Skip to Main content Skip to Navigation

Modeling and regularization in tomographic reconstruction for Compton camera imaging

Abstract : The Compton camera is an imaging device for SPECT (Single Particle Emission Computed Tomography) of increased sensitivity compared to the Anger camera as it does not require mechanical collimation. The goal of this thesis is to evaluate the improvements that Compton camera may bring for nuclear medicine applications, depending both on technological developments and data processing techniques, among which the tomographic reconstruction is currently a bottleneck. In Compton camera imaging, the acquisition model is based on the integral of the intensity of the source on conical shapes. Modeling the measurement uncertainties in the system matrix can strongly influence the result of the list mode MLEM iterative reconstruction algorithm. One of the contributions of this study is a more precise model validated by Monte Carlo simulation. Another contribution concerns regularization methods. We developed a total variation denoising algorithm for Poisson distributed data that we introduced in the MLEM reconstruction as a regularization step, which allows to improve the image quality in low-counts experiments. A total variation regularized EM reconstruction with PSF deconvolution in the image space is also proposed for ameliorating the conditioning of the inverse problem and restoring the resolution of reconstructed images. All the proposed methods were validated on Monte Carlo simulation.
Complete list of metadatas

Cited literature [193 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Thursday, July 16, 2020 - 12:14:21 PM
Last modification on : Tuesday, July 21, 2020 - 4:54:02 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02900652, version 1


Yuemeng Feng. Modeling and regularization in tomographic reconstruction for Compton camera imaging. Signal and Image processing. Université de Lyon, 2019. English. ⟨NNT : 2019LYSEI084⟩. ⟨tel-02900652⟩



Record views


Files downloads