3b and Figure 3c, respectively. The red trace in Figure 3b represents the rectified pulse while the
shaded blue area represents the reference pulse used to drive the gating. Note that the reference pulse is significantly
broader in time than the rectified THz pulse. Amplifiers were selected to produce ~ 12 GHz of bandwidth on the
rectified THz pulse and ~ 3 GHz of bandwidth on the reference pulse. This broad gate reduces the sensitivity to pulse
arrival time synchronicity and results in significant robustness to changes in THz signal path length (or, equivalently,
signal round trip time) due to animal model curvature, breathing, twitching, etc. In other words, the fixed delay line does
not need to be reset even if the arrival time of the rectified THz pulse changes significantly with respect to THz pulse
width. This is somewhat equivalent to operating pulse detection at the zero-path case which is in stark contrast to time
domain imaging which must scan a delay line repeatedly at each pixel given the extremely broad instantaneous detection
bandwidth of the electro-optic or photoconductive sampling.
The red curve in Figure 3c represents the free space power spectral density (PSD) of the photoconductive switch and the
blue shaded area represents the approximate TE 10 operational band of the WR1.5 waveguide in which the detector is
mounted. Although the spectrum should be displayed as “sampled” with a frequency spacing of 20 MHz (pulse
repetition rate of the femtosecond fiber laser), it is represented as a continuum for comparison to time domain spectra
which are typically sampled in the 3 GHz – 10 GHz range. The overall operation of this receiver is to provide a voltage
output directly proportional to the total area under the curve of the band of interest (blue shaded region). The direct
detector performs this operation and the gated receiver provides significant linear gain while minimizing degradation to
the SNR.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
3.3 Photoconductive Switch Performance
A key to the success of our incoherent transceiver system is the power of the PC-switch transmitter in the frequency
band of interest. Unlike THz time domains systems that usually strive to create the greatest possible overall bandwidth,
well beyond 1 THz, our PC switch is designed to emit most of its power below 1 THz where atmospheric absorption and
scattering losses from rough surfaces (like skin burns) are tolerably low. The ultrafast PC material is ErAs:GaAs having
an electron-hole recombination time of ~0.2 ps. The PC switch lies at the geometric center of a self-complementary
square-spiral antenna as shown in the photograph of Figure 4(a). The square spiral antenna provides a higher sub-THz
driving-point resistance than log-spiral and other alternative wideband planar antennas commonly used at THz
frequencies. And so the power levels from a PC switch – average and peak – are considerably higher.
TDS THz Amplitude
TDS THz Amplitude with G25 polarizers horizontal
TDS THz Amplitude with G200 polarizers horizontal
10
88
FTIR
Amp[mV]
(mV)
TD THz
Signal
THz
6
44
2
TDS
00
-2
-4-4
-6
-8-8
-10
-5-5
00
55
10
10
15
15
20
20
25
25
30
30
Rel Time [ps]
Relative
Time [ps]
(a)
(b)
(c)
Figure 4: High efficiency photoconductive switch. (a) Micrograph of photoconductive switch and center
region of self-complementary square spiral antenna. (b) Time domain pulse acquired with ZnTe electro-optic
detection. (c) Power spectra of the photoconductive switch. The temporal pulse in (b) was transformed to
compute the free space power spectral density (TDS – blue trace). This is plotted against power spectral
density measurements of the same switch as acquired with an FTIR and LHe bolometer (FTIR – red trace).
To better quantify the PC-switch power levels, we have carried out separate measurements of the time- and frequencydomain behavior. The time-domain spectrometric (TDS) characterization was conducted by ZOmega Terahertz Corp
using the PC switch as the transmitter and a ZnTe electro-optic crystal as the receiver. The time-domain pulse, shown in
Figure 4(b) shows a significant amount of ringing at times even past 10 ps, consistent with having a majority of transmit
power below 1 THz. The Fourier transform of this pulse results in the power spectrum shown in Figure 4(c). The
frequency-domain measurements were carried out by Fourier transform infrared (FTIR) spectroscopy, yielding the
power spectrum also plotted in Figure 4(c). Notice that the TDS power spectrum rolls of much faster than the FTIR
spectrum, whicih is attributable in part to the roll-off of the ZnTe electro-optic efficiency. Another factor is that the
FTIR detector was a LHe bolometer, which generally displays increasing responsivity with frequency across the THz
region, and therefore makes measured power spectra roll-off artificially slow
4. UCLA THZ IMAGING SYSTEM
A CAD design of the UCLA THz optical subsystem is displayed in Figure 5a and Figure 5b. Details of the system and
its operation can be found here [14].
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
(a)
(b)
(c)
Figure 5: Pulsed THz medical imaging system. (a) CAD design of imaging system with integrated 780 nm optics
and THz optics. (b) CAD design of system optical subassembly. (c) Preliminary optical modeling using GRASP.
The zero-bais schottky diode rectifier is provided by Virginia Diodes and has been characterized by the manufacturer in
the band spanning ~ 500 GHz – 750 GHz. This detector was fabricated without back biasing protection circuitry and
thus its video bandwidth is limited only by device parasitic impedances and the RF coupling between the device and the
SMA port. The power spectrum of the amplified (BW: 10 MHz – 10 GHz, 40 dB gain, 2.5 dB noise figure), rectified
THz pulse was acquired with an HP RF spectrum analyzer and a 3 dB point of ~ 10 GHz was measured. Since this 3 dB
point matches that of the amplifier at a 10 GHz post detection bandwidth, it should be treated as a lower bound. Note,
the signal was too low to couple directly from detector to spectrum analyzer.
Preliminary optical simulations using GRASP have been run exploring the resolution limits and aberrations imposed by
the off-axis parabolic mirrors. The OAP optical train as modeled in GRASP is displayed in Figure 5c. Results from this
simulation indicate that our reported resolution is diffraction limited and our depth of focus is defined by the amount of
beam walk-off at the feedhorn aperture.
5. BURN IMAGING
5.1 Background
500,000 patients are treated for burns in the United States annually [15]. Of these, ~50,000 require hospitalization, and
400 die each year, making mortality from burns the fifth leading cause of injury-related death in the United States [15].
Care for burn wounds is expensive accounting for 13% of all medical claims [16]. It has been hypothesized that earlier
and more accurate identification of burn severity can reduce the morbidity and costs associated with burn injuries.
Improved burn assessment may also identify patients requiring surgical treatment from those that can be treated more
conservatively [17]. This clinical decision requires accurate assessment of burn wound depth and extent as well as the
potential for healing.
Clinically, burn wounds are divided into three categories: superficial, partial, and full thickness [18]. Superficial
thickness burns are limited to the epidermis. Partial thickness burns involve both epidermis and dermis, and are
subdivided into superficial and deep by their involvement of the papillary and reticular dermis, respectively. Full
thickness burns are those that have completely obliterated the epidermal and dermal layers, and involve underlying
structures such as subcutaneous fat, fascia, and muscle [19]. The critical issue in burn wound assessment is
distinguishing between superficial and deep partial thickness burns. Deep partial and full thickness burns require skin
grafting, whereas more superficial burns can be managed non- operatively [20]. The goal of this work is to use spatially
resolved hydration maps to diagnosis burn severity soon after thermal insult allowing earlier and more accurate
treatment and improved patient outcomes.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
5.2 Experimental setup
We have previously reported on the results of imaging and partial and full thickness burns in a rat model with THz
illumination [14, 21, 22]. The results obtain indicate that the hydration maps of burned tissue can be used to predict
wound severity soon after insult. To expand on this study we have improved image registration, processing, and analysis
to better understand the utility of the time series. A THz image of a partial thickness burn 7 hours after thermal insult is
displayed in Figure 6a where increased reflectivity indicates significant increases of tissue hydration [22]. These results
suggested that THz imaging can create transverse maps of surface wound hydration that delineate the zones of
coagulation and stasis (as shown in Figure 6a). These results were correlated with histopathologic analysis of tissue
sections acquired from the zones circumscribed by the dotted lines in Figure 6a. A number of key observations were
made from this study. First, tissue sections that spanned all the zones on each arm of the cross were difficult to obtain
because each desired section often overlapped the adjacent sections. This hindered some of the histologic analysis.
Second, image sets acquired through time were difficult to correlate between each time point due to FOV registration
problems. In the work described in [22] and [21], we lifted the mylar window from contact with the rat abdomen
between each image acquisition to allow the inflammatory response to develop and progress as normally as possible.
Finally, the dearth of THz, visible, and histologic fiducial markers somewhat hindered our accuracy when comparing
results from different modalities.
(a)
(b)
Figure 6: Experimental protocols updates. (a) THz image of partial thickness burn with labeled zones of stasis
and coagulation and labeled tissue section harvest locations. (b) Framed full thickness burn wound following the
updated experimental protocol.
An image of a partial thickness burn using our revised experimental protocol is shown in Figure 6b displaying some the
details of the new experimental protocol. First the cross-shaped burn has been changed to a thin rectrangle. This allows
collections of numerous tissue samples that span equal parts of burned skin and unburned skin. Second, numerous
fiducial markers have been added to improve intramodal and intermodal registration. The blue tape section in the shape
of a triangle on the left edge of the FOV is a THz fiducial marker and acts as an absorber due to the two-pass scattering
and absorption experienced by the THz illumination beam. The black ink dots applied to the skin provide visible light
registration and aided the tissue sectioning process prior to histologic examination. The mylar window frame has been
machined from brass for increased reflectivity to facilitate continuous calibration throughout image acquisition and the
clear aperture diameter has been expanded from 25 mm to 35 mm.
5.3 Improved THz imagery of partial and full thickness burn wounds
A rat model was prepared with the same protocol as in our previous studies [21, 22]. Image acquisition was also
performed in the same manner with the notable difference that the mylar was not moved throughout the duration of the
experiment. As mentioned above, this was done to facilitate improved image registration across image sets. A
rectangular brand was heated to 200 degree C with contact pressure for 10 seconds. THz, visible and thermal images
were acquired every 15 – 30 minutes for 7 hours. A visible image of the burn wounds at 7 hours is displayed in Figure
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
6b. The visual appearance of the wound did not change throughout the 7 hour study. Following image acquisition the
rat was kept alive for 72 hours to allow the burn wound to progress. The rat was then euthanized and tissue sections
collected for histologic analysis. This analysis confirmed that the burn displayed in Figure 6b was a full thickness burn.
(a) Uninjured
(b) Immediate
(c) 15 mins
(d) 2.5 hrs
(e) 6.5 hrs
Figure 7: THz burn wound imaging time series displaying uninjured tissue and the burn wound at varying time
points throughout the trial. The top row is displayed in false color and the bottom gray scale.
A time series of THz images acquired of the uninjured tissue and following thermal insult at varying time points are
displayed in Figure 7 where the top row is represented with a false color map and the bottom row in gray scale. The
FOV is bordered by a thick ring of high reflectivity corresponding to the brass frame of the Mylar window. Also note
the triangular area of decreased reflectivity on the left edge of each displayed field of view corresponding to the blue
tape fiducial marker. The adhesive is lossy at THz frequencies and the drop in signal in this area is due to the two-pass
absorption of THz illumination as it transmits through tape, reflects from the brass frame, and traverses one final path
through the tape before being collected by the off-axis parabolic mirrors and arriving at the detector. The dark triangular
shaped areas in the corners of each image located beyond the perimeter of the brass ring and correspond to zero
reflectivity (air). These areas as well as the brass ring were used as spatially localized calibration values throughout the
experimental trial.
The rat skin prior to burn imaging is displayed in column (a) and displays low average general reflectivity with slight
variations around the edges of the FOV. A THz image of the burn immediately following thermal insult is displayed in
column (b) and a response similar to prior experiments [22] is observed with an increase in tissue reflectivity across the
unburned area and varying hydration in the contact area. Columns (c), (d), and (e) display THz imagery of burn wound
hydration at 15 minutes, 2.5 hours, and 7 hours following injury. Increases in hydration in the tissue immediately
surrounding the contact are visible in the 15 minute image and significant in the 2.5 hour image. These areas appear to
travel outwards from the burn wound to the periphery of the field of view as evidenced by the 6.5 hour image.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
(a)
(b)
Figure 8: Evolution of burn wound hydration over time. (a) location of burn wound and pixel arrays of interest
on the rat model. (b) Evolution of pixel intensity (tissue hydration) over time.
Hydration distributions and water flows within the burned and surrounding tissues were investigated in varying regions
of interest across the field of view. Figure 8a shows the locations of 2 pixel arrays of interest, one in the edge of the field
of view very close to what was estimated to be the zone of stasis (red ‘x’) and one farther away from the burn in what is
potentially the zone of hyperemia (blue ‘x’). Plots of the pixels intensities (whose values are linearly proportional to the
tissue hydration) are displayed in Figure 8b over a 7 hour period. The proximal pixel (red ‘x’) displays a significant
increase in hydration immediately after thermal insult and then a slow decrease in hydration after peaking at 1 hour, 40
minutes. The distal point (blue ‘x’) displays a slight decrease in hydration in then a slow increase over the majority of
the experiment followed with a factor of four increase in the final hour of the 7 hour trial. We are currently working on
identifying image classifiers based on image features at specific time points and image classifiers generated from an
ensemble of images and evaluating the statistical significance of tests based on these classifiers.
6. CONCLUSIONS
An imaging system was presented that uses direct detection to image hydration in burn wounds. The system architecture
was motivated by the lack of THz spectral signatures observed in physiologic tissues. THz images of burn wounds in rat
models were presented using an improved experimental protocol and enhanced image registration process. Image results
of a full thickness burn display significant shifts in tissue hydration and indicate that wound hydration collects at the
periphery of the burn and moves outward in our animal model. The results presented suggest that THz imaging can be
used to assess burn wound hydration at any arbitrary time point following injury and that classifiers generated from the
image ensembles may provide sensitive and specific diagnostic markers for burn severity.
ACKNOWLEDGEMENTS
This work was sponsored by the Combat Wound Initiative (Henry M. Jackson Foundation, Dr. Alexander Stojadinovic)
Award# 708961
REFERENCES
1.
2.
3.
Ferguson, B. and X.-C. Zhang, Materials for terahertz science and technology. Nat Mater, 2002. 1(1): p. 26-33.
Wallace, V.P., et al., Terahertz pulsed imaging of basal cell carcinoma ex vivo and in vivo. British Journal of
Dermatology, 2004. 151(2): p. 424-432.
Woodward, R.M., et al., Terahertz Pulse Imaging of ex vivo Basal Cell Carcinoma. J Investig Dermatol, 2003.
120(1): p. 72-78.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Taylor, Z.D., et al., Reflective terahertz imaging of porcine skin burns. Opt. Lett., 2008. 33(11): p. 1258-1260.
Taylor, Z.D., et al. THz imaging based on water-concentration contrast. in SPIE Terahertz for Military and
Security Applications VI. 2008. Orlando, FL, USA: SPIE.
Arbab, M.H., et al. Characterization of burn injuries using terahertz time-domain spectroscopy. in SPIE
Advanced Biomedical and Clinical Diagnostic Systems IX. 2011. San Francisco, California, USA: SPIE.
Bennett, D.B., et al., Terahertz sensing in corneal tissues. J. Biomed. Opt. 16(5): p. 8.
Singh, R.S., et al. Terahertz sensing of corneal hydration. in Engineering in Medicine and Biology Society
(EMBC), 2010 Annual International Conference of the IEEE. 2010.
Mittleman, D.M., et al., Recent advances in terahertz imaging. Applied Physics B: Lasers and Optics, 1999.
68(6): p. 1085-1094.
Arnone, D.D., et al. Applications of terahertz (THz) technology to medical imaging. 1999. Munich, Germany:
SPIE.
NIH. http://report.nih.gov/. Available from: http://report.nih.gov/.
Han, P.Y. and X.-C. Zhang, Free-space coherent broadband terahertz time-domain spectroscopy. Measurement
Science and Technology, 2001. 12(11): p. 1747.
Tonouchi, M., Cutting-edge terahertz technology. Nat Photon, 2007. 1(2): p. 97-105.
Taylor, Z.D., et al., THz Medical Imaging: in vivo Hydration Sensing. Terahertz Science and Technology, IEEE
Transactions on, 2011. 1(1): p. 201-219.
American Burn Association (ABA), Burn incidence and treatment in the US: 2007 fact sheet 2011.
American Burn Association (ABA) National Burn Repository (2005 Report). 2011.
Mandal, A., Quality and cost-effectiveness--Effects in burn care. Burns, 2007. 33(4): p. 414-417.
Jackson, D., The diagnosis of the depth of burning. British Journal of Surgery, 1953. 40: p. 588-596.
Monstrey, S., et al., Assessment of burn depth and burn wound healing potential. Burns, 2008. 34(6): p. 761769.
Jaskille, A.D., et al., Critical Review of Burn Depth Assessment Techniques: Part I. Historical Review. Journal
of Burn Care & Research, 2009. 30(6): p. 937-947 10.1097/BCR.0b013e3181c07f21.
Tewari, P., et al., In vivo terahertz imaging of rat skin burns. Journal of Biomedical Optics, 2012. 17(4): p.
040503-3.
Tewari, P., et al. Advances in biomedical imaging using THz technology with applications to burn-wound
assessment. in SPIE Terahertz Technology and Applications V. 2012. San Francisco, CA.
*zdeis@seas.ucla.edu phone 310-267-4985 fax 310-794-5956