< Back to previous page

Publication

Spatially Variant Resolution Modelling for Iterative List-Mode PET Reconstruction

Journal Contribution - Journal Article

A spatially variant resolution modelling technique is presented which estimates the system matrix on-the-fly during iterative list-mode reconstruction. This is achieved by redistributing the endpoints of each list-mode event according to derived probability density functions describing the detector response function and photon acollinearity, at each iteration during the reconstruction. Positron range is modelled using an imagebased convolution. When applying this technique it is shown that the maximum-likelihood expectation maximisation (MLEM) algorithm is not compatible with an obvious acceleration strategy. The image space reconstruction algorithm (ISRA), however, after being adapted to a list-mode based implementation, is wellsuited to the implementation of the model. A comparison of ISRA and MLEM is made to confirm that ISRA is a suitable alternative to MLEM. We demonstrate that this model agrees with measured point spread functions and we present results showing an improvement in resolution recovery, particularly for off-centre objects, as compared to commercially available software, as well as the standard technique of using a stationary Gaussian convolution to model the resolution, for equal iterations and only slightly higher computation time.
Journal: IEEE Transactions on Medical Imaging
ISSN: 0278-0062
Issue: 7
Volume: 35
Pages: 1707 - 1718
Publication year:2016
BOF-keylabel:yes
IOF-keylabel:yes
BOF-publication weight:10
CSS-citation score:1
Authors from:Higher Education
Accessibility:Closed