Intercomparison of MR-Informed Methods for PET Image Reconstruction

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review

Abstract

Numerous PET reconstruction methods incorporating MR information have been proposed in the literature that seek to utilise the shared PET-MR boundaries to supress noise and reduce the partial volume effect. The inclusion of prior MR information into the PET reconstruction algorithm can be achieved using either an MR weighted penalty e.g. maximum a posteriori expectation maximisation (MAPEM) [1] , or MR derived spatial basis functions to reparameterise the emission image e.g. kernel expectation maximisation (KEM) [2]. MR information can be included into MAPEM with a quadratic penalty function through MR dependent weighting factors. Prominent choices of these weighting factors include Bowsher [3] and Gaussian weights. The kernel method can also be used to determine the weighting factors for the MR-guided MAPEM quadratic penalty. The limitation of all of these aforementioned methods lies in their susceptibility to suppress PET unique features. Therefore, many have been extended to incorporate PET information into the weighting factors or basis functions, with current state of the art methods including methods which use an anato-functional quadratic prior [4] , and methods like hybrid KEM [5].

Original languageEnglish
Title of host publication2018 IEEE Nuclear Science Symposium and Medical Imaging Conference Proceedings (NSS/MIC)
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781538684948
ISBN (Print)9781538684955
DOIs
Publication statusPublished - 5 Sept 2019
Event2018 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2018 - Sydney, Australia
Duration: 10 Nov 201817 Nov 2018

Conference

Conference2018 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2018
Country/TerritoryAustralia
CitySydney
Period10/11/201817/11/2018

Fingerprint

Dive into the research topics of 'Intercomparison of MR-Informed Methods for PET Image Reconstruction'. Together they form a unique fingerprint.

Cite this