ArticlePDF Available

Towards monitoring dysplastic progression in the oral cavity using a hybrid fiber-bundle imaging and spectroscopy probe

Authors:

Abstract

Intraepithelial dysplasia of the oral mucosa typically originates in the proliferative cell layer at the basement membrane and extends to the upper epithelial layers as the disease progresses. Detection of malignancies typically occurs upon visual inspection by non-specialists at a late-stage. In this manuscript, we validate a quantitative hybrid imaging and spectroscopy microendoscope to monitor dysplastic progression within the oral cavity microenvironment in a phantom and pre-clinical study. We use an empirical model to quantify optical properties and sampling depth from sub-diffuse reflectance spectra (450–750 nm) at two source-detector separations (374 and 730 μm). Average errors in recovering reduced scattering (5–26 cm−1) and absorption coefficients (0–10 cm−1) in hemoglobin-based phantoms were approximately 2% and 6%, respectively. Next, a 300 μm-thick phantom tumor model was used to validate the probe’s ability to monitor progression of a proliferating optical heterogeneity. Finally, the technique was demonstrated on 13 healthy volunteers and volume-averaged optical coefficients, scattering exponent, hemoglobin concentration, oxygen saturation, and sampling depth are presented alongside a high-resolution microendoscopy image of oral mucosa from one volunteer. This multimodal microendoscopy approach encompasses both structural and spectroscopic reporters of perfusion within the tissue microenvironment and can potentially be used to monitor tumor response to therapy.
1
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
www.nature.com/scientificreports
Towards monitoring dysplastic
progression in the oral cavity using
a hybrid ber-bundle imaging and
spectroscopy probe
Gage J. Greening1, Haley M. James2, Mary K. Dierks3, Nontapoth Vongkittiargorn1,
Samantha M. Osterholm1, Narasimhan Rajaram1,* & Timothy J. Muldoon1,*
Intraepithelial dysplasia of the oral mucosa typically originates in the proliferative cell layer at the
basement membrane and extends to the upper epithelial layers as the disease progresses. Detection
of malignancies typically occurs upon visual inspection by non-specialists at a late-stage. In this
manuscript, we validate a quantitative hybrid imaging and spectroscopy microendoscope to monitor
dysplastic progression within the oral cavity microenvironment in a phantom and pre-clinical study. We
use an empirical model to quantify optical properties and sampling depth from sub-diuse reectance
spectra (450–750 nm) at two source-detector separations (374 and 730 μm). Average errors in recovering
reduced scattering (5–26 cm1) and absorption coecients (0–10 cm1) in hemoglobin-based phantoms
were approximately 2% and 6%, respectively. Next, a 300 μm-thick phantom tumor model was used
to validate the probe’s ability to monitor progression of a proliferating optical heterogeneity. Finally,
the technique was demonstrated on 13 healthy volunteers and volume-averaged optical coecients,
scattering exponent, hemoglobin concentration, oxygen saturation, and sampling depth are presented
alongside a high-resolution microendoscopy image of oral mucosa from one volunteer. This multimodal
microendoscopy approach encompasses both structural and spectroscopic reporters of perfusion within
the tissue microenvironment and can potentially be used to monitor tumor response to therapy.
Intraepithelial dysplastic progression within the oral mucosa is a dynamic process that typically arises at the base-
ment membrane and is classied into stages based on how far it has spread towards the upper epithelial layers.1–4
For example, mild dysplasia occurs in the basal epithelial layers, directly above the basement membrane. As
dysplasia progresses upwards towards the apical epithelial surface, the stages are characterized as moderate and
severe (or carcinoma in-situ), respectively2–4. ese stages are not considered invasive cancer since they have not
yet penetrated the basement membrane and metastasized, although the severity of dysplasia increases this risk2,4.
It has been found that < 5%, 3–15%, and > 15% of patients with mild, moderate, and severe dysplasia, respectively,
progressed to carcinoma2,4. Oral squamous cell carcinoma (OSCC) is the most common form of this carcinoma in
the oral cavity and patients diagnosed with OSCC have a 5-year survival rate of less than 60–70% and this num-
ber decreases in developing countries2,5–7. is is because primary detection of dysplastic malignancies typically
occurs upon visual inspection by non-specialized dentists, who then refer patients to specialists5,8,9. Diagnoses at
this point are oen late-stage8. erefore, detection of oral dysplasia at its various stages via aordable, available,
and non-invasive techniques is crucial in limiting the number of cases that progress to OSCC. Several recent
non-invasive translational endoscopy-based techniques have aimed at improving detection.
One such technique is high-resolution microendoscopy, which can provide clinicians rapid, high-resolution
visualization of tissue architecture and histology when compared to that of the naked eye alone. ese tech-
niques provide a step towards point-of-care “optical biopsy,” potentially reducing the number of biopsies per-
formed each year7,10. Preclinical and clinical studies using high-resolution microendoscopy techniques have been
1Department of Biomedical Engineering, University of Arkansas, Fayetteville, Arkansas 72701, USA. 2Department of
Chemistry and Biochemistry, University of Arkansas, Fayetteville, Arkansas 72701, USA. 3Department of Biological
Sciences, University of Arkansas, Fayetteville, Arkansas 72701, USA. *These authors contributed equally to this
work. Correspondence and requests for materials should be addressed to N.R. (email: nrajaram@uark.edu) or T.J.M.
(email: tmuldoon@uark.edu)
Received: 18 January 2016
Accepted: 06 May 2016
Published: 25 May 2016
OPEN
www.nature.com/scientificreports/
2
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
demonstrated in various body organs including the oral cavity7,11, esophagus12–15, lower gastrointestinal tract16–21,
cervix22,23, ear24–26, and liver and pancreas27. Furthermore, several studies have developed high-resolution imag-
ing techniques compatible with the biopsy port of conventional white-light endoscopes, making it more attractive
for clinicians to adopt these new techniques10,13,19. Work has also been performed in quantifying high-resolution
microendoscopy image data, but for the most part this remains a qualitative screening technique14,15,21,28. e
advantages of high-resolution microendoscopy are low cost, portability, and instantaneous imaging of tissue
architecture. However, a drawback of high-resolution microendoscopy is lack of depth sectioning, meaning it
can only resolve tissue architecture at the apical epithelial surface. More complex instrumentation does exist to
overcome this drawback, including laser scanning confocal systems, but this instrumentation requires galvanom-
eters or microelectromechanical (MEMS)-based technology to do so. Additionally, information gathered by these
more complex depth-sensitive technologies are primarily qualitative29–32. High-resolution microendoscopy can
thus benet from additional depth sensitive modalities since mild and moderate dysplasia are oen sub-epithelial
surface phenomena, but relatively simple and quantitative techniques are desirable.
One depth sensitive technique that has demonstrated diagnostic ecacy is diuse reectance spectroscopy
(DRS), a well-established method capable of non-invasively quantifying volume-averaged tissue optical parame-
ters using simple probe designs33–39. Raw DRS data is given in terms of reectance, that is, the percentage of light
recovered from a detection ber to light delivered by a source ber. Studies have shown that volume-averaged
optical properties, such as reduced scattering coecient (μ
s) and absorption coecient (μ
a) can be determined
from in vivo samples34,38,40–47. It should be noted that these extracted values are based on the delivery and collec-
tion of light throughout an oen inhomogeneous layered media, such as tissue, and extracted optical properties
thus represent volume averaged, rather than axially resolved, values. Several in vivo DRS studies have extracted
other clinically relevant optical parameters including blood volume fraction, hemoglobin concentration, oxygen
saturation, mean blood vessel diameter, nicotinamide adenine dinucleotide (NADH) concentration, and tissue
thickness34–37,48–52. Furthermore, DRS is an appealing non-invasive screening technique because it is sensitive to
optical changes beneath the apical tissue layer33–52. However, a drawback of DRS is inability to spatially resolve
tissue architecture.
We have recently reported on a probe-based technique that combines high-resolution microendoscopy
imaging, and a form of DRS called broadband sub-diffuse reflectance spectroscopy (sDRS) within a single
ber-bundle29,53. e term “sub-diuse reectance” is used here to be distinguished from “diuse reectance”
to describe the cases in which our source-detector separations (SDS) are less than one reduced mean-free path
within a sample, which will vary based on a samples optical properties40,54–58. is hybrid ber-bundle spectros-
copy and imaging probe is capable of co-registering qualitative high-resolution images of tissue surface microar-
chitecture with complimentary quantitative and depth-sensitive spectral data29,53. Furthermore, our design uses
two SDSs (shallow and deep channels) to collect data at two dierent sampling depths with the goal of sampling
dierent tissue volumes. erefore, the high-resolution imaging modality may be benecial in providing image
data of later stage moderate and severe dysplasia while the sDRS modality may be sensitive to tissue optical
changes associated with early dysplasia arising at the basement membrane29.
In this manuscript, we validate the sDRS portion of the quantitative hybrid imaging and spectroscopy
microendoscope and present a pilot phantom and pre-clinical study to extract in vivo optical parameters of the
human oral mucosa. First, a set of calibration phantoms was used to generate reectance lookup tables (LUT)
describing the relationship between reectance and optical properties (μ
s and μ
a) for the sDRS modality40.
en, to validate the LUT, the probe and LUT-based inverse model was used to extract μ
s and μ
a from a set of
hemoglobin-based validation phantoms with known μ
s and μ
a40. Extracted optical properties were compared
to theoretical values and reported as percent errors. Next, we quantify sampling depth for the shallow and deep
SDSs of the sDRS modality and validate results using the same calibration and validation phantoms59. Following
this, we present a simple phantom study simulating the physical layered progression from healthy tissue to severe
dysplasia to show how reectance changes with an optically scattering heterogeneity buried at various depths1,2,4.
Finally, the LUT-based inverse model was demonstrated on in vivo human oral mucosa from thirteen healthy
volunteers in a laboratory setting to determine volume-averaged scattering exponent, hemoglobin concentration,
oxygen saturation, and sampling depth. e extracted in vivo quantitative optical parameters were compared
to an in vivo high-resolution image of healthy, non-keratinized oral tissue. ese studies validate our hybrid
ber-bundle imaging and spectroscopy technique and demonstrate the translational potential to a clinical setting.
is technique can potentially be used to for diagnostic purposes as well as dynamically monitoring personalized
tumor response to therapy.
Materials and Methods
Instrumentation. e rst objective of this study was to design the multimodal instrumentation and associ-
ated contact ber-bundle probe to co-register qualitative image data with quantitative spectroscopy data29,53. For
the high-resolution uorescence imaging modality, a 455 nm LED (20 FWHM) light source (Philips, USA) is cou-
pled through a 1 mm-diameter image ber (FIGH-50-1100N, Myriad Fiber Imaging, USA) consisting of approx-
imately 50,000 individual 4.5 μ m-diameter bers. e distal blue 455 nm LED light excites a contrast agent, such
as proavine, which emits uorescence signal (peak emission of ~515 nm with quantum eciency of ~0.5) back
into the image ber and is delivered to an 8-bit monochrome CMOS camera (FL3-U3-32S2M-CS, Point Grey,
Canada)29,53,60. A lter set (Chroma Technology Corp., USA) separates the excitation and emission signals. Next,
for the sub-diuse reectance spectroscopy (sDRS) modality, a tungsten-halogen lamp (HL-2000-LL, Ocean
Optics, USA) delivers broadband light through a multimode 200 μ m (NA = 0.22) delivery ber (FBP200220240,
Molex Inc., USA) to a material. Four adjacent and identical 200 μ m (NA = 0.22) multimode bers, with center-
to-center source-detector separations (SDS) of 374, 730, 1,051, and 1,323 μ m, respectively, collect the broadband
sub-diusely reected light and deliver it to a spectrometer (USB2000+ UV-VIS, Ocean Optics, USA) with a
www.nature.com/scientificreports/
3
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
spectral resolution of 0.35 nm29. Although this technique is capable of having four SDSs, only two (374 and 730 μ m)
are presented in the following studies. Figure1 shows the instrumentation and probe design.
Generation of and validation of lookup tables for volume-averaged optical property extrac-
tion. e second objective of this study was to use the sDRS modality to extract volume-averaged optical
parameters. To accomplish this, reectance lookup tables (LUTs) were generated describing the relationship
between absolute reectance and optical properties (μ
s and μ
a) for the two SDSs (374 and 730 μ m). e target
ranges of the LUTs were μ
s and μ
a between 5–26 cm1 and 0–10 cm1, respectively. ese LUTs required calibra-
tion phantoms of similar order of magnitude as biological tissue40,61.
Calibration phantoms were constructed to exceed the target range using deionized water as the solvent40. e
scattering agent was 1.0 μ m-diameter polystyrene microspheres (07310-15, Polysciences, USA) and the associated
μ
s range (3–31 cm1) was calculated using Mie theory49,50,62. e absorbing agent was a combination of yellow,
red, and blue food dye (McCormick & Company, USA), in ratio of 20:6:2, which contained propylene glycol,
Yellow 5, Red 40, Red 3, Blue 1, and 0.1% propylparaben. e μ
a range (0–47 cm1) was calculated by measuring
the dye solution in deionized water using a spectrophotometer (5102-00, PerkinElmer, USA) and Beer’s Law. All
calibration phantoms were homogenous so μ
s and μ
a were identical throughout the phantom volume.
A total of 12 liquid calibration phantoms was created which was sucient to build the LUTs. Six of the 12
phantoms were considered “scattering-only” and contained only deionized water and polystyrene microspheres
without dye. Deionized water and polystyrene microspheres were gently mixed inside 7 mL scintillation vials
(66022-300, VWR, USA) to yield six μ
s ranges of 3.0–4.9, 4.4–7.1, 6.4–10.2, 9.2–14.7, 13.2–21.2, and 19.5–
31.0 cm1. ese values were chosen so there was sucient overlap between the maximum μ
s value of one phan-
tom at 450 nm and the minimum μ
s value of another phantom at 750 nm. Sucient overlap was determined such
that the minimum μ
s value of one phantom was no greater than 90% of the maximum μ
s value of the proceeding
phantom. is ensured the six scattering-only phantoms spanned a continuous μ
s range. e following equation
expresses this relationship in which n is the phantom number.
µµ
≤.
09 ,(1)
smin phantomn smax phantomn,, () ,, (1)
Figure 1. Representation of the hybrid ber-bundle imaging and spectroscopy system showing (a) the major
instrumentation components including (from le to right) ber switch, imaging portion, and spectroscopy
portion, (b) a SolidWorks representation of the distal probe (scale bar = 1 cm) showing the (c) en face view of
the central 1 mm-diameter image ber and 5 surrounding 200 μ m multimode bers (scale bar = 2.5 mm),
(d) distal probe (scale bar = 1 cm), and (e) en face view of the distal probe tip (scale bar = 2.5 mm).
www.nature.com/scientificreports/
4
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
e remaining six phantoms contained both polystyrene microspheres and the dye combination. Deionized
water, polystyrene microspheres, and dye were gently mixed inside 7 mL scintillation vials to yield a continuous
μ
s range of 3–31 cm1 and continuous μ
a range of 0–47 cm1. e wavelength-dependent variations in μ
s and μ
a
provide the wide range of scattering and absorbing values.
To generate the reectance LUTs, the probe was placed in each phantom so it was completely submerged
at a distance of 2 cm from the bottom of the 7 mL scintillation vial. Broadband sDRS data (450–750 nm) were
recorded at each SDS (374 and 730 μ m) with an integration time of 400 ms. Five spectra were averaged for all
measurements. Spectra were converted to absolute reectance values by calibrating with a Spectralon® 20% dif-
fuse reectance standard (SRS-20-010, Labsphere, USA) which was spectrally at between 200–2600 nm. All
spectra were corrected for background noise33,34,40,47,49. Aer acquiring absolute reectance spectra at a resolution
of 0.35 nm, the LUTs relating reectance (R) to μ
s and μ
a were generated using MATLAB. Raw data from the 12
calibration phantoms (C.P. #1-12) was interpolated to generate a color-mapped mesh with an optical property
resolution of 0.02 cm1. e reectance LUTs were interpolated in the target μ
s and μ
a ranges of 5–26 cm1 and
0–10 cm–1, respectively.
To validate the reectance LUTs, a set of liquid validation phantoms with known optical properties was built
of similar order of magnitude as biological tissue40,61. Validation phantoms were constructed in a similar manner
to calibration phantoms, but contained bovine hemoglobin (H2625, Sigma-Aldrich, USA), rather than food dye,
as the absorbing agent. e μ
s was calculated using Mie theory and μ
a was calculated by measuring a solution of
bovine hemoglobin in deionized water using a spectrophotometer (5102-00, PerkinElmer, USA) and Beer’s Law.
It was necessary to validate the LUTs using a dierent absorber and dierent scattering ranges than those used to
generate the LUTs so that the interpolated range of the LUTs were tested. All validation phantoms were homoge-
nous so μ
s and μ
a were identical throughout the phantom volume.
A 3 × 3 (9 total) set of validation phantoms was created, corresponding to three μ
s ranges and three μ
a ranges.
Deionized water, polystyrene microspheres and diluted bovine hemoglobin were gently mixed inside 7 mL scin-
tillation vials. is yielded μ
s values from 5–26 cm1 and μ
a values from 0–10 cm–1 to validate 100% of the reec-
tance LUTs. Figure2 shows the μ
s and μ
a for the calibration phantoms (C.P. 1-12) and validation phantoms
(V.P. 1–9).
Broadband sDRS data on validation phantoms were collected in the same method as the calibration phantoms.
e LUT-based inverse model was used to extract μ
s and μ
a from the validation phantoms. eoretical optical
properties of the validation phantoms were compared to extracted optical properties and reported as percent
errors. To quantify percent errors, the LUT-based inverse model extracted μ
s and μ
a for the 3 × 3 validation
phantoms at a spectral resolution of 0.35 nm and percent errors were calculated using the following formulas,
µµ
µ
=′−
µ
Error100%,
(2)
sextracted stheoretical
stheoretical
%,
,,
,
s
µµ
µ
=
µ
Error100%,
(3)
aextract ed atheoretical
atheoretical
%,
,,
,
a
Generation of and validation of lookup tables for sampling depth quantication. e third
objective of this study was to determine the sampling depth of the sDRS modality. To accomplish this, sam-
pling depth lookup tables (LUTs) were generated describing the relationship between sampling depth and
volume-averaged optical properties (μ
s and μ
a) for the two SDSs (374 and 730 μ m). e target ranges of the sam-
pling depth LUTs were μ
s and μ
a between 5–26 cm1 and 0–10 cm1, respectively. e same calibration phantoms
as described previously were used to generate the sampling depth LUTs.
Figure 2. Comparison of the optical properties of the (a,b) 6 × 2 (12 total) calibration phantoms (C.P.) and the
(c,d) 3 × 3 (9 total) validation phantoms (V.P.). Calibration phantoms were made with polystyrene microspheres
and a combination of yellow, red, and blue dye and the validation phantoms were made with polystyrene
microspheres and bovine hemoglobin as the scattering and absorbing agents, respectively. Calibration
phantoms had μ
s spanning 3–31 cm1 and μ
a spanning 0–47 cm1 and the validation phantoms had a μ
s
spanning 5–26 cm1 and μ
a spanning 0–10 cm1 to validate the target LUT range.
www.nature.com/scientificreports/
5
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
A highly absorbing phantom layer (μ
a 100 cm1 for all wavelengths between 450–750 nm) was created in a
5 mL beaker using 6.5% w/w India Ink in PDMS59,63. Contributions from specular reection were proven neg-
ligible by placing the probe in contact with the absorbing layer and acquiring sDRS data between distances of
0–2 mm in 50 μ m increments59.
Next, the six dye-containing calibration phantoms (Fig.2, C.P. 7–12) were placed on top of the highly absorb-
ing layer within the beaker. Spectra (450–750 nm) at each SDS were taken by varying the distance of the probe-tip
and absorbing layer between 0–2 mm in 50 μ m increments59. Sampling depth is been dened as the depth reached
by 50% of photons59. At a certain probe-absorbing layer distance (around 2 mm), there were no signicant
changes in signal intensity, meaning that nearly 100% of incident photons were not reaching the highly absorbing
layer. Figure3 shows how sampling depth was quantied for the sDRS modality in representative data59. As the
probe is translated away from the absorbing layer, as shown in Fig.3a, reectance increases until plateauing as
shown in Fig.3b. A depth (x-axis) can then be identied that correlates with the 50% cuto point (y-axis) which
is dened as the sampling depth as shown in Fig.3c 59.
e process from Fig.3 was repeated for all wavelengths at a spectral resolution of 0.35 nm for the 6 calibration
phantoms (C.P. 7–12). Raw data was interpolated in Matlab to generate a color-mapped mesh with a maximum
optical property resolution of 0.02 cm1. e sampling depth LUTs were interpolated in a target μ
s range of
5–26 cm1 and μ
a range of 0–10 cm1.
To validate sampling depth, spectra (450–750 nm) at each SDS of the previously described validation phan-
toms were acquired by varying the distance of the probe-tip and absorbing layer between 0–2 mm in 50 μ m incre-
ments. To quantify percent errors, sampling depths of the validation phantoms were compared to the sampling
depths (D) from the calibration phantoms. Percent errors were calculated using the following formula,
=Error
DD
D100%,
(4)
Dcalibrationvalidation
validation
%,
Semi-infinite phantom model of dysplastic progression. Once optical property extraction and
sampling depth were validated, we tested the capabilities of the sDRS modality of the hybrid ber-bundle in a
dysplasia-mimicking phantom model1. Figure4a–c shows a simplied representation of dysplastic progression
starting at the basement membrane and proliferating upwards into surrounding healthy tissue2,3. Early dysplasia
is known to signicantly increase epithelial scattering by nearly two-fold64–66. To simulate this phenomenon,
three solid scattering-only phantoms, shown in Fig.4d–f, were created1. Since scattering contributes much more
to reectance intensity compared to absorption, the μ
a was held constant at 0 cm1 66. Additionally, the phantom
“epithelia” was made to be 300 μ m thick to approximately simulate the thickness of oral mucosa67. With the
understanding that the 374 and 730 μ m SDSs sample dierent depths, it was expected that the 374 μ m SDS may
be more sensitive to shallower, epithelial-conned scattering changes associated with early dysplasia.
e three phantom models have a semi-innite geometry, a common geometry used in various models of
photon transport in tissues with sub-surface optical heterogeneities1. e semi-innite geometry requires an
optically thick base layer (bottom gray layer in Fig.4d–f) that can be considered innitely thick in the z direction
since no photons penetrate through this layer. In this experiment, the semi-innite base layer was 1 cm thick.
Additionally, all layers can be considered innite in the x and y directions since no photons penetrate laterally
outside this plane1.
Phantoms were created using poly(dimethylsiloxane) (PDMS) as the substrate material, and titanium diox-
ide (TiO2) as the scattering agent. PDMS was used because of its optical clarity (μ
s and μ
a = 0 cm1 between
500–750 nm), comparable refractive index to human tissue (~1.4), optical stability over time, physical durability,
Figure 3. e probe is placed (a) in contact with the highly absorbing (μ
a 100 cm1 for 450–750 nm) inside a
5 mL beaker and translated upwards in 50 μ m increments to (b) acquire sDRS data from a calibration phantom
(C.P. 11) at a 374 μ m SDS. (c) Representative data from the 374 μ m SDS shows the percentage of photons not
reaching the highly absorbing layer as a function of depth for C.P. 11 at 585 nm. Sampling depth is dened as the
depth reached by 50% of photons.
www.nature.com/scientificreports/
6
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
and ability to form multilayer geometries68. Since μ
s contributes to reectance intensity much more than μ
a, no
absorbing agent was used66.
e semi-innite layer and 150 μ m thick healthy tissue-mimicking layers were designed with 0.25% w/w TiO2
in PDMS (2.5 mg TiO2 per 1.0 g PDMS) to yield a μ
s of ~7 cm1 at 630 nm which is comparable to healthy tis-
sue68,69. e 150 μ m thick dysplasia-mimicking layers were designed with 0.50% w/w TiO2 in PDMS (5.0 mg per
1.0 g PDMS) to yield a μ
s of ~14 cm1 at 630 nm68,69. is represented a two-fold increase in scattering which is
representative of the increased scattering ratio of dysplastic to healthy epithelial tissue64–66. For each geometry
in Fig.4, two 150 μ m layers were stacked to generate the desired phantom67,68. e total phantom “epithelial”
thickness was thus 300 μ m, not including the “stromal” semi-innite base layer, which was 1 cm thick. All thin
phantom layers were created using a previously described spin coating technique67,68.
e volume-averaged μ
s was extracted between 500–750 nm for each phantom. Ten sDRS measurements were
averaged for each geometry (Phantoms 1–3) and SDS with an integration time of 500 ms. We hypothesized that
the 374 μ m SDS would show larger deviations in volume-averaged μ
s compared to the 730 μ m SDS because the
changes in scattering were conned to the upper 300 μ m of the phantom. e 730 μ m would be sampling signi-
cantly more into the underlying “stromal” semi-innite layer, in which μ
s was held constant for this experiment.
Results from this study were expected to indicate that the shorter SDS would be more sensitive to scattering
changes associated with dysplastic epithelia.
In vivo assessment of oral structural and optical properties. e nal objective of this study was to
extract optical parameters from in vivo oral mucosa and elucidate the dierences of the optical parameters for
each SDS (374 and 730 μ m). e multimodal technique was demonstrated in the inner lip of thirteen healthy vol-
unteers, with no history of tobacco use, between the ages of 18–35. Institutional Review Board approval (IRB #15-
09-149) was obtained from the Human Subjects Research program at the University of Arkansas for all aspects
of this study. e methods described were carried out in accordance with the approved guidelines, and informed
consent was obtained from all participants.
Extracting optical parameters required two steps. First, in vivo data acquisition was carried out with custom
LabVIEW soware29. e probe was directly placed in contact with the inner lip and broadband sDRS were
acquired at both SDSs (374 and 730 μ m). e tungsten-halogen lamp delivered 0.35 mW of power at the probe
tip for 500 ms. Additionally, in one volunteer, a single high-resolution uorescence image was taken using topical
proavine (0.01% w/v in saline) as a contrast agent with an exposure of 100 ms and gain of 5 dB, thus demon-
strating the capability of the probe to sequentially and non-invasively extract image and optical property data.
Second, for post-processing, raw broadband sDRS data was imported into custom MATLAB soware which was
Figure 4. A simplied representation of dysplastic proliferation arising at the basement membrane in the
oral cavity (a–c) showing normal cells (gray with nuclei), dysplastic cells (light gray with nuclei), basement
membrane (dark gray), and the stroma (gray). e associated dysplasia-mimicking phantom models
(d–f) simulate this progression. Two SDSs (374 and 730 μ m) deliver and collect broadband light at dierent
depths (detected photons shown here as blue and red crescents, respectively). Each of thin phantom layers was
150 μ m thick for a total phantom thickness of 300 μ m to simulate the thickness of oral epithelium.
www.nature.com/scientificreports/
7
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
integrated with the LUT-based inverse model and sampling depth LUT to extract optical parameters. e use of
this post-processing algorithm to extract optical parameters has been previously described34–36.
e optical parameters extracted in this study were volume-averaged scattering exponent (B), hemoglobin
concentration ([Hb]), and oxygen saturation (SaO2). Sampling depth was also quantied which is a function of
the underlying optical parameters35,40,51,59. e scattering exponent relates to the size of a tissue’s scattering par-
ticles, and thus can provide reasoning for changes in scattering when comparing groups within the same SDS70.
Hemoglobin concentration and oxygen saturation are commonly derived measurements in optical spectroscopy
to assess angiogenesis, and since blood vessel density has been shown to increase as oral tissue progresses from
normal to dysplastic, extracting these parameters was important71. ese optical parameters and their relation to
μ
s and μ
a are given in equations5 and 6. e μ
s was calculated based on the following equation,
µλ µλ λ
λ
=
() () ,
(5)
ss
B
0
0
where μs’(λ) is the reduced scattering coecient (cm1) at any wavelength, λ is a wavelength (nm), λ0 is 630 nm,
and B is the scattering exponent51. e μ
a was calculated based on the following equation,
µλ αε αε=. ⋅⋅
+−
() ()
Hb
MW
() 2303 []
1
(1 ),
(6)
aHb Hb
oxydeoxy
where μa is the absorption coecient (cm1) , [Hb] is the bulk tissue hemoglobin concentration (mg/mL), MW
is the gram molecular weight of hemoglobin which is assumed to be 64,500 g/mole72, α is the bulk tissue oxygen
saturation, and ε is the molar extinction coecient (cm1M1) of oxygenated hemoglobin (Hboxy) and deoxygen-
ated hemoglobin (Hbde-oxy). Some groups have also included a packaging correction factor when calculating μ
a
for sampling wavelengths below 450 nm, but this was shown to have no impact on the LUT-based inverse model
ts presented here since spectra were taken between 500–750 nm35.
Figure5 shows the experimental setup with the instrumentation, hybrid fiber-bundle probe, and
post-processing soware. For this experiment, it was hypothesized that the 730 μ m SDS would yield reduced B
values due longer SDSs having greater reectance from longer wavelengths. Alternatively, the 730 μ m SDS should
yield greater [Hb] values because of increased sampling into the sub-epithelia, where the blood vessels exist51,69.
SaO2 was expected to be comparable when sampling at dierent depths since changes in SaO2 have been shown
to not be depth dependent73. Finally, we expected increased sampling depth for the longer SDS51,59. Results from
this study were expected to show the value of including two dierent sub-diuse reectance spectroscopy SDSs
along with a high-resolution uorescence imaging capability.
Results
Generation of and validation of lookup tables for volume-averaged optical property extrac-
tion. Figure6a,b shows the reectance LUTs (μ
s = 5–26 cm1 and μ
a = 0–10 cm1) overlaid with the respective
reectance data from the dye-based calibration phantoms. Similarly, Fig.6d,e shows the reectance LUTs overlaid
with the respective data from the bovine hemoglobin-based validation phantoms. Validation phantom data that
perfectly overlays the LUT would indicate a 0% error; however, minor errors do exist. Additionally, Fig.6c,f shows
a ratio of the 730 to 374 μ m SDS LUTs. e mean ratio is 1.14, with a standard deviation of 0.27, indicating a var-
iable reectance ratio as μ
s and μ
a vary. Notice that at high reduced mean free paths (low μ
s and μ
a) in Fig.6c,f,
the reectance ratio is at a maximum of 1.69, and at low reduced mean free paths (high μ
s and μ
a), the reectance
ratio is at a minimum of 0.58. is trend supports the observation that longer SDSs are more sensitive to lower
Figure 5. An image of the experimental setup showing the optical instrumentation, post-processing soware
based in MATLAB showing a high-resolution uorescence image of the inner lip, LUT-based inverse model t
of raw reectance data, sampling depth, μs, and μa from one volunteer (image center), and the proximal and
distal hybrid ber-bundle probe.
www.nature.com/scientificreports/
8
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
scattering values, especially at longer wavelengths. Similarly, shorter SDSs are more sensitive to higher scattering
values. us, this reectance ratio trend supports the validity of our LUTs.
e LUT-based inverse model correctly estimated μ
s of the validation phantoms with average percent errors
of 1.6% and 2.5% for the 374 and 730 μ m SDS, respectively. Minimum and maximum percent errors for μ
s extrac-
tion were 0.1% and 5.3% for the 374 μ m SDS and 1.2% and 11.4% for the 730 μ m SDS, respectively. Additionally,
the LUT-based inverse model correctly estimated μ
a of the validation phantoms with average percent errors of
4.2% and 7.2% for the 374 and 730 μ m SDS, respectively. Minimum and maximum percent errors for μ
a extraction
were 2.1% and 18.4% for the 374 μ m SDS and 0.1% and 22.1% for the 730 μ m SDS, respectively.
Average percent errors were comparable to similar studies (< 10%) and considered acceptable34–36,40,42–44,49–51.
Thus, 100% of the optical property range of the LUTs were validated, and could be used to reliably extract
volume-averaged optical properties from unknown samples. Figure7 shows the ability of the reectance LUTs to
extract accurate μ
s and μ
a.
Figure 6. 100% (μs = 5–26 cm1, μa = 0–10 cm1) of both reectance LUTs were validated with acceptable
percent errors less than 10%. Following validation, optical properties can be reliably extracted from samples
with unknown optical properties using the LUT-based inverse model. (a,b) Reectance LUTs were interpolated
with raw data from calibration phantoms and (c) shows a ratio of the 730 μ m SDS to 374 μ m SDS LUTs.
(d,e) Reectance LUTs were validated with raw data from the bovine hemoglobin-based validation phantoms
and (f) shows the validated ratio of the 730 μ m SDS to 374 μ m SDS LUTs.
Figure 7. e LUT-based inverse model correctly estimated μs with average percent errors of 1.6% and 2.5%
for the 374 and 730 μm SDS, respectively, and correctly estimated μa with average percent errors of 4.2% and
7.2% for the 374 and 730 μm SDS, respectively. e ability to extract optical properties is shown with a perfect
t line.
www.nature.com/scientificreports/
9
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
Generation of and validation of lookup tables for sampling depth quantication. Sampling
depth ranged between 240 to 530 μ m and 300 to 680 μ m for the 374 and 730 μ m SDSs, respectively (Fig.8). In
both cases, maximum sampling depth occurred when μ
s and μ
a were 0 cm1 and minimum sampling depth
occurred at the maximum μ
s (26 cm1) and maximum μ
a (10 cm1) in the target range of the LUTs. Aer valida-
tion with hemoglobin-based validation phantoms, sampling depth was estimated with average percent errors of
1.9% and 1.6% for the 374 and 730 μ m SDS, respectively. Minimum and maximum percent errors for μ
s extrac-
tion were 1.8% and 5.3% for the 374 μ m SDS and 1.1% and 2.1% for the 730 μ m SDS, respectively. Average percent
errors, all under 2%, were considered acceptable in this study. Additionally, the ratio of sampling depths for the
730 to 374 μ m SDS were calculated for the entire LUT range (Fig.8c,f). On average, the sampling depth ratio was
1.20 with a standard deviation of 0.08, and relatively at as expected. is indicates the sampling depth of the
longer SDS is approximately 1.2× that of the shorter SDS across all wavelengths.
Extraction of sampling depth from semi-infinite phantom model of dysplastic progres-
sion. ree dierent phantom geometries, simulating the progression from healthy tissue to severe dysplasia,
underwent sDRS evaluation using both SDSs (374 and 730 μ m). Fig.9 shows that the extracted μ
s for phan-
tom 1 (blue line) was approximately 7 cm1 at 630 nm, as expected from the phantom generation protocol68.
As the higher scattering (μ
s = 14 cm1) layers proliferated upwards towards the probe tip (phantoms 2 and 3),
an increase in volume-averaged μ
s occurred for both SDSs, although more so for the shorter SDS, as expected.
For the shorter SDS, there was a signicant increase in volume-averaged μ
s from phantoms 1 to 2 and 2 to 3.
However, for the longer SDS, there was only a signicant increase in volume-averaged μ
s from phantoms 2 to 3.
is indicates the 374 μ m SDS is more sensitive to scattering heterogeneities at upper layers compared to the 730 μ
m SDS.
is phenomenon is further quantied in Table1 by the percent increase in volume-averaged μ
s at 630 nm
for Phantoms 1–3 for each SDS. e data indicates that the μ
s percent increase for the 374 μ m SDS is signi-
cantly greater compared to the 730 μ m SDS. is is because the shorter SDS has a decreased sampling depth, and
therefore scattering is mostly aected by more supercial heterogeneities, as seen in early dysplasia, compared to
the longer SDS. However, it is important to note that the 374 μ m SDS still does not exclusively sample the upper
layers, as indicated by the fact that the volume-averaged μ
s of phantom 3 (300 μ m thick heterogeneity) is approx-
imately 9 cm1 rather than 14 cm1 at the reference 630 nm. Additionally, sampling depth of the 374 μ m SDS at
a μ
s of 14 cm1 is ~400 μ m, indicating a sampling depth deeper than the 300 μ m scattering heterogeneity. ese
results demonstrate the value of including a shorter SDS for detection of more supercial scattering changes. e
value of including an additional longer SDS was shown in the following section describing in vivo results from
healthy human oral mucosa.
In vivo assessment of oral structural and optical properties. Thirteen volunteers underwent
data collection in the oral mucosa via the hybrid imaging and spectroscopy microendoscope (Fig.10). One
high-resolution uorescence image is presented in Fig.10a which shows the 1 mm-diameter image circle of the
image ber in direct contact with proavine-stained oral mucosa. Individual cell nuclei appear as distinct white
Figure 8. 100% (μs = 5–26 cm1, μa = 0–10 cm1) of both sampling depth LUTs were validated with acceptable
percent errors much less than 10%. (a,b) Sampling depth LUTs were interpolated with raw data from calibration
phantoms and (c) shows a ratio (1.2× ) of the 730 μ m SDS to 374 μ m SDS sampling depths. (d,e) Sampling depths
LUTs were validated with raw data from the bovine hemoglobin-based validation phantoms and (f) shows
the validated ratio of the 730 μ m SDS to 374 μ m SDS sampling depths.
www.nature.com/scientificreports/
10
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
spots in the image. Fig.10b shows representative absolute reectance data from both the 374 and 730 μ m SDS
from a single volunteer. Reectance is presented as black dots and the LUT-based inverse model (Fig.6) and
an established hemoglobin absorption spectrum72 was used to t the data via custom post-processing soware
based in MATLAB. e tted reectance is a function of the volume-averaged optical parameters, B, [Hb], and
SaO2 (Eqs5 and 6). ese values are presented as averages with standard deviations from the 13 volunteers in
Fig.10d–f and Table2. Sampling depth was quantied and presented in Fig.10c aer μ
s and μ
a were determined
using the LUT-based inverse model (Fig.7).
e 730 μ m SDS typically demonstrates increased reectance values, especially at wavelengths greater than
600 nm, indicating a greater contribution from the red and near-infrared region at larger source-detector separa-
tions. is phenomenon was responsible for the decreased B values at the longer SDS of 0.48 compared to 0.80 of
the shorter SDS as shown in Fig.10d. Average [Hb] was signicantly dierent at 2.39 and 2.91 mg/mL for the 374
and 730 μ m SDS, respectively (Fig.10e). ese values support our hypothesis and demonstrate increased [Hb]
for the longer SDS compared to the shorter SDS. Average SaO2 was not signicantly dierent at 94.1% and 91.7%
for the 374 and 730 μ m SDS, respectively (Fig.10f), supporting our hypothesis that oxygen saturation does not
signicantly vary with sampling depth. Finally, sampling depth ranged between 355 and 447 μ m for the 374 μ m
SDS and between 435 and 563 μ m for the 730 μ m SDS, with the sampling depth minima occurring at the rst
Figure 9. e volume-averaged μ
s (a,b) increased as the proliferating scattering heterogeneity moved upwards
towards the phantom surface (going from P1 to P3) showing a vertical line at 630 nm, in which percent increase
in volume-averaged μ
s was measured from. ere was a signicantly greater μ
s increase in these values for
the 374 μ m SDS compared to the 730 μ m SDS, indicating that the shorter SDS is more sensitive to supercial
scattering changes associated with early epithelial dysplasia.
Phantom
Comparison
374 μm SDS
(n = 10)
730 μm SDS
(n = 10)
P- Va lu e
Signicance (Y/N),
α = 0.01Mean St d. Dev. Mean St d. Dev.
P1 to P2 (%) 4.97 0.40 1.42 1.93 1.67 × 104Y
P2 to P3 (%) 16.18 5.95 9.19 1.54 4.58 × 103Y
P1 to P3 (%) 21.96 6.42 10.72 0.93 1.23 × 104Y
Table 1. Paired t-test statistics for percent increases in μs (λ = 630 nm) for dysplasia-mimicking phantom
model.
www.nature.com/scientificreports/
11
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
Q-band of hemoglobin at 542 nm and the sampling depth maxima occurring at the furthest tested wavelength at
750 nm. Complete paired t-test statistics for optical parameters are shown in Table2.
Discussion
We have demonstrated a hybrid spectroscopy and imaging probe capable of acquiring qualitative and quanti-
tative data by combining high-resolution microendoscopy and broadband sDRS. High-resolution ber-bundle
microendoscopy provides a highly resolved and magnied image of apical epithelial architecture in a small
1 mm-diameter eld-of-view while sDRS provides quantitative optical parameters of tissue in approximately the
same image region (Fig.1). By combining these two modalities, we can co-register qualitative image data and
quantitative spectral data within a single probe. Co-registration is important because this technique can be poten-
tially used to not only detect dysplasia using two dierent modalities, but also to monitor personalized response
of sub-surface dysplastic lesions to anti-tumor therapy at two dierent source-detector separations.
In this study, we designed two sets of liquid phantoms (Fig.2) to generate and validate a LUT-based inverse
model that was used to extract material optical parameters from raw sDRS data for each SDS (Fig.6). As of
the current report, the LUTs are valid for μ
s between 5–26 cm1 and μ
a between 0–10 cm1. ese ranges of
optical properties are sufficient to acquire accurate sDRS data for many tissue types between 500–750 nm.
Furthermore, our calibration and validation methods were optimized until all average percent errors were below
10% (Figs6and7), a benchmark error value comparable to many similar studies34–37,39,40,42–45,47,49–51.
In the same set of calibration phantoms (Fig.2), sampling depth was determined for each SDS59. A demon-
stration of calculating sampling depth was presented (Fig.3) and an empirical relationship was determined
Figure 10. Comparison of qualitative and quantitative data acquired by the hybrid imaging and spectroscopy
technique from 13 healthy volunteers showing (a) a high-resolution uorescence image of apical oral mucosa
from the inner lip of one volunteer (scale bar = 200 μ m), (b) representative absolute reectance proles showing
reectance data and the overlaid LUT-based inverse model ts from the same volunteer from (a,c) average
sampling depths for each SDS, (d) scattering exponent (B), (e) hemoglobin concentration ([Hb]), and
(f) oxygen saturation (SaO2). Error bars from (c–f) represent standard deviation.
Optical Proper ty
374 μm SDS (n = 13) 730 μm SDS (n = 13)
P- Va lu e
Signicance (Y/N),
α = 0.01Mean St d. Dev. Mean Std . D ev.
B 0.80 0.19 0.48 0.25 8.8 × 104Y
[Hb] (mg/mL) 2.39 0.44 2.91 0.65 8.8 × 103Y
SaO2 (%) 94.1 10.0 91.7 9.10 4.6 × 101N
Table 2. Paired t-test statistics for extracted in vivo oral optical properties from LUT-based inverse model.
www.nature.com/scientificreports/
12
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
for sampling depth as a function of μ
a and μ
s (Fig.8). Sampling depths were comparable to a similar study by
Hennessy et al.59.
Once the reectance LUTs (Fig.6) and sampling depth LUTs (Fig.8) were validated, a semi-innite phantom
model was used to simulate dysplastic progression in the oral mucosa (Fig.4)1–3. Results conrmed that the
shorter 374 μ m SDS was more sensitive to the scattering heterogeneity at supercial layers (Fig.9), where epithe-
lial dysplasia is known to have a profound eect on the scattering properties in such layers64–66. ese experiments
demonstrate the potential for monitoring scattering changes associated with early epithelial dysplasia which is
oen conned above the basement membrane1–4.
Next, the bench-top technique was applied to in vivo oral mucosa by collecting sDRS data from the inner lip
of 13 healthy volunteers (Fig.5). e LUT-based inverse model was used to extract the wavelength-dependent B,
[Hb], and SaO2 values from all 13 volunteers (Fig.10). e representative reectance data demonstrates increased
reectance for the 730 μ m SDS compared to the 374 μ m SDS at wavelengths greater than approximately 600 nm,
consistent with previous ndings29,74. It is well known that longer SDSs penetrate deeper into tissue, and thus
longer wavelengths will dominate reectance for longer SDSs51,59,74. is phenomenon is apparent when analyzing
the scattering exponent (B). At longer separations, B values decrease because of greater reectance from longer
wavelengths.
e extracted absorption-based optical properties, [Hb] and SaO2, were comparable to other studies35,75. e
longer 730 μ m SDS extracted greater [Hb] compared to the shorter 374 μ m SDS. is supports our hypothe-
sis that the longer SDS sampled deeper into the tissue vasculature, although it is clear the vasculature is still
being sampled with the 374 μ m SDS51,69,72. is penetration into the vasculature was expected since sampling
depth in the short SDS was greater than 300 μ m, which exceeds the non-vascularized epithelial thickness of the
oral cavity67. We anticipate the standard deviations for [Hb] and SaO2 values (Fig.10 and Table2) to be most
likely due to variations in the pressure applied between the probe tip and volunteer’s inner lip. It has been shown
that probe-pressure variations among measurements can induce large errors in [Hb] and SaO2, so future studies
will seek to develop a real-time probe-pressure monitoring system similar in concept to those reported in other
studies76.
e study presented here was an extensive validation of the quantitative spectroscopy modality of this tech-
nique. Since this technique has been validated, its ability to monitor tissue health in response to anti-tumor
therapy can be further evaluated in pre-clinical and clinical studies. Additionally, future studies will explore
quantitative measures regarding the high-resolution fluorescence imaging modality, such as automated
nuclear-to-cytoplasmic ratio and cells-per-area calculations, and co-register these values with sDRS extracted
optical parameters. Finally, since this hybrid imaging and spectroscopy technique lacks a wideeld imaging
modality, future trials will explore designing probes with identical probe-tip geometries that are compatible with
conventional endoscopes.
Conclusion
We have developed a hybrid spectroscopy and imaging technique comprising of a conventional uorescence
ber-bundle microendoscopy platform coupled with a series of o-axis broadband spectroscopy (sDRS) chan-
nels. Since dysplasia can initially arise near the epithelial basement membrane, collecting structural and func-
tional information from deeper within the tissue microenvironment is critical for many applications, including
detection of early dysplasia, analysis of tumorigenesis, and monitoring of therapeutic response. As a result,
this hybrid imaging and spectroscopy platform may be capable of collecting a wealth of information about the
structural and functional properties of tissue at various imaging sites in ex vivo and in vivo models. Finally, the
potential of this technique to be coupled to the biopsy port of a conventional endoscope makes further clinical
translation and complimentary optical biopsy in the oral cavity and other epithelial tissues feasible.
References
1. Zhu, C. & Liu, Q. Validity of the semi-innite tumor model in diuse reectance spectroscopy for epithelial cancer diagnosis: a
Monte Carlo study Optics Express 19, 17799–17812, doi: 10.1364/OE.19.017799 (2011).
2. Speight, P. M. Update on oral epithelial dysplasia and progression to cancer. Head and Nec Pathology 1, 61–66, doi: 10.1007/s12105-
007-0014-5 (2007).
3. Warnaulasuriya, S., eibel, J., Bouquot, J. & Dabelsteen, E. Oral epithelial dysplasia classication systems: predictive value, utility,
weanesses and scope for improvement. Journal of Oral Pathology & Medicine 37, 127–133, doi: 10.1111/j.1600-0714.2007.00584.x
(2008).
4. Bouquot, J. E., Speight, P. M. & Farthing, P. M. Epithelial dysplasia of the oral mucosa—Diagnosticproblems and prognostic features.
Current Diagnostic Pathology 12, 11–21 (2006).
5. Davies, . et al. Point of care optical diagnostic technologies for the detection of oral and oropharyngeal squamous cell carcinoma.
e Surgeon 13, 321–329, doi: 10.1016/j.surge.2015.06.004 (2015).
6. Cheng, Y. S., ees, T. & Wrig ht, J. Updates egarding Diagnostic Adjuncts for Oral Squamous Cell Carcinoma. Texas Dental Journal
132, 538–549 (2015).
7. Muldoon, T. J. et al. Noninvasive imaging of oral neoplasia with a high-resolution ber-optic microendoscope. Head & Nec 34,
305–312, doi: 10.1002/hed.21735 (2011).
8. Brailo, V. Dentist and early detection of oral carcinoma. Acta Medica Croatica 69, 45–48 (2015).
9. Broclehurst, P. et al. Comparative accuracy of dierent members of the dental team in detecting malignant and non-malignant oral
lesions. British Dental Journal 218, 525–529, doi: doi: 10.1038/sj.bdj.2015.344 (2015).
10. Shula, ., Abidi, W. M., ichards-ortum, . & Anandasabapathy, S. Endoscopic imaging: How far are we from real-time
histology? World Journal of Gastroenterology 3, 183–194, doi: 10.4253/wjge.v3.i10.183 (2011).
11. Pierce, M. C. et al. Accuracy of in vivo multi-modal optical imaging for detection of oral neoplasia. Cancer Prevention esearch 5,
801–809, doi: 10.1158/1940-6207 (2012).
12. Hur, C. et al. High-resolution microendoscopy for esophageal cancer screening in China: A cost-eectiveness analysis. Wor ld
Journal of Gastroenterology 21, 5513–5523, doi: 10.3748/wjg.v21.i18.5513 (2015).
13. Muldoon, T. J., Anandasabapathy, S., Maru, D. & ichards-ortum, . High-resolution imaging in Barrett’s Esophagus: a novel,
low-cost endoscopic microscope. Gastrointestinal Endoscopy 68, 737–744, doi: 10.1016/j.gie.2008.05.018 (2008).
www.nature.com/scientificreports/
13
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
14. Muldoon, T. J. et al. Evaluation of quantitative image analysis criteria for the high-resolution microendoscopic detection of neoplasia
in Barrett’s esophagus. Journal of Biomedical Optics 15, 026027, doi: 10.1117/1.3406386 (2010).
15. Shin, D. et al. Quantitative analysis of high-resolution microendoscopic images for diagnosis of esophageal squamous cell
carcinoma. Clinical Gastroenterology and Hepatology 13, 272–279, doi: 10.1016/j.cgh.2014.07.030 (2015).
16. Carns, J., eahey, P. A., Quang, T., Anandasabapathy, S. & ichards-ortum, . Optical molecular imaging in the gastrointestinal
tract. Gastrointestinal Endoscopy Clinics of North America 23, doi: 10.1016/j.giec.2013.03.010 (2013).
17. Chang, S. S. et al. High resolution microendoscopy for classication of colorectal polyps. Endoscopy 45, 553–559, doi: 10.1055/s-
0032-1326502 (2013).
18. Elahi, S. F., Miller, S. J., Joshi, B. & Wang, T. D. Targeted imaging of colorectal dysplasia in living mice with fluorescence
microendoscopy. Biomedical Optics Express 2, 981–986, doi: 10.1364/BOE.2.000981 (2011).
19. Louie, J. S., ichards-ortum, . & Anandasabapathy, S. Applications and advancements in the use of high-resolution
microendoscopy for detection of gastrointestinal neoplasia. Clinical Gastroenterology and Hepatology 12, 1789–1792, doi: 10.1016/j.
cgh.2014.08.004 (2014).
20. Parih, N. et al. In vivo diagnostic accuracy of high resolution microendoscopy in dierentiating neoplastic from non-neoplastic
colorectal polyps: a prospective study. e American Journal of Gastroenterology 109, 68–75, doi: 10.1038/ajg.2013.387 (2014).
21. Prieto, S. P. et al. Qualitative and quantitative comparison of colonic microendoscopy image features to histopathology. Proceedings
of SPIE 9328, doi: 10.1117/12.2079816 (2015).
22. Pierce, M. C. et al. A pilot study of low-cost, high-resolution microendoscopy as a tool for identifying women with cervical
precancer. Cancer Prevention esearch 5, 1273–1279, doi: 10.1158/1940-6207.CAP-12-0221 (2012).
23. Quinn, M. . et al. High-resolution microendoscopy for the detection of cervical neoplasia in low-resource settings. Plos One 7, doi:
10.1371/journal.pone.0044924 (2012).
24. Campbell, A. P., Suberman, T. A., Buchman, C. A., Fitzpatric, D. C. & Aduna, O. F. Flexible cochlear microendoscopy in the gerbil.
Laryngoscope 120, doi: 10.1002/lary.20979 (2010).
25. Levy, L. L., Jiang, N., Smouha, E., ichards-ortum, . & Siora, A. G. Optical imaging with a high resolution microendoscope to
identify cholesteatoma of the middle ear. Laryngoscope 123, 1016–1020, doi: 10.1002/lary.23710 (2013).
26. Monfared, A. et al. In vivo imaging of mammalian cochlear blood ow using uorescence microendoscopy. Otology & Neurotology
27, 144–152 (2006).
27. egunathan, . et al. Feasibility and preliminary accuracy of high-resolution imaging of the liver and pancreas using FNS
compatible microendoscopy. Gastrointestinal Endoscopy 76, 293–300, doi: 10.1016/j.gie.2012.04.445 (2012).
28. Ishijimi, A. et al. Automated frame selection process for high-resolution microendoscopy. Journal of Biomedical Optics 20, 46014,
doi: 10.1117/1.JBO.20.4.046014 (2015).
29. Greening, G. J. et al. Fiber-bundle microendoscopy with sub-diuse reectance spectroscopy and intensity mapping for multimodal
optical biopsy of stratied epithelium. Biomedical Optics Express 6, 4934–4950, doi: 10.1364/BOE.6.004934 (2015).
30. ivera, D. . et al. Compact and exible raster scanning multiphoton endoscope capable of imaging unstained tissue. PNAS 108,
17598–17603, doi: 10.1073/pnas.1114746108 (2011).
31. Chen, X., Xu, X., McCormic, D. T., Wong, . & Wong, S. T. C. Multimodal nonlinear endo-microscopy probe design for high
resolution, label-free intraoperative imaging. Biomedical Optics Express 6, 2283–2293, doi: 10.1364/BOE.6.002283 (2015).
32. Piyawattanametha, W. & Wang, T. D. MEMS-Based Dual Axes Confocal Microendoscopy. IEEE Journal of Selected Topics in
Quantum Electronics 16, 804–814, doi: 10.1109/JSTQE.2009.2032785 (2010).
33. Glennie, D. L., Hayward, J. E., Mcee, D. E. & Farrell, T. J. Inexpensive diuse reectance spectroscopy system for measuring
changes in tissue optical properties. Journal of Biomedical Optics 19, 105005, doi: 10.1117/1.JBO.19.10.105005 (2014).
34. ajaram, N. et al. Design and validation of a clinical instrument for spectral diagnosis of cutaneous malignancy. Applied Optics 49,
142–152 (2010).
35. ajaram, N., Gopal, A., Zhang, X. & Tunnell, J. W. Experimental validation of the eects of microvasculature pigment pacaging on
in vivo diuse reectance spectroscopy. Lasers in Surgery and Medicine 42, 680–688, doi: 10.1002/lsm.20933 (2010).
36. ajaram, N., eichenberg, J. S., Migden, M. ., Nguyen, T. H. & Tunnell, J. W. Pilot clinical study for quantitative spectral diagnosis
of non-melanoma sin cancer. Lasers in Surgery and Medicine 42, 716–727, doi: 10.1002/lsm.21009 (2010).
37. Bish, S. F. et al. Handheld Diuse eectance Spectral Imaging (DSi) for in-vivo characterization of sin. Biomedical Optics Express
5, 573–586, doi: 10.1364/BOE.5.000573 (2014).
38. arsten, A. E., Singh, A., arsten, P. A. & Braun, M. W. Dius e reectance spectroscopy as a tool to measure the absorption coecient
in sin: South African sin phototypes. Photochemistry and Photobiology 89, 227–233, doi: 10.1111/j.1751-1097.2012.01220.x (2013).
39. Hennessy, ., Lim, S. L., Marey, M. . & Tunnell, J. W. Monte carlo looup table-based inverse model for extracting optical
properties from tissue-simulation phantoms using diuse reectance spectroscopy. Journal of Biomedical Optics 18, 037003, doi:
10.1117/1.JBO.18.3.037003 (2013).
40. ajaram, N., Nguyen, T. H. & Tunnell, J. W. Looup table-based inverse model for determining optical properties of turbid media.
Journal of Biomedical Optics 13, 050501, doi: 10.1117/1.2981797 (2008).
41. arsten, A. E., Singh, A., arsten, P. A. & Braun, M. W. Diuse reectance spectroscopy as a tool to measure the absorption
coecient in sin: system calibration. Lasers in Surgery and Medicine 28, 437–444, doi: 10.1007/s10103-012-1079-2 (2013).
42. Vishwanath, . et al. Portable, Fiber-Based, Diuse eection Spectroscopy (DS) Systems for Estimating Tissue Optical Properties.
Applied Spectroscopy 62, 206–215 (2011).
43. Yu, B. et al. Cost-eective diuse reectance spectroscopy device for quantifying tissue absorption and scattering in vivo. Journal of
Biomedical Optics 13, 060505, doi: 10.1117/1.3041500 (2008).
44. Yu, B., Fu, H. L. & amanujam, N. Instrument independent diuse reectance spectroscopy. Journal of Biomedical Optics 16,
011010, doi: 10.1117/1.3524303 (2010).
45. Bish, S. F., ajaram, N., Nichols, B. & Tunnell, J. W. Development of a noncontact diuse optical spectroscopy probe for measuring
tissue optical properties. Journal of Biomedical Optics 16, 120505, doi: 10.1117/1.3662459 (2011).
46. Pimenta, S., Castanheira, E. M. S. & Minas, G. Optical microsystem for analysis of diuse reectance and uorescence signals
applied to early gastrointestinal cancer detection. Sensors 15, 3138–3153, doi: 10.3390/s150203138 (2012).
47. Nichols, B. S. et al. A quantitative diuse reectance imaging (QDI) system for compresensive surveillance of the morphological
landscape in breast tumor margins. Plos One 10, e0127525, doi: 10.1371/journal.pone.0127525 (2015).
48. Lim, L., Nichols, B., ajaram, N. & Tunnell, J. W. Probe pressure effects on human sin diffuse reflectance and fluorescence
spectroscopy measurements. Journal of Biomedical Optics 16, 011012, doi: 10.1117/1.3525288 (2011).
49. Nichols, B. S., ajaram, N. & Tunnell, J. W. Performance of a looup table-based approach for measuring tissue optical properties
with diuse optical spectroscopy. Journal of Biomedical Optics 17, 057001, doi: 10.1117/1.JBO.17.5.057001 (2012).
50. Sharma, M., Marple, E., eichenberg, J. S. & Tunnell, J. W. D esign and characterization of a novel multimodal ber-optic probe and
spectroscopy system for sin cancer applications. eview of Scientc Instruments 85, 083101, doi: 10.1063/1.4890199 (2014).
51. Sharma, M., Hennessy, ., Marey, M. . & Tunnell, J. W. Verication of a two-layer inverse Monte Carlo absorption model using
multiple source-detector separation diuse reectance spectroscopy. Biomedical Optics Express 5, 40–53, doi: 10.1364/BOE.5.000040
(2013).
52. Hennessy, ., Marey, M. . & Tunnell, J. W. Impact of one-layer assumption on diuse reectance spectroscopy of sin. Journal of
Biomedical Optics 20, 27001, doi: 10.1117/1.JBO.20.2.027001 (2015).
www.nature.com/scientificreports/
14
Scientific RepoRts | 6:26734 | DOI: 10.1038/srep26734
53. Greening, G. J. et al. Design and validation of a diuse reectance and spectroscopic microendoscope with poly(dimethylsiloxane)-
based phantoms. Proceedings of SPIE 9332, doi: 10.1117/12.2076300 (2015).
54. Durduran, T., Choe, ., Baer, W. B. & Yodh, A. G. Diuse optics for tissue monitoring and tomography. eports on Progress in
Physics 73, 076701, doi: 10.1088/0034-4885/73/7/076701 (2010).
55. Dehghani, H., Srinivasan, S., Pogue, B. W. & Gibson, A. Numerical modelling and image reconstruction in diffuse optical
tomography. Philosophical Transactions of the oyal Society A 367, doi: 10.1098/rsta.2009.0090 (2009).
56. Subramanian, H., Pradhan, P., im, Y. L. & Bacman, V. Penetration depth of low-coherence enhanced bacscattered light in
subdiusion regime. Physical eview E: Statistical, Nonlinear, and So Matter Physics 75, 041914 (2007).
57. Turzhitsy, V., adosevich, A., ogers, J. D., Taove, A. & Bacman, V. A predictive model of bacscattering at subdiusion length
scales. Biomedical Optics Express 1, 1034–1046, doi: 10.1364/BOE.1.001034 (2010).
58. Turzhitsy, V., ogers, J. D., Mutyal, N. N., oy, H. . & Bacman, V. Characterization of light transport in scattering media at sub-
diusion length scales with Low-coherence Enhanced Bacscattering. IEEE Journal of Selected Topics in Quantum Electronics 16,
619–626, doi: 10.1109/JSTQE.2009.2032666 (2010).
59. Hennessy, ., Goth, W., Sharma, M., Marey, M. . & Tunnell, J. W. Eect of probe geometry and optical properties on the sampling
depth for diuse reectance spectroscopy. Journal of Biomedical Optics 19, 107002, doi: 10.1117/1.JBO.19.10.107002 (2014).
60. Prieto, S. P., Powless, A. J., Boice, J. W., Sharma, S. G. & Muldoon, T. J. Proavine Hemisulfate as a Fluorescent Contrast Agent for
Point-of-Care Cytology. Plos One 10, e0125598, doi: 10.1371/journal.pone.0125598 (2015).
61. Sandell, J. L. & Zhu, T. C. A review of in-vivo optical properties of human tissues and its impact on PDT. Journal of Biophotonics 4,
773–787 (2011).
62. Fu, H. L. et al. A low-cost, portable, and quantitative spectral imaging system for application to biological tissues. Optics Express 18,
12630–12645, doi: 10.1364/OE.18.012630 (2010).
63. Di Ninni, P., Martelli, F. & Zaccanti, G. e use of India in in tissue-simulating phantoms. Optics Express 18, 26854–26865, doi:
10.1364/OE.18.026854 (2010).
64. Arier, D. et al. Light scattering from normal and dysplastic cervical cells at dierent epithelial depths: nite-dierence time-domain
modeling with a perfectly matched layer boundary condition. Journal of Biomedical Optics 8, 484–494, doi: 10.1117/1.1578640
(2003).
65. Collier, T., Arier, D., Malpica, A., Follen, M. & ichards-ortum, . Determination of epithelial tissue scattering coecient using
confocal microscopy. IEEE Journal of Selected Topics in Quantum Electronics 9, 307–313, doi: 10.1109/JSTQE.2003.814413 (2003).
66. Clar, A. L., Gillenwater, A. M., Alizadeh-Naderi, ., El-Naggar, A. . & ichards-ortum, . Detection and diagnosis of oral
neoplasia with an optical coherence microscope. Journal of Biomedical Optics 9, 1271–1280, doi: 10.1117/1.1805558 (2004).
67. Greening, G. J., James, H. M. & Muldoon, T. J. Optical Phantoms: Diuse and Sub-diuse Imaging and Spectroscopy Validation. 1–37
(SPIE Spotlights, 2015).
68. Greening, G. J. et al. Characterization of thin poly (dimethylsiloxane)-based tissue simulating phantoms with tunable reduced
scattering and absorption coecients at visible and nearinfrared wavelength. Journal of Biomedical Optics 19, 115002, doi: 10.1117/1.
JBO.19.11.115002 (2014).
69. Chang, V. T. et al. Quantitative physiology of the precancerous cervix in vivo through optical spectroscopy. Neoplasia 11, 325–332
(2009).
70. Mourant, J. . et al. Mechanisms of light scattering from biological cells relevant to noninvasive optical-tissue diagnostics. Applied
Optics 37, 3586–3593, doi: 10.1117/1.JBO.19.3.037004 (1998).
71. Mourant, J. ., Marina, O. C., Hebert, T. M., aur, G. & Smith, H. O. Hemoglobin parameters from diuse reectance data. Jouornal
of Biomedical Optics 19, doi: 0.1117/1.JBO.19.3.037004 (2014).
72. Prahl, S. A. Optical Absorption of Hemoglobin, http://omlc.org/spectra/hemoglobin/ (1999) 01/01/2016.
73. Bezemer, . et al. Simultaneous multi-depth assessment of tissue oxygen saturation in thenar and forearm using near-infrared
spectroscopy during a simple cardiovascular challenge. Critical Care 13, doi: 10.1186/cc8003 (2009).
74. Mirabal, Y. N. et al. eectance spectroscopy for in vivo detection of cervical precancer. Journal of Biomedical Optics 7, 587–594, doi:
10.1117/1.1502675 (2002).
75. Amelin, A., Serenborg, H. J.,  oodenburg, J. L. & Witjes, M. J. Non-invasive measurement of the microvascular properties of non-
dysplastic and dysplastic oral leuoplaias by use of optical spectroscopy. Oral Oncology 47, 1165–1170, doi: 10.1016/j.
oraloncology.2011.08.014 (2011).
76. Yu, B., Shah, A., Nagarajan, V. . & Ferris, D. G. Diuse reectance spectroscopy of epithelial tissue with a smart ber-optic probe.
Biomedical Optics Express 5, 675–689, doi: 10.1364/BOE.5.000675 (2014).
Acknowledgements
is material is based on work supported by the National Institutes of Health (1R03-CA182052, 1R15-CA202662),
the National Science Foundation Graduate Research Fellowship Program (G.G., DGE-1450079), the Arkansas
Biosciences Institute, and the University of Arkansas Doctoral Academy Fellowship. Any opinions, ndings, and
conclusions or recommendations expressed in this material are those of the authors and do not necessarily reect
the views of the acknowledged funding agencies.
Author Contributions
G.J.G. designed the experimental procedures, prepared the experimental devices and soware, helped prepare all
gures, and draed the manuscript. H.M.J. and N.V. collected data for and assisted in preparing Figures 4 and 9.
M.K.D. and S.M.O. collected data for and assisted in preparing Figures 5 and 10. N.R. and T.J.M. contributed
equally to the overall supervision of the project and helped design experimental procedures. All authors have
given approval to the nal version of the manuscript.
Additional Information
Competing nancial interests: e authors declare no competing nancial interests.
How to cite this article: Greening, G. J. et al. Towards monitoring dysplastic progression in the oral cavity using
a hybrid ber-bundle imaging and spectroscopy probe. Sci. Rep. 6, 26734; doi: 10.1038/srep26734 (2016).
is work is licensed under a Creative Commons Attribution 4.0 International License. e images
or other third party material in this article are included in the article’s Creative Commons license,
unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material. To view a copy of this
license, visit http://creativecommons.org/licenses/by/4.0/
... Diffuse reflectance spectroscopy (DRS) is a non-invasive method which can be used to quantify volumetric total hemoglobin concentration (THC), tissue oxygen saturation (StO 2 ), and tissue scattering at or within accessible tissue sites [1][2][3][4][5][6][7][8][9][10]. This technique has been adapted for studies of tumor perfusion and response to therapy, since THC and StO 2 can be used to differentiate therapeutic responders from non-responders over the course of treatment [11][12][13]. ...
... The four physiological parameters were quantified by inputting raw DRS spectra into an experimental lookup-table (LUT)-based post-processing software with a priori values for oxygenated and deoxygenated hemoglobin extinction coefficients [8,35,36]. The software performed an iterative model fit, based on a standard damped least-squares nonlinear fitting method, on raw DRS data to quantify THC, StO 2 , HbO 2 , and μ s ' [8]. ...
... The four physiological parameters were quantified by inputting raw DRS spectra into an experimental lookup-table (LUT)-based post-processing software with a priori values for oxygenated and deoxygenated hemoglobin extinction coefficients [8,35,36]. The software performed an iterative model fit, based on a standard damped least-squares nonlinear fitting method, on raw DRS data to quantify THC, StO 2 , HbO 2 , and μ s ' [8]. Additionally, the chi- squared (Χ 2 ) value indicated goodness-of-fit between the model fit and raw DRS data; for this study, if Χ 2 values exceeded 1.0, data was rejected and re-acquired as this was likely due to user-induced movement artifacts during data collection (i.e. ...
Article
Diffuse reflectance spectroscopy (DRS) has been used in murine studies to quantify tumor perfusion and therapeutic response. These studies frequently use inhaled isoflurane anesthesia, which depresses the respiration rate and results in the desaturation of arterial oxygen saturation, potentially affecting tissue physiological parameters. However, there have been no controlled studies quantifying the effect of isoflurane anesthesia on DRS-derived physiological parameters of murine tissue. The goal of this study was to perform DRS on Balb/c mouse (n = 10) tissue under various anesthesia conditions to quantify effects on tissue physiological parameters, including total hemoglobin concentration, tissue oxygen saturation, oxyhemoglobin and reduced scattering coefficient. Two independent variables were manipulated including metabolic gas type (pure oxygen vs. medical air) and isoflurane concentration (1.5 to 4.0%). The 1.5% isoflurane and 1 L/min oxygen condition most closely mimicked a no-anesthesia condition with oxyhemoglobin concentration within 89% ± 19% of control. The time-dependent effects of isoflurane anesthesia were tested, revealing that anesthetic induction with 4.0% isoflurane can affect DRS-derived physiological parameters up to 20 minutes post-induction. Finally, spectroscopy with and without isoflurane anesthesia was compared for colon tumor Balb/c-CT26 allografts (n = 5) as a representative model of subcutaneous murine tumor allografts. Overall, isoflurane anesthesia yielded experimentally-induced depressed oxyhemoglobin, and this depression was both concentration and time dependent. Investigators should understand the dynamic effects of isoflurane on tissue physiological parameters measured by DRS. These results may guide investigators in eliminating, limiting, or managing anesthesia-induced physiological changes in DRS studies in mouse models. © 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement.
... Then, the LUT was used as an inverse model to fit measured spectral data and extract optical properties. [18][19][20] DRS data at each SDS represent a weighted average of physiological parameters collected at increasing depths. Therefore, a one-layer inverse experimental model was chosen to quantity volume-averaged, rather than layerspecific, physiological parameters without assuming precise thickness of overlying skin layers. ...
... The μ a was calculated by measuring a diluted solution of teal India ink in distilled water using a spectrophotometer (5102-00, PerkinElmer) and the Beer-Lambert Law. [18][19][20]26 A 5 × 3 (15 total) set of calibration phantoms was created, corresponding to five scattering ranges and three absorbing ranges (Fig. 3). Five of the 15 phantoms were considered "scattering-only" and contained only polystyrene microspheres without India ink. ...
... The LUT was considered accurate when average percent errors for μ 0 s and μ a were each <10%, a standard cutoff across the literature. 18 Spectra were converted to absolute diffuse reflectance values by calibrating with a Spectralon® 20% diffuse reflectance standard and background noise subtraction as previously described. ...
Article
Diffuse reflectance spectroscopy (DRS) is a probe-based spectral biopsy technique used in cancer studies to quantify tissue reduced scattering (μs') and absorption (μa) coefficients and vary in source-detector separation (SDS) to fine-tune sampling depth. In subcutaneous murine tumor allografts or xenografts, a key design requirement is ensuring that the source light interrogates past the skin layer into the tumor without significantly sacrificing signal-to-noise ratio (target of ≥15 dB). To resolve this requirement, a DRS probe was designed with four SDSs (0.75, 2.00, 3.00, and 4.00 mm) to interrogate increasing tissue volumes between 450 and 900 nm. The goal was to quantify percent errors in extracting μa and μs', and to quantify sampling depth into subcutaneous Balb/c-CT26 colon tumor allografts. Using an optical phantom-based experimental method, lookup-tables were constructed relating μa,μs', diffuse reflectance, and sampling depth. Percent errors were <10 % and 5% for extracting μa and μs', respectively, for all SDSs. Sampling depth reached up to 1.6 mm at the first Q-band of hemoglobin at 542 nm, the key spectral region for quantifying tissue oxyhemoglobin concentration. This work shows that the DRS probe can accurately extract optical properties and the resultant physiological parameters such as total hemoglobin concentration and tissue oxygen saturation, from sufficient depth within subcutaneous Balb/c-CT26 colon tumor allografts. Methods described here can be generalized for other murine tumor models. Future work will explore the feasibility of the DRS in quantifying volumetric tumor perfusion in response to anticancer therapies.
... The utility of SRDRS for quantitative optical characterization of tissues has been widely recognized [18][19][20][21][22]. The increased dimensionality of the DRS data yields increased information density for the unique determination of tissue optical properties, and the illumination/detection separations of SRDRS probes may be optimized for specific tissue applications to reduce noise and target specific interrogation depths [23,24]. ...
... The increased dimensionality of the DRS data yields increased information density for the unique determination of tissue optical properties, and the illumination/detection separations of SRDRS probes may be optimized for specific tissue applications to reduce noise and target specific interrogation depths [23,24]. These advantages have motivated several investigations into the endoscopic implementation of SRDRS for characterization of GI screening for dysplasia and cancer using fiber bundle probes for in-vivo characterization of stomach tissues [10], colon tissues [11], and oral tissues [22]. However, fiber bundle probes have several disadvantages for SRDRS, including low collection efficiency due to low numerical aperture (NA), low fill factor, and limited geometrical collection configurations (typically round) [25,26]. ...
Article
Early detection and surveillance of disease progression in epithelial tissue is key to improving long term patient outcomes for colon and esophageal cancers, which account for nearly a quarter of cancer related mortalities worldwide. Spatially resolved diffuse reflectance spectroscopy (SRDRS) is a non-invasive optical technique to sense biological changes at the cellular and sub-cellular level that occur when normal tissue becomes diseased, and has the potential to significantly improve the current standard of care for endoscopic gastrointestinal (GI) screening. Herein the design, fabrication, and characterization of the first custom SRDRS device to enable endoscopic SRDRS GI tissue characterization using a custom silicon (Si) thin film multi-pixel endoscopic optical sensor (MEOS) is described.
... Subsequently, the system was validated in the clinical environment and in vivo sub-DRS data were collected from the inner lip of 13 volunteers. This hybrid system was reported to be capable of gathering information on functional and structural properties of tissue [93]. ...
Article
Full-text available
Oral cancer is the 16th most common cancer worldwide. It commonly arises from painless white or red plaques within the oral cavity. Clinical outcome is highly related to the stage when diagnosed. However, early diagnosis is complex owing to the impracticality of biopsying every potentially premalignant intraoral lesion. Therefore, there is a need to develop a non-invasive cost-effective diagnostic technique to differentiate non-malignant and early-stage malignant lesions. Optical spectroscopy may provide an appropriate solution to facilitate early detection of these lesions. It has many advantages over traditional approaches including cost, speed, objectivity, sensitivity, painlessness, and ease-of use in clinical setting for real-time diagnosis. This review consists of a comprehensive overview of optical spectroscopy for oral cancer diagnosis, epidemiology, and recent improvements in this field for diagnostic purposes. It summarizes major developments in label-free optical spectroscopy, including Raman, fluorescence, and diffuse reflectance spectroscopy during recent years. Among the wide range of optical techniques available, we chose these three for this review because they have the ability to provide biochemical information and show great potential for real-time deep-tissue point-based in vivo analysis. This review also highlights the importance of saliva-based potential biomarkers for non-invasive early-stage diagnosis. It concludes with the discussion on the scope of development and future demands from a clinical point of view.
... Each DRS measurement resulted in a value for THC, StO 2 , HbO 2 , dHb (630 nm) [34], and a chi-square (χ 2 ) value. THC, StO 2 , HbO 2 , dHb were quantified by inputting raw DRS spectra into custom lookup-table (LUT)based MATLAB software with a priori values for oxygenated and deoxygenated hemoglobin extinction coefficients [35][36][37]. The software performed an iterative model fit (1 × 10 4 iterations) to the raw DRS data to quantify THC, StO 2 , HbO 2 , dHb. ...
Article
Full-text available
Background Immunotherapy in colorectal cancer (CRC) regulates specific immune checkpoints and, when used in combination with chemotherapy, can improve patient prognosis. One specific immune checkpoint is the recruitment of circulating monocytes that differentiate into tumor-associated macrophages (TAMs) and promote tumor angiogenesis. Changes in vascularization can be non-invasively assessed via diffuse reflectance spectroscopy using hemoglobin concentrations and oxygenation in a localized tumor volume. In this study, we examine whether blockade of monocyte recruitment via CCL2 (macrophage chemoattractant protein-1) leads to enhanced sensitivity of 5-fluorouracil (5-FU) in a CT26-Balb/c mouse model of CRC. It was hypothesized that the blockade of TAMs will alter tumor perfusion, increasing chemotherapy response. A subcutaneous tumor model using Balb/c mice injected with CT26 colon carcinoma cells received either a saline or isotype control, anti-CCL2, 5-FU, or a combination of anti-CCL2 and 5-FU. Results Findings show that 12 days post-treatment, monocyte recruitment was significantly reduced by approximately 61% in the combination group. This shows that the addition of anti-CCL2 to 5-FU slowed the fold-change (change from the original measurement to the final measurement) in tumor volume from Day 0 to Day 12 (~ 5 fold). Modest improvements in oxygen saturation (~ 30%) were observed in the combination group. Conclusion The findings in this work suggest that the blockade of CCL2 is sufficient in the reduction of TAMs that are recruited into the tumor microenvironment and has the ability to modestly alter tumor perfusion during early-tumor response to treatment even though the overall benefit is relatively modest.
... In the case of biological tissues, hemoglobin, water or even proteins or other pigments can dominate the spectral response [27]. Hemoglobin peaks and valleys are usually around 425 and 555 nm for the deoxygenated state, and around 410, 540 and 575 for the oxygenated one [7], proteins present another peak at 280 nm [28], water shows peaks in the IR at 970 or 1197 nm [29], and lipids at 930 or 1210 nm [19]. Usually spectra coming from biological tissues present a decrease that starts around 650 nm [30]. ...
Article
Full-text available
Biological tissue identification in real clinical scenarios is a relevant and unsolved medical problem, particularly in the operating room. Although it could be thought that healthy tissue identification is an immediate task, in practice there are several clinical situations that greatly impede this process. For instance, it could be challenging in open surgery in complex areas, such as the neck, where different structures are quite close together, with bleeding and other artifacts affecting visual inspection. Solving this issue requires, on one hand, a high contrast noninvasive technique and, on the other hand, powerful classification algorithms. Regarding the technique, optical diffuse reflectance spectroscopy has demonstrated such capabilities in the discrimination of tumoral and healthy biological tissues. The complex signals obtained, in the form of spectra, need to be adequately computed in order to extract relevant information for discrimination. As usual, accurate discrimination relies on massive measurements, some of which serve as training sets for the classification algorithms. In this work, diffuse reflectance spectroscopy is proposed, implemented, and tested as a potential technique for healthy tissue discrimination. A specific setup is built and spectral measurements on several ex vivo porcine tissues are obtained. The massive data obtained are then analyzed for classification purposes. First of all, considerations about normalization, detrending and noise are taken into account. Dimensionality reduction and tendencies extraction are also considered. Featured spectral characteristics, principal component or linear discrimination analysis are applied, as long as classification approaches based on k-nearest neighbors (k-NN), quadratic discrimination analysis (QDA) or Naïve Bayes (NB). Relevant parameters about classification accuracy are obtained and compared, including ANOVA tests. The results show promising values of specificity and sensitivity of the technique for some classification algorithms, even over 95%, which could be relevant for clinical applications in the operating room.
... This variability is due to anatomical differences in the sample, instrument variation and/or slight variations in the position of the sample with respect to the spectroscopic system. This effect has been experimentally proved [11,12]. An adequate quantification and statement of a non-random character of the magnitude of this spectral variability is essential for any optical diagnostic technique based on DRS. ...
Article
and discrimination of biological tissues. Discrimination is based on massive measurements that conform training sets. These sets are then used to classify tissues according to the biomedical application. Classification accuracy depends strongly on the training dataset, which typically comes from different samples of the same class, and from different points of the same sample. The variability of these measurements is not usually considered and is assumed to be purely random, although it could greatly influence the results. In this work, spectral variations within and between samples of different animals of ex-vivo porcine adipose tissue are evaluated. Algorithms for normalization, dimensionality reduction by principal component analysis, and variability control are applied. The PC analysis shows the dataset variability, even when a variability removal algorithm is applied. The projected data appear grouped by animal in the PC space. Mahalanobis distance is calculated for every group, and an ANOVA test is performed in order to estimate the variability. The results confirm that the variability is not random and is dependent at least on the anatomical location and the specific animal. The variability magnitude is significant, particularly if the classification accuracy is needed to be high. As a consequence, it should be taken generally into account in classification problems. © 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement.
Article
Subdiffusive reflectance captured at short source-detector separations provides increased sensitivity to the scattering phase function and hence allows superficial probing of the tissue ultrastructure. Consequently, estimation of subdiffusive optical parameters has been the subject of many recent studies focusing on lookup-table-based (LUT) inverse models. Since an adequate description of the subdiffusive reflectance requires additional scattering phase function related optical parameters, the LUT inverse models, which grow exponentially with the number of estimated parameters, become excessively large and computationally inefficient. Herein, we propose, to the best of our knowledge, the first artificial-neural-network-based inverse Monte Carlo model that overcomes the limitations of the LUT inverse models and thus allows efficient real-time estimation of optical parameters from subdiffusive spatially resolved reflectance. The proposed inverse model retains the accuracy, is about four orders of magnitude faster than the LUT inverse models, grows only linearly with the number of estimated optical parameters, and can be easily extended to estimate additional optical parameters.
Article
Sensitive methods that can enable early detection of dental diseases (caries and calculus) are desirable in clinical practice. Optical spectroscopic approaches have emerged as promising alternatives owing to their wealth of molecular information and lack of sample preparation requirements. In the present study, using multispectral fluorescence imaging, we have demonstrated that dental caries and calculus can be objectively identified on extracted tooth. Spectral differences among control, carious and calculus conditions were attributed to the porphyrin pigment content, which is a byproduct of bacterial metabolism. Spectral maps generated using different porphyrin bands offer important clues to the spread of bacterial infection. Statistically significant differences utilizing fluorescence intensity ratios were observed among three groups. In contrast to laser induced fluorescence, these methods can provide information about exact spread of the infection and may aid in long term dental monitoring. Successful adoption of this approach for routine clinical usage can assist dentists in implementing timely remedial measures.
Article
Full-text available
In an ongoing effort to address the clear clinical unmet needs surrounding breast conserving surgery (BCS), our group has developed a next-generation multiplexed optical-fiber-based tool to assess breast tumor margin status during initial surgeries. Specifically detailed in this work is the performance and clinical validation of a research-grade intra-operative tool for margin assessment based on diffuse optical spectroscopy. Previous work published by our group has illustrated the proof-of-concept generations of this device; here we incorporate a highly optimized quantitative diffuse reflectance imaging (QDRI) system utilizing a wide-field (imaging area = 17 cm(2)) 49-channel multiplexed fiber optic probe, a custom raster-scanning imaging platform, a custom dual-channel white LED source, and an astronomy grade imaging CCD and spectrograph. The system signal to noise ratio (SNR) was found to be greater than 40 dB for all channels. Optical property estimation error was found to be less than 10%, on average, over a wide range of absorption (μa = 0-8.9 cm(-1)) and scattering (μs' = 7.0-9.7 cm(-1)) coefficients. Very low inter-channel and CCD crosstalk was observed (2% max) when used on turbid media (including breast tissue). A raster-scanning mechanism was developed to achieve sub-pixel resolution and was found to be optimally performed at an upsample factor of 8, affording 0.75 mm spatially resolved diffuse reflectance images (λ = 450-600 nm) of an entire margin (area = 17 cm(2)) in 13.8 minutes (1.23 cm(2)/min). Moreover, controlled pressure application at the probe-tissue interface afforded by the imaging platform reduces repeated scan variability, providing
Article
Full-text available
Despite significant advances in treatment modalities, the 5 year survival rate in oral and oropharyngeal squamous cell carcinoma (SCC) is less than 60%. Clinical examination, white light endoscopy followed by blind biopsies and histopathological analysis remains the gold standard for diagnosis and surveillance. These modalities continue to have a limited diagnostic accuracy of less than 55%. Novel optical-based diagnostic methods are promising new technologies for improving both screening and detection of cancer. This review will discuss their role in oral and oropharyngeal cancer detection with particular emphasis on optical imaging in oral and oropharyngeal cancer diagnosis, including the use of surface enhanced Raman spectroscopy, optical coherence tomography, fluorescence diagnosis, confocal laser endomicroscopy, confocal reflectance microscopy and narrow band imaging. Aided by the use of differing wavelengths of light, these methods are capable of detecting physical and biochemical changes that precede and mirror malignant change within tissue. Our review of the currently utilized optical diagnostic modalities suggests the possibility of a cost effective, point of care diagnosis that could facilitate early detection, reduce healthcare costs and improve patient survival and quality of life. Copyright © 2015 Royal College of Surgeons of Edinburgh (Scottish charity number SC005317) and Royal College of Surgeons in Ireland. Published by Elsevier Ltd. All rights reserved.
Article
Full-text available
We present a portable, multimodal, nonlinear endo-microscopy probe designed for intraoperative oncological imaging. Application of a four-wave mixing noise suppression scheme using dual wavelength wave plates (DWW) and a polarization-maintaining fiber improves tissue signal collection efficiency, allowing for miniaturization. The probe, with a small 14 mm transversal diameter, includes a customized miniaturized two-axis MEMS (micro-electromechanical system) raster scanning mirror and micro-optics with an illumination laser delivered by a polarization-maintaining fiber. The probe can potentially be integrated into the arms of a surgical robot, such as da Vinci robotic surgery system, due to its minimal cross sectional area. It has the ability to incorporate multiple imaging modalities including CARS (coherent anti-Stokes Raman scattering), SHG (second harmonic generation), and TPEF (two-photon excited fluorescence) in order to allow the surgeon to locate tumor cells within the context of normal stromal tissue. The resolution of the endo-microscope is experimentally determined to be 0.78 µm, a high level of accuracy for such a compact probe setup. The expected resolution of the as-built multimodal, nonlinear, endo-microscopy probe is 1 µm based on the calculation tolerance allocation using Monte-Carlo simulation. The reported probe is intended for use in laparoscopic or radical prostatectomy, including detection of tumor margins and avoidance of nerve impairment during surgery.
Article
Full-text available
Colorectal cancer is the second leading cause of cancer deaths in the United States, affecting more than 130,000 Americans every year(1). Determining tumor margins prior to surgical resection is essential to providing optimal treatment and reducing recurrence rates. Colorectal cancer recurrence can occur in up to 20% of cases, commonly within three years after curative treatment. Typically, when colorectal cancers are resected, a margin of normal tissue on both sides of the tumor is required. The minimum margin required for colon cancer is 5 cm and for the lower rectum 2 cm. However, usually more normal tissue is taken on both sides of the tumor because the blood supply to the entire segment is removed with the surgery and therefore the entire segment must be removed. Anastomotic recurrences may result from inadequate margins. Pathologists look at the margins to ensure that there is no residual tumor and this is usually documented in the pathology report. We have developed a portable, point-of-care fiber bundle microendoscopy imaging system for detection of abnormalities in colonic epithelial microstructure. The system comprises a laptop, a modified fiber bundle image guide with a 1mm active area diameter and custom Lab VIEW interface, and is approved for imaging surgically resected colon tissue at the University of Arkansas for Medical Sciences. The microendoscopy probe provides high-resolution images of superficial epithelial histology in real-time to assist surgical guidance and to localize occult regions of dysplasia which may not be visible. Microendoscopy images of freshly resected human colonic epithelium were acquired using the microendoscopy device and subsequently mosaicked using custom post-processing software. Architectural changes in the glands were mapped to histopathology H&E slides taken from the precise location of the microendoscopy images. Qualitatively, glandular distortion and placement of image guide was used to map normal and dysplastic areas of the colonic tumor and surrounding region from microendoscopy images to H&E slides. Quantitative metrics for correlating images were also explored and were obtained by analyzing glandular diameter and spatial distribution as well as image texture.
Article
The 5-year survival rate for oral cancer (66%) is still one of the lowest among major human cancers, and delayed diagnosis until an advanced stage is thought to be the main factor contributing to this low survival rate. The detection and diagnosis of oral cancer is currently based on clinical visual examination and histopathological evaluation of a biopsy specimen. In response to the need for early detection of oral cancer, several diagnostic adjuncts have been developed and sold commercially over the years, including vital tissue staining, brush cytology, light-based visualization adjuncts, and the most recently developed test for salivary biomarkers for oral cancer. The purpose of this article is to review the current knowledge and research regarding these diagnostic adjuncts developed for early detection of oral cancer. Clinicians are best served by an awareness of the advantages and disadvantages of each adjunct, and to always consider and correlate with the clinical findings when interpreting the test results from these adjuncts.
Article
Oral cancer is one of the ten cancers with the worst prognosis. Higher mortality rate is due to the fact that the majority of cases are diagnosed at an advanced stage. Early stage of the disease causes no discomfort and symptoms start to present in later stages. Oral cancer can be detected in its earliest stage by simple examination of oral mucosa. Oral mucosa can be easily examined during routine dental examination. The examination itself is noninvasive, takes no more than one minute, does not require any special equipment, and poses no discomfort to the patient. Apart from better knowledge about the physiological and pathological characteristics of oral mucosa, dentist is in an ideal position to perform oral mucosa examinations every day, since he/she regularly sees the patient oral cavity. In Croatia, between 300 and 400 new cases of oral carcinoma per year are diagnosed at the sites that are visible during routine dental examination. Since oral cancer in its early stage does not cause any symptoms, dentists should be sensitized/motivated to perform mucosal as well as dental examinations in all patients irrespective of the patient's lack of complaint.
Article
Early detection of structural or functional changes in dysplastic epithelia may be crucial for improving long-term patient care. Recent work has explored myriad non-invasive or minimally invasive "optical biopsy" techniques for diagnosing early dysplasia, such as high-resolution microendoscopy, a method to resolve sub-cellular features of apical epithelia, as well as broadband sub-diffuse reflectance spectroscopy, a method that evaluates bulk health of a small volume of tissue. We present a multimodal fiber-based microendoscopy technique that combines high-resolution microendoscopy, broadband (450-750 nm) sub-diffuse reflectance spectroscopy (sDRS) at two discrete source-detector separations (374 and 730 μm), and sub-diffuse reflectance intensity mapping (sDRIM) using a 635 nm laser. Spatial resolution, magnification, field-of-view, and sampling frequency were determined. Additionally, the ability of the sDRS modality to extract optical properties over a range of depths is reported. Following this, proof-of-concept experiments were performed on tissue-simulating phantoms made with poly(dimethysiloxane) as a substrate material with cultured MDA-MB-468 cells. Then, all modalities were demonstrated on a human melanocytic nevus from a healthy volunteer and on resected colonic tissue from a murine model. Qualitative in vivo image data is correlated with reduced scattering and absorption coefficients.
Article
To study the cost-effectiveness of high-resolution microendoscopy (HRME) in an esophageal squamous cell carcinoma (ESCC) screening program in China. A decision analytic Markov model of ESCC was developed. Separate model analyses were conducted for cohorts consisting of an average-risk population or a high-risk population in China. Hypothetical 50-year-old individuals were followed until age 80 or death. We compared three different strategies for both cohorts: (1) no screening; (2) standard endoscopic screening with Lugol's iodine staining; and (3) endoscopic screening with Lugol's iodine staining and an HRME. Model parameters were estimated from the literature as well as from GLOBOCAN, the Cancer Incidence and Mortality Worldwide cancer database. Health states in the model included non-neoplasia, mild dysplasia, moderate dysplasia, high-grade dysplasia, intramucosal carcinoma, operable cancer, inoperable cancer, and death. Separate ESCC incidence transition rates were generated for the average-risk and high-risk populations. Costs in Chinese currency were converted to international dollars (I$) and were adjusted to 2012 dollars using the Consumer Price Index. The main outcome measurements for this study were quality-adjusted life years (QALYs) and incremental cost-effectiveness ratio (ICER). For the average-risk population, the HRME screening strategy produced 0.043 more QALYs than the no screening strategy at an additional cost of I$646, resulting in an ICER of I$11808 per QALY gained. Standard endoscopic screening was weakly dominated. Among the high-risk population, when the HRME screening strategy was compared with the standard screening strategy, the ICER was I$8173 per QALY. For both the high-risk and average-risk screening populations, the HRME screening strategy appeared to be the most cost-effective strategy, producing ICERs below the willingness-to-pay threshold, I$23500 per QALY. One-way sensitivity analysis showed that, for the average-risk population, higher specificity of Lugol's iodine (> 40%) and lower specificity of HRME (< 70%) could make Lugol's iodine screening cost-effective. For the high-risk population, the results of the model were not substantially affected by varying the follow-up rate after Lugol's iodine screening, Lugol's iodine test characteristics (sensitivity and specificity), or HRME specificity. The incorporation of HRME into an ESCC screening program could be cost-effective in China. Larger studies of HRME performance are needed to confirm these findings.