| Autor: |
Plumat, Jérôme |
| Přispěvatelé: |
UCL - SST/ICTM/ELEN - Pôle en ingénierie électrique, Macq, Benoît, De Vleechouwer, Christophe, Draye, Xavier, Labeau, Fabrice, Jacques, Laurent, Pizurica, Aleksandra |
| Jazyk: |
angličtina |
| Rok vydání: |
2012 |
| Předmět: |
|
| Popis: |
The medical imaging requires fast and accurate volume reconstruction. This may be used to evaluate, before any treatment, differences between previsions, formulated in an MRI scanner and the actual patient anatomy, captured with X-Ray shoots. Based on a set of input projections, the task is to define a value in each volume's position such that the reconstruction's projections are as close as possible to the inputs. In this thesis we investigate two volume reconstruction problems. The first is formulating in the Markov Random Field (MRF). The particular studied problem investigates the reconstruction with only two orthogonal projections. The MRF formulation makes possible to reconstruct a volume close to the target, even with this restricted set of projections. We present algorithmic solutions and graph formulations to reconstruct volumes with a known set of pixels efficiently. Also, this framework provides an efficient formulation to solve imaging recognition problems. We present a mathematical formalism and an algorithmic solution to extract biological roots efficiently. Then, we investigate another framework: Compressed Sensing. The presented solution aims to reconstruct an MRI volume efficiently with a minimal number of measurements. We present an evaluation of the quality of the reconstruction that uses only reconstruction's wavelet coefficients. This criterion makes possible to stop the acquisition process at constant quality, and to capture only the needed measurements. Also, in our particular scenario, we use the previous reconstructed slice as a prior knowledge and perform the reconstruction on a part of the support only. This improvement increases the quality of the reconstruction. All culminates in an efficient volume reconstruction with a dynamic restricted number of measurements. (FSA 3) -- UCL, 2012 |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
