Speaker
Dr
Lionel Hervé
(LETI - CEA Recherche Technologique)
Description
Near-infrared fluorescence-enhanced Diffuse Optical Tomography (fDOT) has been
proven to be an efficient tool for image reconstruction of the bio-distribution of
fluorescent markers. By labeling regions of interest such as tumors with target-
specific fluorescing molecular probes, fDOT enables both the three-dimensional (3D)
localization of the targeted areas and the quantization of the local concentration
of the fluorochromes.
Within this framework, the reconstruction algorithms use either analytical or
numerical forward solvers that are required to be fast as well as robust. Analytical
approaches are time efficient but are usually restricted to the examination of
objects with simple geometries (infinite, semi-infinite media, slabs, cylinders…),
leading to the necessity of using an index matching fluid for immersion of the
animal in order to satisfy this assumption; whereas numerical ones, such as the
Finite Element Method (FEM), are more versatile but are known to be time and memory
consuming.
The present work is motivated by the fact that one of the requirements for small
animal imaging is to get rid of the constraint of the index matching fluid and to
image the animal in free space. Here, a comparison between two different approaches,
analytical and numerical, for the establishment of an efficient forward solver for
using fDOT technique in a non contact geometry is presented. Both are adapted to the
use of the CCD technology for the detection.
The experiments have been performed with our laboratory made tomographer. The
optical system is composed of a laser source (690 nm, 8 mW) for illumination and a
CCD camera (Orca ER, Hamamatsu) for detection. The source is guided to the object
via an optical fiber. The movements of the scanning fiber are driven by two
translating plates (Microcontrol) and monitored with a computer. The CCD camera is
focused at the top surface of the object. For fluorescence light detection, a filter
(high pass RG9, Schott) is placed in front of the camera.
A 2.5cm wide half cylindrical solid phantom has been designed for the experiment. It
is composed with a mixture of epoxy resin (Solloplast), titanium dioxide powder
(Sigma-Aldrich) as the scatterers, and black ink as the absorber. The index of
refraction is estimated to be 1.54, the diffusion coefficient 10 cm-1, and the
absorption coefficient 0.2 cm-1. In order to easily introduce fluorescent
inclusions, four holes have been drilled, at different positions. To model the
fluorescent regions, we considered two thin glass tubes (external diameter: 3 mm,
internal diameter: 1 mm, length: 3 cm) filled with commercial fluorescent dyes
(Alexa750, 10 microM, Molecular Probes), introduced in the phantom at 12 mm from the
ground surface. Both are separated by a distance of 2 cm.
In this experiment, 12×9 different sources positions, equally spaced, and separated
by a distance of 3 mm have been considered, corresponding to a rectangular field of
view of approximately 3.3×2.5 cm2.
The numerical approach is based on the resolution of both the forward and the
adjoint problems by using the FEM. A segmentation-based method is adopted for
constraining the reconstructions with a classical ART algorithm.
In the present analytical treatment, the morphology of the object is determined
indirectly by an appropriate reconstruction of optical heterogeneities. As a
consequence, this accounts self-consistently for the errors produced by using an
analytical model.
An example of reconstructions obtained by using the two kind of forward solvers was
performed. There are no obvious differences between the results obtained with one
approach or the other. An even more surprising remark, if a close-up is made on
these results, is that those performed by using the analytical forward solver are
slightly better. This is due to the fact that this approach takes into account most
of the causes of the discrepancies between the actual object and the assumptions on
the geometry or on the homogeneity made. It accounts efficiently not only for the
morphology of the object but also for the possible presence of heterogeneities (in
this case the tubes) inside the object.
A complete description of the models and a precise comparison between the two
approaches will be presented at the conference.
We developed two viable techniques which address our goal of suppressing the
matching fluid to perform fluorescence tomography on small animal.
Author
Dr
Lionel Hervé
(LETI - CEA Recherche Technologique)
Co-authors
Dr
Anabela Da Silva
(LETI - CEA Recherche Technologique)
Mrs
Anne Frassati
(LETI - CEA Recherche Technologique)
Dr
Jean-Marc Dinten
(LETI - CEA Recherche Technologique)
Dr
Jérôme Boutet
(LETI - CEA Recherche Technologique)
Dr
Philippe Peltié
(LETI - CEA Recherche Technologique)
Dr
Philippe Rizo
(LETI - CEA Recherche Technologique)
