To reconstruct medical images from truncated radiographic projections of organs that are in motion during the data acquisition. <br />To develop solid theoretical bases for the development of robust and efficient algorithms. <br />To test and evaluate these methods in the context of interventional imaging, in orthopedics in the framework of pedicle screw insertion and in radiotherapy of lung tumours.
In interventional radiographic imaging, the problem of reconstructing patient information from truncated projections of moving organs often arises. This situation is due to the digital radiography detector usually being too small; is because of the need to reduce patient dose; and is because interventional radiography systems are slow compared to CT scanners: the acquisition of roughly one hundred radiographs generally takes several tens of seconds. This is the case, for example, in radiologic imaging for monitoring dosimetry during radiotherapy of the lung. The radiographic projections are acquired on the radiotherapy machine for several tens of seconds while the patient is breathing normally. In interventional radiographic imaging for monitoring pedicle screw insertion using a C-arm x-ray system, images of the vertebrae must be reconstructed from truncated projections of a patient who is breathing. <br /> <br />The objective of the DROITE project is to develop methods of ROI (Region-of-Interest) reconstruction from truncated projections of dynamic objects. DROITE addresses the mathematics of dynamic ROI tomography et the medical applications in radiotherapy and in orthopedic surgery. The project concerns geometric calibration of imaging systems (parameter estimation of acquisition geometries), ROI reconstruction methods in 2D and 3D, and analytic methods for compensation of motion-related deformations in tomography and their link to ROI reconstruction.
The reconstruction methods that we are considering are primarily analytic. Our main objective is to demonstrate the existence, uniqueness and stability of ROI reconstruction from truncated projections and from partial trajectories or trajectories modified due to the measured object been deformed during the acquisition, i.e. during the motion of the source along its trajectory.
Reconstruction with analytic compensation of new deformation classes.
We want to study the possibility of analytically compensating for local deformations. We initially assume that an interior part of the organ (a lung tumour) is translating inside a region of homogeneous attenuation. We will then generalize to other deformation classes, more general than simple translations.
ROI Reconstruction with analytic compensation of new deformation classes:
Initially we will generalize the ROI reconstruction methods to the case of organs which are deforming during the acquisition, for deformations for which we know how to analytically compensate when the projections are complete. We will then generalize these approaches to new classes of deformations such as those described in the previous paragraph.
Identification of geometric acquisition parameters and of the deformations.
We have begun working on consistency conditions for cone-beam projections. This work may contribute to the estimation of the dynamic deformation parameters and/or the estimation of geometric calibration parameters of the acquisition systems.
Preliminary results on Consistency Conditions for radiographic data in 3D cone beam geometry were obtaine
Validation in the medical context.
We will evaluate methods that we are developing in the framework of radiotherapy for lung cancer at the Leon Berard Centre in Lyon, and in the framework of interventional imaging in the orthopedics department at the CHU of Grenoble.
 Clackdoyle R, Desbat L. 2013. «Full cone-beam consistency conditions for sources on a plane.« Proceedings of The 12th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine. Lake Tahoe, CA. June 16-21, 2013. pp.253-256.
 Clackdoyle R, Desbat L. «Full data consistency conditions for cone-beam projections with sources on a plane.« Physics in Medicine and Biology (submitted July 2013)
DROITE, Dynamic and ROI Tomography, Theory and Experiments, is a fundamental research project. Computed tomography aims at reconstructing images of internal physical quantities (attenuation coefficients, radioactivity concentrations) from external measurements (X-ray projections, radiation detectors). To first order, these measurements can be mathematically modeled by the Radon transform: straight line integrals of the unknown function. The problem to solve is the reconstruction of a function from a set of line integrals. Solving this problem led to the development of CT medical scanners, PET and SPECT. However, open questions remain that restrict the use of such systems in certain circumstances:
In case of patient movement during the data acquisition, solving the inverse problem (even when the exact movement is known) is an open problem, except for some very particular motion classes.
The use of a relatively small detector compared to the patient size, and the need to minimize the X-ray dose to the patient, lead to the problem of reconstruction from truncated projections. This is a challenging open problem that will be addressed in the project. Over the past decade, much progress has been made on reconstruction from truncated data, but open questions still remain related to the interior problem. The interior problem is well known to have no unique solution, but the uniqueness question for the reconstruction of an interior region from truncated projections when some supplementary non-truncated projections are available, remains an open problem.
Finally, the third open question which we will attack is the combination of patient motion with truncated projections, for which virtually no results exist today.
The objective of DROITE is to make theoretical contributions to the field of dynamic CT (reconstruction of moving objects) and Regions Of Interest (ROI) reconstruction (reconstruction from truncated projections). The objectives are to solve mathematical problems arising from the open questions presented above (obtaining results on existence, uniqueness, and stability for dynamic ROI reconstruction), to develop the associated reconstruction algorithms, and to experimentally validate the results using simulated and real data.
DROITE is a project with strong interaction between mathematics and medical imaging. The mathematical problems under investigation are motivated by image-guided therapy at Centre Léon Bérard (CLB, Lyon) and the Centre d'Investigation Clinique - Innovation Technologique (CIC-IT, Grenoble University Hospital). Thus, the DROITE consortium brings together on one hand, research teams in the mathematics of tomography (TIMC-IMAG: Laurent Desbat, LHC: Rolf Clackdoyle, Catherine Mennessier), and on the other hand physicians and medical imaging experts (for image-guided radiotherapy: Myriam Ayadi, Simon Rit and David Sarrut at CLB/CREATIS; for image guided orthopeadic surgery: Ivan Bricault, Philippe Cinquin, Alexandre Moreau Gaudry at CIC-IT/TIMC-IMAG). They will participate on the experimental validation using medical data.
Therefore, the DROITE project addresses two priorities of the SNRI: numerical sciences society, with the mathematical analysis and simulation of imaging systems in the context of medical applications and health care.
DROITE was first submitted to the ANR last year. We have amended the project according to the evaluation remarks. First, the involvement of the medical partners has been reinforced and a new partner, CLB/CREATIS, has been introduced for expertise on image-guided radiotherapy at CLB. Furthermore, the project cost has been reduced by more than a factor of two (one partner has been removed, and the PhD grants will be found elsewhere).
Monsieur Laurent Desbat (Techniques de l’Ingénierie Médicale et de la Complexité - Informatique, Mathématiques et Applications de Grenoble) – firstname.lastname@example.org
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
UJF - Grenoble 1 Techniques de l’Ingénierie Médicale et de la Complexité - Informatique, Mathématiques et Applications de Grenoble
CREATIS - CNRS Centre de Recherche en Acquisition et Traitement de l'Image pour la Santé
LHC Laboratoire Hubert Curien
Help of the ANR 139,248 euros
Beginning and duration of the scientific project: March 2013 - 48 Months