This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging
Comparison of NIR versus SWIR fluorescence
image device performance using working
standards calibrated with SI units
Banghe Zhu, Sunkuk Kwon, John C. Rasmussen, Maritoni Litorja, and Eva M. Sevick-Muraca*
Abstract— Recently, fluorescence imaging using shortwave
infrared light (SWIR, 1,000-2,000 nm) has been proposed as
having advantage over conventional near-infrared fluorescence
(NIRF) imaging due to the reduced tissue scattering, negligible
autofluorescence, comparable tissue absorption, and the discovery
that indocyanine green (ICG), used clinically as a NIRF contrast
O ver the past two decades, techniques of near-infrared
agent, also has fluorescence emission in SWIR regime. Images of
ICG in small animals acquired by commercial Si-based and
InGaAs-based imaging cameras have been qualitatively
compared, however the lack of working standards to quantify
performance of these imaging systems limits quantitative
comparison. Without quantification using a traceable in vitro test,
clinical adoption of rapidly evolving advances in both NIRF and
SWIR imaging devices will become limited. In this work, we
developed an ICG based fluorescent solid working standard
calibrated with SI units (mW · cm-2· sr-1) for quantification of
measurement sensitivity of Si, GaAs-intensified Si, and InGaAs
based camera systems, their signal-to-noise ratio (SNR), and
contrast in non-clinical tests. In addition, we present small animal
and large animal imaging with ICG for qualitative comparison of
the same SWIR fluorescence and NIRF imaging systems. Results
suggest that SWIR fluorescence imaging of ICG may have
superior resolution in small animal imaging compared to NIRF
imaging, but lack of measurement sensitivity, SNR, contrast, as
well as water absorption limits deep penetration in large animals.
Index Terms— Molecular and cellular imaging, Optical
imaging/OCT/DOT, Evaluation and performance, System design,
Validation.
Manuscript received May 30, 2019; accepted August 21, 2019. This work was
supported in parts by the Translational Research Institute through NASA
Cooperative Agreement NX16AO69A.
B. Zhu is with the Brown Foundation Institute of Molecular Medicine, The
University of Texas Health Science Center, Houston, Texas, 77030 USA (e-
mail: Banghe.Zhu@uth.tmc.edu).
S. Kwon is with the Brown Foundation Institute of Molecular Medicine, The
University of Texas Health Science Center, Houston, Texas, 77030 USA (e-
mail: Sunkuk.Kwon@uth.tmc.edu).
J. C. Rasmussen is with the Brown Foundation Institute of Molecular
Medicine, The University of Texas Health Science Center, Houston, Texas,
77030 USA (e-mail: John.Rasmussen@uth.tmc.edu).
M. Litorja is with the National Institute of Standards and Technology,
Gaithersburg, MD 20899 USA (e-mail: maritoni.litorja@nist.gov).
*E. M. Sevick-Muraca is with the Brown Foundation Institute of Molecular
Medicine, The University of Texas Health Science Center, Houston, Texas,
77030 USA (e-mail: Eva.Sevick@uth.tmc.edu).
JCR, BZ and EMS-M are listed as inventors on patents related to near-
infrared fluorescence lymphatic imaging and may receive future financial
benefit from its commercialization. JCR, BZ, EMS-M and The University of
Texas Health Science Center at Houston have research-related financial
interests in Lymphatics Technologies, Inc. The remaining authors have no
financial relationships relevant to this article to disclose.
Copyright (c) 2019 IEEE. Personal use of this material is permitted. However, permission to use this material for any other
purposes must be obtained from the IEEE by sending a request to pubs-permissions@ieee.org.
0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
1
I. INTRODUCTION
fluorescence (NIRF) imaging have been developed and
translated into clinical use [1-3]. The technique utilizes incident
light within the wavelength range of 760 - 900 nm, within which
the hemoglobin and water in tissues exhibit low absorption such
that incident near-infrared (NIR) light penetrates deeply into the
tissue, activates exogenous NIRF agents, and allows imaging of
large tissue volumes from the emitted fluorescent light. ICG
was originally approved by the US Food and Drug
Administration (FDA) in 1956 as a reagent for clinical use for
determining cardiac output, hepatic function and ophthalmic
angiography following i.v. administration and has an excellent
record of safety. Today there are now approximately 25
commercially available NIRF imaging devices including open-
field surgery camera systems, endoscopes, and laparoscopes
with the majority designed for intraoperative detection of ICG
to assess perfusion of skin flaps [3]. These devices employ 700
– 806 nm excitation light and collect fluorescent emission
between 800 nm and 900 nm. Yet unlike all other conventional
imaging modalities, these NIRF imaging devices lack IEC
(International Electrotechnical Commission) traceable
measurements of performance in International System (SI) of
units that enable accurate predictions of sensitivity/resolution
for detecting ICG or other NIRF imaging agents with non-
clinical tests. The seven base SI units are defined using the
physical constants of the universe, lending confidence that all
scientific and commercial measurements that are SI-traceable
are comparable across instrumentation, methods, time and
location. The lack of non-clinical, metrologically-traceable test
standards complicate the efficient adaptation and efficient
translation of ongoing, rapid technological advancements of
optical components driven primarily by the military and
communications industries [4, 5]. One recent example of
technological advancement is the recent advent of InGaAs-
based devices that expand biomedical fluorescence imaging
devices into the shortwave infrared (SWIR, 1,000 –2,000 nm)
wavelength regime [6]. Because the degree of tissue scattering
is inversely proportional to wavelength λ-α (α=0.2~4 for
biological tissues), fluorescence at longer SWIR wavelengths
experiences reduced tissue light scattering and less
autofluorescence than at NIRF wavelengths, albeit at the
expense of somewhat increased water absorption [7]. The
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging
benefits of reduced photon scattering, negligible autofluores-
cence, and comparably low absorbance in SWIR window have
been hypothesized to enable high-resolution imaging at greater
penetration depths than achievable in NIR window [8-12].
Moreover, ICG has recently been found to have SWIR
fluorescence emission (>1064 nm) following excitation at 808
nm and its SWIR emission spectra can be found in [13]. ICG
has been successfully used in small animals to demonstrate
SWIR intravital microscopy, noninvasive SWIR imaging in
blood and lymph vessels and SWIR assessment of hepatobiliary
clearance using emerging commercial InGaAs array detectors
[14]. Despite the small fluorescent yield of ICG at 1064 nm
following the NIR excitation, the large 200 nm Stokes shift may
be particularly advantageous for SWIR fluorescence imaging.
Even with a comparatively larger fluorescent yield, the
incomplete rejection of excitation light due to the small (10 –
60 nm) Stokes shift experienced in NIRF systems limits their
performance [15], a problem potentially avoided by the large
Stokes shift probed by SWIR imaging. As a result, SWIR has
thus been proposed as having advantage over NIRF imaging for
clinical applications [13, 14], yet no quantitative comparison
using a non-clinical or in vitro test assessment has been
presented. Non-clinical or in vitro testing provides an
inexpensive, repeatable, and when reported in SI units,
traceable measurement that can hasten direct comparison and
enable evaluation of improvement in device advancements such
as potentially offered by SWIR detectors. Figure 1 is a
schematic depicting the typical spectral responses of the
systems currently deployed in clinical imaging of ICG.
Fig. 1. The typical spectral response curves of detectors made of Si, GaAs,
and InGaAs materials.
The main objective of this work is to evaluate performance
of Si-, InGaAs- and GaAs intensified- based camera systems
for detecting ICG using an in vitro, fluorescent solid phantom
calibrated with SI-traceable units. A secondary objective is to
relate performance assessed from the quantitative in vitro test
to qualitative images from large and small animals. It should
be noted that the comparisons made herein are specific to the
devices compared, and that the methodology for assessing
performance in SI-traceable units could be used to drive
technological advancements that improve existing and future
NIRF and SWIR fluorescent devices.
0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
2
II. MATERIALS AND METHODS
A. ICG based working standards calibrated with SI units
Previously we have used stable, working standards based upon
the small fluorescent yield of QDots800 to assess performance
of various custom and commercial NIRF devices that operate
within the same or similar wavelength regimes. In this work,
detection of ICG at NIR and SWIR wavelengths defines the
task for which device performance is assessed. For this reason,
we constructed working standards similar to that of our prior
work [4, 5], using ICG as opposed to QDots800. The standard
consists of 6 wells machined from a 96 well plate and filled with
a mixture of polyurethane, TiO2 (1.5 mg/mL) and increasing
concentrations of ICG diluted in water or in ethanol.
Polyurethane was chosen as the base material for the phantom
due to its optical clarity, fast curing time, and affordability;
TiO2 was added to the phantom to exaggerate the optical
scattering properties of tissue that lead to incomplete excitation
rejection; and 6 wells of the phantom had varying
concentrations of ICG. A low radiance phantom was
constructed to have similar fluorescent radiances to our prior
work with ICG concentrations of 0 (acting as a non-fluorescent
background), 20, 40, 60, 80, and 100 nM and a high radiance
phantom was constructed with ICG concentrations of 200, 400,
600, 800 and 1000 nM. To calibrate the radiance from the ICG
based fluorescent solid phantoms with SI units in NIR window,
we used the same methodology described previously for
calibration of QDots 800 based luminescent solid tissue
phantoms with SI units [5]. Briefly, a NIST calibrated power
meter is used to assign a radiance scale to an 830 nm diffuse
source (830 nm laser diode and integrating sphere, known as
the “reference source”) with output matching emission from the
fluorescent tissue phantoms. Both reference source and tissue
phantoms are viewed by the camera system under the same
radiometric conditions. Thus, the reference source is used to
calibrate the arbitrary unit (AU) output of the camera system
configured under specific radiometric conditions and
transferring the radiance scale (mW ∙ cm−2 ∙ sr −1 ) to the tissue
phantoms illuminated at a calibrated irradiance at 785 nm. The
tissue phantoms can then be used as working standards at this
wavelength to evaluate camera systems at any other radiometric
configuration. To calibrate SI units in SWIR window, we
replaced 785 nm excitation source with 808 nm excitation
source, and the 830 nm reference source with 1064 nm
reference source with all other configurations held constant.
Radiance from the integrating sphere reference sources were
measured by NIST-calibrated InGaAs- and Si-photodiode
power meters at the wavelengths of 1064 nm and 830 nm
(NIST15, Gaithersburg, MD; S2281-04, Hamamatsu, Japan).
B. NIR and SWIR fluorescence imaging systems
In NIRF imaging set-ups shown in Fig. 2(a), a 785 nm laser
diode (HPD1005-9mm-78503 model, High Power Devices
Inc., NJ) that was typically employed in our investigational
devices was used to illuminates the tissue/standard surfaces of
interest through a diffuser. The collection of fluorescence
signals were implemented using two 830 nm band pass filters
(830FS10, optical density > 5 at 785 nm, Andover, Salem, NH)
separated with the Nikon focus lens (AF NIKKOR 28 mm
f/2.8D, Nikon, NY). The Nikon lens introduced a small amount
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging 3

Fig. 2. Schematics of NIR (a) and SWIR (b) fluorescence imaging systems installed with Si, GaAs coupled Si, and InGaAs based cameras.

of loss between two filters to efficiently cancel multiple-path
interference effects and increase the combined optical density.
This reduced the leakage of excitation light through the filters.
The collected fluorescence signals were then registered on a Si-
based commercial electron-multiplying couple-charged device
(EMCCD) camera (Photon Max 512, Princeton Instruments,
Trenton, NJ) or a GaAs-based intensified scientific
complementary metal–oxide–semiconductor (IsCMOS)
camera (a customized fiber optic coupled Zyla 5.5 sCMOS,
ANDOR, Concord, MA) as developed in our past works [5]. In
the SWIR fluorescence imaging system shown in Fig. 2(b), a
laser source operating at 808 nm (L808P1000MM model,
Thorlabs, NJ) was utilized to illuminate the tissue/standards.
The choice of 808 nm as the excitation wavelength was based
upon prior literature reports of SWIR fluorescence detection of
ICG. The generated fluorescence signals were collected with a
1064 nm long-pass filter (BLP01-1064R, optical density >5 at
808 nm, Semrock, Rochester, NY) and then focused through a
SWIR lens (SR2343-A01 model, StingRay Optics, Keene, NH)
onto an InGaAs-based PIN Photodiode camera (Ninox SWIR
640, Raptor Photonics, kindly loaned by Phoenix Engineering,
Carson, CA). The working distances and hence radiometric
conditions were identical for all systems. All cameras were
cooled to their corresponding temperatures suggested by the
manufacturers and the field of views (FOVs) of both NIR and
SWIR fluorescence imaging systems were adjusted to 10 cm in
diameter.
C. Figures of merit for fluorescence imaging
Signal-to-noise ratio (SNR) and contrast were used to compare
and quantify the measurement sensitivity of both NIR and
SWIR fluorescence imaging systems. To calculate the SNR and
contrast, the ICG-based fluorescent solid phantom with SI units
was first positioned under the FOV and the irradiance measured
on the phantom surface was 3.7 mW/cm2 for 785 nm excitation
and 4.2 mW/cm2 for 808 nm excitation. Fluorescence images
were acquired at two different integrating times of 33 and 200
ms with the gain adjusted such that an individual pixel value
within the phantom well with the highest ICG concentration
was just less than the full well capacity of the camera. From the
acquired fluorescence images, the SNR corresponding to each
well in the calibrated ICG-based fluorescent solid phantom was
calculated using the following equation:
S t C   S b 0
SNR  (1)

where S t C  and S b 0 represent the averaged arbitrary units
(AU) values for the pixels corresponding to the entire well
containing C concentration of ICG and to the well without ICG
respectively, and σ represents the standard deviation of AU
values in the well without ICG. The calculated SNR was then
plotted as a function of radiance (mW ∙ cm−2 ∙ sr −1 ) of each
well in ICG based fluorescent solid phantom. The contrast was
computed by the following relationship:
S t C 
Contrast  1 (2)
S b 0
The computed contrast was then plotted as a function of
radiance (mW ∙ cm−2 ∙ sr −1 ) of each well of the working
standard.

0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging
D. Small and large animal imaging with NIRF and SWIR
imaging systems
Animal studies were conducted under protocols approved by
the University of Texas Health Science Center Animal Welfare
Committee. Dynamic imaging was performed using the
imaging systems described above on anesthetized athymic mice
after 100 µL of 323 µM ICG was injected into the tail vein.
Images were acquired dynamically with acquisition times of 50
ms using the EMCCD system and 200 ms using the GaAs-based
IsCMOS and InGaAs systems. The irradiance measured on the
animal surfaces has a spatial variation of 3.0~4.4 mW/cm2 for
785 nm excitation and 3.1-5.1 mW/cm2 for 808 nm excitation.
In ancillary studies of swine, dynamic imaging was conducted
using the IsCMOS and InGaAs systems following
intramuscular injection of 1-2 mL of 62.5 µM using an 18-
gauge needle.
III. RESULTS
A. Radiance of ICG based fluorescent solid phantom in SI
units
When reconstituted in water prior to incorporation with
polyurethane and TiO2, or when diluted in ethanol at
concentrations lower than 200 nM, ICG emission was too weak
for detection or quantification using the SWIR system. Figure
3(a) shows that ICG in ethanol exhibited higher emission at
>1064 nm than ICG in water, which has been attributed to its
higher absorption cross section in ethanol [13], but could also
be impacted by water absorption at SWIR wavelengths.
Because SWIR fluorescence imaging systems can detect
emission of the lowest ethanol-diluted ICG concentration (200
nM) as shown in Fig. 3(b), working standards prepared with
ethanol dilution of ICG at concentrations of 200 to 1000 nM
were adopted. Figure 4 illustrates the ratio of emission radiance
at 830 nm to the incident 785 nm excitation irradiance (ratio
uncertainty = 0.10) as well as the ratio of emission radiance at
1064 nm to the incident 808 nm excitation irradiance (ratio
uncertainty = 0.11) for each of the wells containing different
concentrations of ICG. It can be observed that the ratio of
Fig. 3. (a) The acquired SWIR fluorescence image of Eppedorf tubes filled
with ICG diluted in water and ethanol, demonstrating higher fluorescence
signals in ethanol. (b) Three acquired fluorescence images from ICG based
fluorescent solid phantom using NIR and SWIR fluorescence imaging
systems.
0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
4
emission radiance to excitation irradiance for both NIR and
SWIR fluorescence channels is proportional to the ICG
concentration, although at the lower and upper ranges of
radiance, the radiometric-specific function curves slightly
deviate from linearity, demonstrating the dark noise of the
detector and maximum measurable radiance prior to camera
saturation. The slope of the lines refer to the fluorescent
characteristic of the standard, which for our past work with
inorganic QDots800 is stable for years, but for ICG deteriorates
with light exposure and time slowly [5]. As a result, these
standards were prepared and used in a darkroom and exposure
to illumination for no more than a few seconds over which time
their output appeared to be stable. It is noteworthy, that despite
the small fluorescence yield at SWIR over NIR wavelengths,
the radiance in SWIR fluorescence channel is higher than that
in NIRF channel. We believe that the narrow bandwidth of 830
nm bandpass filter limited the amount of NIR fluorescence
collected as compared to the SWIR fluorescence collected
using the 1064 nm long pass filter. We did not evaluate whether
enhancement of optical effects that occur due to the low
absorption and high refractive index of TiO2 scattering particles
in the presence of a fluorophore could be responsible the large
differences between the NIR and SWIR measurement of
fluorescent characteristics of the working standard.
B. Measurement sensitivity of both NIR and SWIR
fluorescence imaging systems
Figure 5 shows that the SNR of NIRF imaging systems are
much higher than that of SWIR system despite that there is an
2-3 order of magnitude greater radiance in SWIR fluorescence
channel. At longer integrating times of 200 ms, the SNR is
improved as compared to that at 33 ms for both the Si-based
EMCCD and InGaAs systems as depicted in Fig. 5(a) and (c).
For the GaAs-based IsCMOS imager (shown in Fig. 5(b))
shorter integrating time is compensated by the increased gain of
the intensifier. Figure 6 shows that the Si-based EMCCD NIRF
imaging system provides approximately twice the contrast as
that of SWIR system. In comparison to both, GaAs-based-
IsCMOS NIRF imaging system gives greater contrast as shown
Fig. 4. Emission radiance normalized to excitation irradiance of ICG
based working standards in NIR (Left axis) and SWIR (Right axis)
fluorescence imaging channels, where arrows to show which y-axis
corresponds to each dataset.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging 5

Fig. 5. The plots of signal-to-noise ratio as a function of radiance at 33 ms and 200 ms exposure times using EMCCD (Si) (a) or IsCMOS (GaAs intensified)
(b) camera based NIRF imaging system and InGaAs (c) camera based SWIR fluorescence imaging system.

Fig. 6. The plots of contrast as a function of radiance at 33 ms and 200 ms exposure times using EMCCD (Si) (a) or IsCMOS (GaAs Intensified) (b) camera
based NIRF imaging system and InGaAs (c) camera based SWIR fluorescence imaging system.

in Fig. 6(b). Using working standards with lower NIR

radiances that challenge device sensitivity, we previously
showed that the GaAs-based intensification improved SNR and
contrast compared to non-intensified CCD and sCMOS NIRF
systems [4, 5]. Fig. 7 shows a comparison of SWIR and GaAs-
based IsCMOS performance at low radiances and 200 ms
exposure times as provided by standards made with ICG diluted
with water up to 100 nM. The results suggest that SWIR does
not have as much sensitivity as NIRF-based systems to measure
clinically relevant concentrations of ICG for medical imaging.
Typically, a USAF target is generally used to assess device
resolution which will be a function of irradiance, radiometry,
and pixilation of the detector. However, one advantage of
SWIR over NIR may be the reduced scattering by tissues that
results in the improved resolution of fluorescent targets. To Fig. 7. Signal-to-noise ratio versus contrast for low radiance set of working
demonstrate this potential impact on target detection, we standards that included 0 – 100 nM of ICG diluted in water as measured
by GaAs-IsCMOS NIRF system and InGaAs SWIR system. As contained
determined the full-width half maximum of a (27.5 mm long, in the insert, radiance values at SWIR were not sufficiently detected as in
4.7 mm diameter) rod filled with ICG diluted with ethanol to a the high radiance set of working standards that included > 200 nM of ICG
concentration of 322 μM concentration as it was immersed in diluted in ethanol (Figures 5, 6) and therefore could not be calibrated with
1.0% liposyn, to depths of 0, 2 and 4 mm from the surface. SI units. As a result, SNR is plotted as a function of contrast to show
imaging at SWIR had lower contrast with lower SNR than NIRF imaging.
Figure 8 (a) shows the acquired NIR and SWIR fluorescence

0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging
Fig. 8. (a) Fluorescence images in NIR and SWIR fluorescence imaging channels of a rod consisting ICG (322 μM) at depths of 0, 2, and 4 mm in 1.0%
Liposyn. (b) Full-width-half-maximum (FWHM) of fluorescent rod as a function of depth in 1.0% Liposyn, demonstrating improved spatial-resolution at
SWIR as opposed to NIRF wavelengths.
images, from which the full-width-half-maximum (FWHM) of
feature width was calculated, as shown in Fig. 8(b). At this
high concentration of ICG, the FWHM of the SWIR ICG signal
deteriorates less with increasing depth than does the NIRF ICG
signal.
To relate performance characterized by the working
standards and in vitro tests to actual imaging, we also evaluate
the devices with small and large animal tests. Dynamic in vivo
imaging of mice injected i.v. with 100 L of 323 M ICG was
performed using NIRF (EMCCD and GaAs-based IsCMOS)
and SWIR fluorescence demonstrate comparable performance
with potentially less scatter and greater clarity with the SWIR
system, as shown in supplemental videos 1- 3 and Fig. 9. This
result may be expected in the context of the results in Figure 8.
However, when using the larger swine animal, i.m. injection
with 2 mL of 62.5 µM can be seen using the GaAs-based
IsCMOS NIRF system, but not with the InGaAs-based SWIR
system, as shown in Fig. 10. This latter result is likely due to
the relatively strong absorption of water at SWIR wavelengths
that limits penetration depth and is expected with the results
obtained in the ethanol and water-based working standards.
IV. DISCUSSION
Herein we used an ICG-based working standard calibrated with
SI-traceable units to report the quantitative performance of two
NIR and a single SWIR fluorescence imaging system and then
followed with a qualitative comparison in large and small
animal studies. The advantages of SWIR, namely reduced
tissue scattering, a large Stokes shift that minimizes
contributions from autofluorescence and excitation light
leakage, and the larger bandwidth for fluorescent collection
bode well for its deployment in biomedical optical imaging.
However increased water absorption and the smaller
fluorescent yield at SWIR wavelengths may offset these
advantages, especially for clinical applications that use ICG.
Because this study was confined to the three separate systems
studied, we would like to emphasize that we do not broadly
conclude that NIRF imaging is advantageous over SWIR
0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
6
fluorescence imaging of ICG or any other potential fluorescent
contrast agent [16] that could be translated clinically for
molecularly targeted detection of diseased tissues. With more
SWIR sensitive array technologies made of other materials such
as InSb and HgCdTe, or perhaps more impactful, the advent of
SWIR sensitive intensifiers that couple to conventional Si-
based detectors, it may be possible that sensitivity could be
improved beyond that possible with NIRF imaging systems.
Interestingly, a recent work by Carr et al [17] demonstrates that
the SWIR system has highest image contrast with a higher
imaging penetration depth at the peak absorbance of water
(~1450 nm). The SWIR system evaluated herein was less
sensitive in detecting ICG than NIRF systems as depicted in
Figures 5-7. Future comparisons of imaging systems need to
be based on quantitative characterization using traceable
working standards that are calibrated in SI units and that are
appropriately used to objectively address performance in
clearly defined tasks through receiver operator curve analysis.
Before task-based assessments are performed within expensive
in vivo or clinical tests, an in vitro or non-clinical test can
provide a simple means of assessing and comparing
performance of different devices, their operation, and
radiometric geometries in which they are used. Subsequent
evaluation of task-based performance within in vivo or clinical
testing will require calibration with SI units to be effective.
Previously, we developed a QDots 800 based working
standard that, unlike the ICG working standard described
herein, possessed stable emission over a period of a year or
greater, but did not possess emission into the SWIR wavelength
regimes. The broad emission spectra of ICG ranging from NIR
to SWIR spectra region allows performance comparisons of
both NIR and SWIR fluorescence imaging systems possible.
On the downside, these ICG based working standards
photodecay with exposure to light and therefore are not as
desirable as inorganic ones. Future work to develop stable
working standards with durable stability and spanning NIR and
SWIR wavelength ranges could help spur the development of
successful fluorescent imaging devices and the imaging agents
they are designed to detect.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging 7

Fig. 9. Example of ICG imaging of small animals using EMCCD (Si) (a) or IsCMOS (GaAs) (b) camera based NIRF imaging system and InGaAs (c) camera
based SWIR fluorescence imaging system.

Fig. 10. Example ICG imaging in large animals using IsCMOS (GaAs) (a) camera based NIRF imaging system and InGaAs (b) camera based SWIR
fluorescence imaging system.

*
Note: References are made to certain commercially available
products in this paper to adequately specify the experimental
procedures involved. Such identification does not imply
recommendation or endorsement by the National Institute of
Standards and Technology, nor does it imply that these products
are the best for the purpose specified.
Acknowledgement: The authors thank Jeanne Houston at NIST
for the loan of the calibrated InGaAs photodiode used in the
SWIR measurements.
REFERENCES
[1] E. Sevick-Muraca, "Translation of near-infrared fluorescence
imaging technologies: emerging clinical applications," Annual
Review of Medicine, vol. 63, pp. 217-231, 2012.
[2] J. T. Alander et al., "A review of indocyanine green fluorescent
imaging in surgery," Journal of Biomedical Imaging, vol. 2012, p.
7, 2012.
[3] B. Zhu and E. Sevick-Muraca, "A review of performance of near-
infrared fluorescence imaging devices used in clinical studies," The
British Journal of Radiology, vol. 88, no. 1045, p. 20140547, 2014.
[4] B. Zhu, J. C. Rasmussen, and E. M. Sevick‐Muraca, "A matter of
collection and detection for intraoperative and noninvasive near‐
infrared fluorescence molecular imaging: To see or not to see?,"
Medical Physics, vol. 41, no. 2, 2014.

0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMI.2019.2937760, IEEE
Transactions on Medical Imaging
IEEE Transactions on Medical Imaging 8

[5] B. Zhu, J. C. Rasmussen, M. Litorja, and E. M. Sevick-Muraca,
"Determining the performance of fluorescence molecular imaging
devices using traceable working standards with SI units of
radiance," IEEE Transactions on Medical Imaging, vol. 35, no. 3,
pp. 802-811, 2016.
[6] G. Hong, A. L. Antaris, and H. Dai, "Near-infrared fluorophores for
biomedical imaging," Nature Biomedical Engineering, vol. 1, no. 1,
p. 0010, 2017.
[7] S. L. Jacques, "Optical properties of biological tissues: a review,"
Physics in Medicine & Biology, vol. 58, no. 11, p. R37, 2013.
[8] K. Welsher, S. P. Sherlock, and H. Dai, "Deep-tissue anatomical
imaging of mice using carbon nanotube fluorophores in the second
near-infrared window," Proceedings of the National Academy of
Sciences, vol. 108, no. 22, pp. 8943-8948, 2011.
[9] G. Hong et al., "Multifunctional in vivo vascular imaging using
near-infrared II fluorescence," Nature Medicine, vol. 18, no. 12, p.
1841, 2012.
[10] D. Naczynski et al., "Rare-earth-doped biological composites as in
vivo shortwave infrared reporters," Nature Communications, vol. 4,
p. 2199, 2013.
[11] H. Kantamneni et al., "Surveillance nanotechnology for multi-organ
cancer metastases," Nature Biomedical Engineering, vol. 1, no. 12,
p. 993, 2017.
[12] H. Wan et al., "A bright organic NIR-II nanofluorophore for three-
dimensional imaging into biological tissues," Nature
Communications, vol. 9, no. 1, p. 1171, 2018.
[13] Z. Starosolski, R. Bhavane, K. B. Ghaghada, S. A. Vasudevan, A.
Kaay, and A. Annapragada, "Indocyanine green fluorescence in
second near-infrared (NIR-II) window," Plos One, vol. 12, no. 11,
p. e0187563, 2017.
[14] J. A. Carr et al., "Shortwave infrared fluorescence imaging with the
clinically approved near-infrared dye indocyanine green,"
Proceedings of the National Academy of Sciences, vol. 115, no. 17,
pp. 4465-4470, 2018.
[15] E. M. Sevick-Muraca and J. C. Rasmussen, "Molecular imaging
with optics: primer and case for near-infrared fluorescence
techniques in personalized medicine," Journal of Biomedical
Optics, vol. 13, no. 4, p. 041303, 2008.
[16] B. K. Byrd et al., "Characterizing short-wave infrared fluorescence
of conventional near-infrared fluorophores," Journal of Biomedical
Optics, vol. 24, no. 3, p. 035004, 2019.
[17] J. A. Carr, M. Aellen, D. Franke, P. T. So, O. T. Bruns, and M. G.
Bawendi, "Absorption by water increases fluorescence image
contrast of biological tissue in the shortwave infrared," Proceedings
of the National Academy of Sciences, vol. 115, no. 37, pp. 9080-
9085, 2018.

0278-0062 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.