This work reports on a time-domain image reconstruction algorithm for fluorescence lifetime tomography exploiting information contained whole fluorescence TPSFs. The results show the possibility to distinguish 4 different fluorophores with different lifetimes.
We present for the first time the analytical solution for the simplified spherical harmonics equations with partial reflective boundary conditions for a point source inside a spherical homogenous turbid medium.
In this work, synthetic time-domain data are generated as if it were collected with a state-of-the-art multi-view experimental optical scanner developed in our group for small animal imaging, and used in a tomographic image reconstruction algorithm. The collected data comprises full time-dependent optical signals leaving the biological medium and acquired all around the medium. The diffuse optical tomography (DOT) algorithm relies on the time dependent parabolic simplified spherical harmonics (TD-pSPN) equations as the forward model to recover the 3D absorption and diffusion coefficient maps of the medium. The inverse problem is casted and solved as an iterative constrained optimization problem where an objective function determines the accuracy of the forward model predictions at each iteration. Time-dependent adjoint variables are introduced to accelerate the calculation of the gradient of the objective function. A three-dimensional case involving an absorption heterogeneity in a homogeneous medium is presented, reproducing practical situations encountered in our lab. The results support our hypothesis that accurate quantitative 3D maps of optical properties of biological tissues can be retrieved using intrinsic measurements obtained with our experimental scanner along with our DOT algorithm.
KEYWORDS: Luminescence, Positron emission tomography, Cameras, In vivo imaging, Molecules, Signal to noise ratio, Proteins, In vitro testing, Image segmentation, Signal detection
Multimodal imaging is quickly becoming a standard in pre-clinical studies, and new developments have already confirmed the strength of acquiring and analyzing parallel information obtained in a single imaging session. One such application is the introduction of an internal reference moiety (e.g. radioisotope) to an activatable fluorescent probe. One of the limitations of this approach consists of working at concentrations which are within the overlapping range of sensitivities of each modality. In the case of PET/Fluorescence imaging, this range is in the order of 10-9 nM. Working in epi-illumination fluorescence imaging at such concentrations remains challenging. Here, we present in vitro and in vivo detection limits of a new fluorescent compound.
We investigate the problem of retrieving the optical properties (absorption and scattering) of biological tissue from a set of optical measurements. A diffuse optical tomography (DOT) algorithm that incorporates constrained optimization methods is implemented. To improve image quality, the DOT algorithm exploits full time-domain data. The time-dependent parabolic simplified spherical harmonics equations (TD-pSP N ) are used as the forward model. Time-dependent adjoint variables are resorted to in the calculation of the gradient of the objective function. Several numerical experiments for small geometric media with embedded inclusions that mimic small animal imaging are performed. In the experiments, optical coefficient values are varied in the range of realistic values for the near-infrared spectrum, including high absorption values. Single and multiparameter reconstructions are performed with the diffusion equation and higher orders of the TD-pSP N equations. The results suggest the DOT algorithm based on the TD-pSP N model outperforms the DE, and accurately reconstructs optical parameter distributions of biological media both spatially and quantitatively.
The use of Intralipid 20% was recently suggested as a diusive reference standard for tissue simulating phantoms.
We herein extend previously obtained results to other fat emulsions, specically Intralipid 10%, Intralipid 30%,
Lipovenoes 10%, Lipovenoes 20%, Lipofundin S 10%, and Lipofundin S 20%. Of particular importance for
practical applications, our measurements carried out at a wavelength of 751 nm show the following features.
First, batch-to-batch variations of the optical properties of these products are less than 2 % similarly to Intralipid
20%. Second, the absorption coecient of Intralipid, Lipovenoes, and Lipofundin S are very similar and their
measured values are within the experimental errors. Third, the reduced scattering of Intralipid 20%, Lipovenoes
20%, and Lipofundin S 20% are similar and their measured values are within 5%. Forth, the reduced scattering
coecient of Intralipid 10% and Intralipid 30% can be scaled from that of Intralipid 20% with an error of 9%
and 2%, respectively. A similar scaling property is valid for Lipovenoes and Lipofundin S.
We present a diffuse optical tomography (DOT) algorithm for imaging the absorption distribution in a biological
tissue using time-domain optical measurements. The time-dependent parabolic simplified spherical polynomials
approximation of the radiative transfer equation (the TD-pSPN model) serves as the forward model. The DOT
algorithm is implemented using a nested analysis and design (NAND) method developed for solving constrained
optimization problems. Numerical experiments are provided for small geometry media to mimic small animal
imaging. In these experiments, the optical absorption coefficient value is varied within typical values found in
the near infrared range for biological tissues, including high absorption values. The results show good spatial
and quantitative reconstructions and support our TD-pSPN-based DOT algorithm as an accurate approach to
image absorption in biological media.
We recently developed a time-domain technique for localizing in 3D discrete fluorescent inclusions embedded
in a scattering medium. It exploits early photon arrival times (EPATs), that is the time of flight of early
arriving photons at a detector determined via numerical constant fraction discrimination. Our localization
technique requires the knowledge of the speed of propagation of diffuse light pulses in the turbid medium to
convert measured propagation times to distances. We have developed an experimental method for measuring
the speed of propagation of such pulses. We have shown that time differences between a reference detector
position and other positions around the medium allow finding the position of the inclusion. Our technique allows
localizing inclusions to millimeter precision in a thick 5 cm diameter turbid medium. Herein, we analyze the
stability of EPAT differences introduced above and propagation speeds with respect to changes in the medium's
optical properties for optical properties typical of biological tissues. As we target small animal imaging, we
concentrate on optical properties of mouse organs and tissues. Our objective is to determine bounds to be
expected on the precision that can be achieved when media properties can vary and determine the limits of
validity of our localization technique. Our results show that EPAT differences and propagation speeds obtained
by our approach can vary; these values depend on the medium. We study 5 kinds of mouse organs and tissues.
Propagations speeds are between 2.97 × 107ms-1 and 5.52 × 107ms-1. Thus, it becomes important to evaluate
the discrepancy between true geometrical distance differences and distances as obtained by our approach using a
constant propagation speed and the measurement of EPAT differences. It is such discrepancies that ultimately
determine the localization accuracy of our algorithm because if distance differences based on EPATs are far from
true distances, our algorithm although it has a certain tolerance will have to consider that. The distance error
and so the localization accuracy of our algorithm is between 2.5mm and 8.6mm.
The present work will serve in a diffuse optical tomography (DOT) scanner that we are developing for small
animal non-contact molecular imaging. We present a new method for deconvoluting time-domain signals for use
in DOT. Time-domain signals represents reemitted light intensity as a function of time when the medium is
excited by ultra-short laser pulses. Actually, each signal equals the convolution between the light propagation in
the medium and the impulse response of the detection system, so-called the instrument response function (IRF).
Moreover, Poisson noise present in the system has to be considered. Time-domain signals directly depend on
the optical properties of a medium and so contain additional information (compared to continuous-wave signals)
that should be exploited in reconstruction algorithms. As an advantage, our deconvolution method does not use
a priori information about the signal. It is important to remove the IRF and noise from measured signals in
order to keep only the true signal, which has a direct link to medium properties.
Current efforts in tissue engineering target the growth of 3D volumes of tissue cultures in bioreactor conditions.
Fluorescence optical tomography has the potential to monitor cells viability and tissue growth non-destructively
directly within the bioreactor via bio-molecular fluorescent labelling strategies. We currently work on developing
the imaging instrumentation for tissue cultures in bioreactor conditions. Previously, we localized in 3D
thin fluorescent-labelled capillaries in a cylindrically shaped bioreactor phantom containing a diffusive medium
with our time-of-flight localization technique. Here, we present our first reconstruction results of the spatial
distribution of fluorophore concentrations for labelled capillaries embedded in a bioreactor phantom.
The equation of radiative transfer (ERT) is generally accepted as the most accurate model for light propagation in biological tissues. The ERT is notoriously expensive to solve numerically. Recently, Klose and Larsen have approximated the time-independent ERT using the simplified spherical harmonics equations ( SPN approximation). In this work, we outline how to derive the SPN approximation of the time-dependent ERT and obtain the associated integro- partial differential equations involving temporal convolution integrals. No approximation is made as regards the time variable in our derivation. To simplify the numerical solution of these equations, we introduce a "memory function". We discuss the numerical solution for N = 1 in the 2D and homogeneous case. We provide time evolution maps of the solution and compare it with the diffusion approximation of the ERT. The findings presented here straightforwardly extend to 3D inhomogeneous media and for higher values of N. These more complicated cases along with further details will be reported elsewhere.
We introduce a novel approach for localizing a plurality of discrete
fluorescent inclusions embedded in a thick scattering medium using
time-domain (TD) experimental data. It relies on numerical constant fraction
discrimination (NCFD), a signal processing technique for extracting in a
stable manner the arrival time of early photons emitted by one or many
fluorescent inclusions from measured photons time of flight (TOF)
distributions. Our experimental set-up allows multi-view TD data acquisition
from multiple tomographic projections over 360 degrees without contact with
the medium. Fluorescence time point-spread functions (FTPSFs) are acquired
all around the medium with ultra-fast time-correlated single photon counting
(TCSPC) after short pulse laser excitation. From these FTPSFs, the early
photons arrival time (EPAT) of a fluorescent wavefront at a detector
position is extracted with our NCFD technique. The key to our localization
algorithm is to combine EPATs from several detection positions and
projections to form 3D surfaces. The digital analysis of the concavities of
the surfaces allows to find the 3D positions of an a priori unknown number
of fluorescent inclusions located in the medium. Indocyanine green (ICG;
absorption peak = 780nm, emission peak = 830nm) is used for the inclusions.
Various experiments were conducted, and we show localization results on
experimental data for up to 5 discrete inclusions distributed at arbitrary
positions in the medium. We expect to extend our method to continuous
distributions of fluorescence (rather than discrete inclusions) in a near
future.
We introduce an improved approach in the 3D localization of discrete fluorescent inclusions in a thick scattering
medium. Previously our approach provided accurate localization of a single inclusion, showing the potential for
direct time-of-flight fluorescence diffuse optical tomography. Here, we localize various combinations of multiple
fluorescent inclusions. We resort to time-domain (TD) detection of emitted fluorescence pulses after short pulse
laser excitation. Our approach relies on a signal processing technique, dubbed numerical constant fraction
discrimination (NCFD), for extracting in a stable manner the arrival time of early photons emitted by one or
many fluorescent inclusions from measured time-of-flight (TOF) distributions. Our experimental set-up allows
multi-view tomographic optical TD measurements over 360 degrees without contact with the medium. It uses an ultra-short pulse laser and ultra-fast time-correlated single photon counting (TCSPC) detection. Fluorescence time point-spread functions (FTPSFs) are acquired all around the phantom after laser excitation. From measured
FTPSFs, the arrival time of a fluorescent wavefront at a detector position is extracted with our NCFD technique.
Indocyanine green (ICG; absorption peak = 780nm, emission peak = 830nm) is used for the inclusions. Various
experiments were conducted with this set-up in a stepwise fashion. First, single inclusion experiments are
presented to provide background information. Second, we present results using two inclusions in a plane. Then,
we move on with two inclusions located in different planes. Finally, we show results with a plurality of inclusions
(>2) distributed at arbitrary positions in the medium. Using an algorithm we have developed and tested on the
acquired data, we successfully achieve to locate the inclusions. Here, results are obtained for discrete inclusions.
In a close future, we expect to extend our method to continuous fluorescence distributions.
Fluorescence optical diffuse tomography (fDOT) is of much interest in molecular imaging to retrieve information
from fluorescence signals emitted from specifically targeted bioprocesses deep within living tissues. An exciting
application of fDOT is in the growing field of tissue engineering, where 3D non-invasive imaging techniques are
required to ultimately grow 3D engineered tissues. Via appropriate labelling strategies and fluorescent probes,
fDOT has the potential to monitor culture environment and cells viability non-destructively directly within
the bioreactor environment where tissues are to be grown. Our ultimate objective is to image the formation
of blood vessels in bioreactor conditions. Herein, we use a non-contact setup for small animal fDOT imaging
designed for 3D light collection around the sample. We previously presented a time of flight approach using a
numerical constant fraction discrimination technique to assign an early photons arrival time to every fluorescence
time point-spread function collected around the sample. Towards bioreactor in-situ imaging, we have shown the
capability of our approach to localize a fluorophore-filled 500 μm capillary immersed coaxially in a cylindrically
shaped bioreactor phantom containing an absorbing/scattering medium representative of experiments on real
tissue cultures. Here, we go one step further, and present results for the 3D localization of thinner indocyanine
green labelled capillaries (250 μm and 360 μm inner diameter) immersed in the same phantom conditions and
geometry but with different spatial configurations (10° and 30° capillary inclination).
We discuss the fabrication of optical phantoms that will serve as calibration and test standards for a diffuse optical tomographic system developed in our laboratory. For the fabrication of a phantom's matrix, two materials are compared: polyester and epoxy resins. We investigate the capacity to use either of these two resins in the fabrication of a phantom's matrix with different shapes and thicknesses. For the absorbing agent we use India ink and for diffusing agent we choose a fine titanium dioxide (TiO2) powder. We test a collimated light measurements set-up to extract: the absorption coefficient μa of a purely absorbing medium, and the scattering coefficient μs of a purely diffusing medium. We also compare the way μa and μs add in a solution of absorbing and diffusing media.
We herein describe a time-of-flight (TOF) technique to localize in 3D the position of a small fluorophore-filled
inclusion immersed in a scattering medium. To achieve this, we exploit the arrival time of early excited and
fluoresced photons. This is an embodiment of fluorescence diffuse optical tomography (FDOT) which aims to
find the position of fluorescent heterogeneities in 3D in thick turbid media non-invasively via optical imaging
techniques. In Ref. 4, we gave a short review of previous work on the problem of localizing a fluorescent
inclusion via time-resolved measurements. This will not be discussed again here.
Recent advances in the design and fabrication of avalanche photodiodes (APDs) and quenching circuits for timecorrelated
single photon counting (TCSPC) have made available detectors with timing resolutions comparable
to microchannel plate photomultiplier tubes (MCP-PMTs). The latter, were until recently the best TCSPC
detectors in terms of temporal resolution (≤30ps). Comparable resolutions can now be obtained with TCSPC
APDs at a much lower cost. It should also be possible to manufacture APDs with standard electronics fabrication
processes in a near future. This will contribute to further decrease their price and ease their integration in
complex multi-channel detection systems, as required in diuse optical imaging (DOI) and tomography (DOT).
We present, to our knowledge for the first time, results which demonstrate that, despite their small sensitive area,
TCSPC APDs can be used in time-domain (TD) DOT and more generally in TD DOI. With appropriate optical
design of the detection channel, our experiments show that it is possible to obtain comparable measurements
with APDs as with PMTs.
We present an optical metrology system for measuring the outer shape of small animals in 3D using a stereo
camera pair. This system will be integrated into a non-contact small animal diffuse optical tomography (DOT)
scanner we are currently developing. The key feature of our approach is to use the same laser beam as that
for the tomographic measurements, thus considerably reducing system complexity. Moreover, the 3D data are
acquired simultaneously with the DOT measurements. Precise measurements (< 1 mm) are achieved via a novel
axis (rotational and translational) optical calibration technique allowing the acquisition of full 3D models. Our
approach allows to measure, rather than indirectly infer, the exact position where laser light is injected into the
animal, in contrast to other approaches. This is extremely useful information for the tomographic reconstruction
algorithm. 3D measurements of a reference shape and of a small animal are presented, showing the precision and
effectiveness of our system.
Currents eorts in tissue engineering (TE) are directed towards growing 3D volumes of tissues. In response to TE
needs, we are developing a non-invasive technique based on fluorescence diuse optical tomography (FDOT) to
image in 3D, via fluorescence labelling, the formation of micro-blood vessels in tissue cultures grown on biodegradable
scaolds in bioreactor conditions. In the present work, we use a non-contact FDOT setup developed for
small animal imaging for our measurements. We present experimental results showing the feasability to localize
a fluorophore-filled 500μm capillary immersed in a scattering medium contained in a cylindrically-shaped glass
tube. These conditions are representative of experiments to be carried on real tissue cultures. Time-resolved
scattering-fluorescence measurements are made via Time-Correlated Single Photon Counting (TCSPC) and we
use numerical constant fraction discrimination (NCFD) to obtain primary localization information from our
time-resolved data.
In previous work, we have introduced a numerical constant fraction discrimination (NCFD) technique for processing
time-resolved optical signals. It allows to extract, in a stable manner, the arrival time of early photons
emitted by a fluorescent inclusion embedded in a scattering medium. We showed experimentally that these
arrival times correlate quasi-linearly with inclusion depth. We now exploit this arrival time vs depth relationship
for inferring the inclusion position by way of a time of flight algorithm. The algorithm uses the relative arrival
times measured at several detector positions around the scattering medium with respect to a reference detector
position. The latter is chosen as that detector position for which the arrival time is shortest. Our approach
provides accurate inclusion localization, showing the potential of direct time-of-flight fluorescence diffuse optical
tomography.
Recent advances in tissue engineering (TE) aim to grow 3D volumes of tissue in bioreactor conditions. This has
proved to be a difficult task thus far, notably due to the lack of non-invasive diagnostic tools to monitor the
growth of a tissue and ensure its appropriate development. To fulfill part of this need, we currently develop a
non-invasive imaging technique based on fluorescence diffuse optical tomography (FDOT) to image in 3D, via
fluorescent tracers, processes relevant to tissue growth in a bioreactor. More particularly, here we are interested
in imaging the formation of micro-blood vessels in tissue cultures grown on biodegradable scaffolds. Blood vessels
are thought to play a fundamental role in tissue growth. Since a bioreactor possesses a known geometry (by
design), we propose an FDOT configuration that uses fiber optics brought in contact with the boundary of the
bioreactor to collect tomographic optical data. We describe an optical fibers-based set-up and experimental
measurements that demonstrate the possibility of localizing a fluorophore-filled 500&mgr;m capillary immersed in
a scattering medium contained in a cylindrically-shaped glass tube. These conditions are representative of
experiments to be carried on real tissue cultures. In our particular implementation, time-resolved scattering-
fluorescence measurements are made via time-correlated single photon counting. Numerical constant fraction
discrimination applied to our time-resolved data allows to extract primary localization information.
We introduce a novel non-contact fluorescence diffuse optical tomography (FDOT) approach for localizing a
fluorescent inclusion embedded in a scattering medium. It uses the time of flight of early photons arriving at several detector positions around the medium. It is a true and direct time-of-flight approach in that arrival
times are converted to distance. The arrival time of early photons is found via a recently introduced numerical
constant fraction discriminator applied to fluoresced photons time-of-flight distributions (fluorescence time pointspread
functions (FTPSFs)). Time-correlated single photon counting and an ultrafast photon counting avalanche
photodiode are used for measuring FTPSFs that form tomographic data sets. The FDOT localization algorithm
proceeds in two steps. The first determines the angular position of the inclusion as the average, over projections, of
angular detector positions with smallest arrival time. The second determines the inclusion's radial position based
on relative arrival times obtained at several detector positions within each tomographic projection relatively to
a reference detector position, the latter being that of shortest arrival time in the projection. The radial position
found minimizes the discrepancy between relative arrival times computed for several possible inclusion positions
and relative arrival times deduced from experimental data. Two methods are presented for this.
The design and fabrication of time-correlated single photon counting (TCSPC) avalanche photodiodes (APDs)
and associated quenching circuits have made significant progresses in recent years. APDs with temporal resolutions
comparable to microchannel plate photomultiplier tubes (MCP-PMTs) are now available. MCP-PMTs
were until these progresses the best TCSPC detectors with timing resolutions down to 30ps. APDs can now
achieve these resolutions at a fraction of the cost. Work is under way to make the manufacturing of TCSPC
APDs compatible with standard electronics fabrication practices. This should allow to further reduce their cost
and render them easier to integrate in complex multi-channel TCSPC electronics, as needed in diffuse optical
tomography (DOT) systems. Even if their sensitive area is much smaller than that of the ubiquitous PMT
used in TCSPC, we show that with appropriate selection of optical components, TCSPC APDs can be used in
time-domain DOT. To support this, we present experimental data and calculations clearly demonstrating that
comparable measurements can be obtained with APDs and PMTs. We are, to our knowledge, the first group using APDs in TD DOT, in particular in non-contact TD fluorescence DOT.
We introduce a novel approach for calibrating an axis of rotation in a 3D optical metrology system. The system
uses a stereo camera pair, along with rotation and translation stages for obtaining a 3D model of the surface
of small animals. The metrology system will be part of a fully non-contact diffuse optical tomography (DOT)
scanner for small animal imaging. The rotation axis calibration technique is based on measuring, with the stereo
pair, the 3D position of a small ball as it is moved by the rotation stage (turntable). Our system has the advantage
of using the tomograph's laser beam to measure the outer shape of the subject, thereby reducing overall system
complexity, and allowing simultaneous surface and DOT measurements. Additionnaly, the exact position where
laser light penetrates the animal is measured, while traditionally, this information is indirectly inferred with
less accuracy. This information plays an important role in a tomographic reconstruction algorithm. Our new
approach for the calibration of the rotation axis is compared to another technique we previously developed, where
a checkerboard pattern is tracked instead of a ball. We present measurements of a reference shape and a small
animal taken by our system.
We develop a novel approach to infer depth information about a small fluorophore-filled inclusion immersed in a scattering medium. It relies on time-resolved measurements of the time of flight distribution of emitted fluorescent photons after short pulse laser excitation. The approach uses a novel numerical constant fraction discrimination technique to assign a stable arrival time to the distribution's early photons. Our experimental results show a linear relationship between these arrival times and the position of the inclusion. This approach will serve as a useful technique in fluorescence diffuse optical tomography.
An autonomous approach for learning the colors of specific objects assumed to have known body spectral reflectances is developed for daylight illumination conditions. The main issue is to be able to find these objects autonomously in a set of training images captured under a wide variety of daylight illumination conditions, and to extract their colors to determine color space regions that are representative of the objects' colors and their variations. The work begins by modeling color formation under daylight using the color formation equations and the semi-empirical model of Judd, MacAdam and Wyszecki (CIE daylight model) for representing the typical spectral distributions of daylight. This results in color space regions that serve as prior information in the initial phase of learning which consists in detecting small reliable clusters of pixels having the appropriate colors. These clusters are then expanded by a region growing technique using broader color space regions than those predicted by the model. This is to detect objects in a way that is able to account for color variations which the model cannot due to its limitations. Validation on the detected objects is performed to filter out those that are not of interest and to eliminate unreliable pixel color values extracted from the remaining ones. Detection results using the color space regions determined from color values obtained by this procedure are discussed.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.