Image Reconstruction Software (TIRIUS Project)
("Open-Source" Version)
During my Ph. D. project, I have implemented many 2D and 3D image
reconstruction algorithms applied to Positron Emission Tomography (PET)
which include the following:
1. "Filtered-back Projection (FBP)"
2. "Maximum Likelihood Expectation Maximisation (MLEM)"
3. "Ordered Subsets Expectation Maximisation (OSEM)"
In
order to validate these algorithm implementations and other image reconstruction algorithm
which constitute the hearth of my Ph. D., I have developed a user
interface for PET image reconstruction. The user interface. The
reconstructed image is then displayed on the screen. Initially, two
different version of the user interface was maintain in order to
support both Windows and Linux OS environment. Works are currently in
progress to develop a new user interface based on Qt librairies which
will be compatible on both Windows and Linux system. This software have
been called TIRIUS which stand for "Tomographic
Image Reconstruction Interface of Université de Sherbrooke". The TIRIUS
software will be available shortly under a GPL licence. Other source
codes will also be available from this web page to allow for example to
read simulated data generated by the GATE Monte Carlo simulator
software and convert them in a suitable form for the TIRIUS software. For more details on the TIRIUS software, consult the "
TIRIUS Project" web pages.
(Commercial Version)
The commercial version of the TIRIUS software is in reality only a
testbench for demonstrating the high potential of the new ultra-fast 3D
image reconstruction algorithms that have been developped during my Ph.
D. project. These methods, which are based on cylindrical coordinate
images, takes advantages of all symmetries between the detector of an
imaging system to reduce the system matrix size and to accelereate by
many order the image reconstruction process. This strategy have been
applied to iterative image reconstruction methods like the MLEM and
OSEM algorithms and to direct methods based on the singular value
decomposition of the system matrix. These methods are currently under
the protection of a patent and are available for commercial applications. See the
Technology Offer pages for more informations.
Monte Carlo Simulations (using GATE)
According to some aspects of my
Ph. D. project, the use of Monte Carlo simulations applied to Positron
Emission Tomography (PET) problems have the following objectives:
- Acquired PET data of multiple image phantoms with camera having
different ring geometries. These MC acquisition data can than be used
to validate and evaluate the performances of the different image
reconstruction methods developped..
- Tool for deriving Monte
Carlo system matrix in order to include more physical aspect of the PET
acquisition process into the system matrix model..
I am
presently using the GATE software for simulating PET acquisition. The
GATE software is based on physical models developed in the Geant4
librairies. Both softwares could be downloaded freely from the web.
Moreover, I have developed several functions allowing the conversion of
the GATE list-mode acquisition data into histogram data files that can
be read and reconstructed using the TIRIUS software.I have realized
different GATE models of some PET cameras that can also be downloaded
freely on this web site.
Publications
1)
J.D. Leroux, D. Rouleau, C. Pepin, J. Cadorette, R. Fontaine and R.
Lecomte, Time determination of BGO-APD detectors by digital signal
processing for Positron Emission Tomography, Article accepté au Trans
on Nuclear Science.
2)
J.-D. Leroux, D. Rouleau, C. Pepin,
J.-B. Michaud, J. Cadorette, R. Fontaine and R. Lecomte, Time
Discrimination Techniques using Artificial Neural Networks for Positron
Emission Tomography, Article accepté au Trans on Nuclear Science.
3)
J.-D. Leroux, D. Rouleau, C. Pepin, J.-B. Michaud, J. Cadorette, R.
Fontaine and R. Lecomte, Time Discrimination Techniques using
Artificial Neural Networks for Positron Emission Tomography, Proc.
NSS/MIC, Rome, octobre 2004.
4)
J.D. Leroux, D. Rouleau, C.
Pepin, J. Cadorette, R. Fontaine and R. Lecomte, Time determination of
BGO-APD detectors by digital signal processing for Positron Emission
Tomography, Proc. IEEE NSS/MIC, Portland, October 2003.
5)
J.-D. Leroux, J.-P. Martin, F. Bélanger, D. Rouleau, C. Pepin, J.
Cadorette, R. Fontaine, R. Lecomte, Time determination by digital
signal processing with BGO-APD detectors in positron emission
tomography, Proc. IEEE/NPSS 13th Real time Conference, Montreal, May
2003.
Conferences
1) Oral presentation, Accelerated iterative image reconstruction methods
based on block-circulant system matrix derived from a polar image
representation, IEEE NSS-MIC Conference, Hawaii, USA, octobre 2007.
2) Poster presentation, Fast 3D image reconstruction method based on
SVD decomposition of a block-circulant system matrix, IEEE NSS-MIC
Conference, Hawaii, USA, octobre 2007.
3) Poster presentation, Fast, accurate and versatile Monte Carlo method for computing
system matrix, IEEE NSS-MIC Conference, Hawaii, USA, octobre 2007.
4) Oral presentation, Time Discrimination Techniques using Artificial
Neural Networks for Positron Emission Tomography, IEEE NSS-MIC
Conference, Rome, octobre 2004.
5) Oral presentation, Time
determination of BGO-APD detectors by digital signal processing for
Positron Emission Tomography, IEEE NSS-MIC, Portland, October 2003.
6) Oral presentation, Time determination by digital signal processing
with BGO-APD detectors in positron emission tomography, Proc. IEEE/NPSS
13th Real time Conference, Montreal, May 2003.
Patent
1)
Jean-Daniel Leroux, Vitali
Selivanov, Roger Lecomte et Réjean Fontaine,
Image reconstruction methods based on block circulant system matrices.
This patent, which is the property of the Bureau de Liaison Entreprise
Université (BLEU) de l'Uninversité de Sherbrooke, protect the use of
iterative and direct image reconstruction methods based on
block-circulant system matrices derived from cylindrical image
representations. This strategies allow to mitigate the two primary
factor which constrained the use of fully 3D image reconstruction
methods by reducing significantly the size of the system matrix and by
accelerating by many order the computational speed (
more details...).
