Universidade de Lisboa Repositório da Universidade de Lisboa

Repositório da Universidade de Lisboa >
Faculdade de Ciências (FC) >
FC - Teses de Doutoramento >

Please use this identifier to cite or link to this item: http://hdl.handle.net/10451/1668

Title: Incorporating accurate statistical modeling in PET: reconstruction for whole-body imaging
Authors: Janeiro, Luis Manuel de Almeida Soares, 1972-
Advisor: Almeida, Pedro Miguel Dinis de, 1968-
Comtat, Claude
Keywords: Biofísica
Teses de doutoramento
Issue Date: 2007
Abstract: The thesis is devoted to image reconstruction in 3D whole-body PET imaging. OSEM ( Ordered Subsets Expectation maximization ) is a statistical algorithm that assumes Poisson data. However, corrections for physical effects (attenuation, scattered and random coincidences) and detector efficiency remove the Poisson characteristics of these data. The Fourier Rebinning (FORE), that combines 3D imaging with fast 2D reconstructions, requires corrected data. Thus, if it will be used or whenever data are corrected prior to OSEM, the need to restore the Poisson-like characteristics is present. Restoring Poisson-like data, i.e., making the variance equal to the mean, was achieved through the use of weighted OSEM algorithms. One of them is the NECOSEM, relying on the NEC weighting transformation. The distinctive feature of this algorithm is the NEC multiplicative factor, defined as the ratio between the mean and the variance. With real clinical data this is critical, since there is only one value collected for each bin the data value itself. For simulated data, if we keep track of the values for these two statistical moments, the exact values for the NEC weights can be calculated. We have compared the performance of five different weighted algorithms (FORE+AWOSEM, FORE+NECOSEM, ANWOSEM3D, SPOSEM3D and NECOSEM3D) on the basis of tumor detectablity. The comparison was done for simulated and clinical data. In the former case an analytical simulator was used. This is the ideal situation, since all the weighting factors can be exactly determined. For comparing the performance of the algorithms, we used the Non-Prewhitening Matched Filter (NPWMF) numerical observer. With some knowledge obtained from the simulation study we proceeded to the reconstruction of clinical data. In that case, it was necessary to devise a strategy for estimating the NEC weighting factors. The comparison between reconstructed images was done by a physician largely familiar with whole-body PET imaging.
Description: Tese de doutoramento em Biofísica, apresentada à Universidade de Lisboa através da Faculdade de Ciências, 2007
URI: http://hdl.handle.net/10451/1668
Appears in Collections:FC - Teses de Doutoramento

Files in This Item:

File Description SizeFormat
3601_Thesis%20-%20L.%20Janeiro.pdf2.11 MBAdobe PDFView/Open
Statistics
FacebookTwitterDeliciousLinkedInDiggGoogle BookmarksMySpaceOrkut
Formato BibTex mendeley Endnote Logotipo do DeGóis 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

 

  © Universidade de Lisboa / SIBUL
Alameda da Universidade | Cidade Universitária | 1649-004 Lisboa | Portugal
Tel. +351 217967624 | Fax +351 217933624 | repositorio@reitoria.ul.pt - Feedback - Statistics
DeGóis
  Estamos no RCAAP Governo Português separator Ministério da Educação e Ciência   Fundação para a Ciência e a Tecnologia

Financiado por:

POS_C UE