King's College London

Research portal

Composite System Modelling for High Accuracy Brain PET Image Reconstruction using GATE

Research output: Chapter in Book/Report/Conference proceedingConference paper

Original languageEnglish
Title of host publication2016 IEEE Nuclear Science Symposium and Medical Imaging Conference
StatePublished - Nov 2016

Documents

  • gate_model_belzunce

    gate_model_belzunce.pdf, 481 KB, application/pdf

    18/04/2017

    Submitted manuscript

King's Authors

Abstract

High resolution and good quantification is needed in specific regions of the brain in a number of PET brain imaging applications. An improvement in the spatial resolution and in the quantification of the tracer uptake can be achieved by using statistical reconstruction methods with an accurate model of the scanner acquisition process. This model is represented by a system response matrix and needs to include all the factors that contribute to the degradation of the reconstructed images. Monte Carlo simulations are the best method to model the complex physical processes involved in PET, but they have an extremely high computational cost and the system matrix needs to be recomputed for every new scan. Furthermore, for 3D PET the system matrix can have billions of elements, therefore at present it is impossible to store in memory during the iterative reconstruction. Consequently, on-the-fly Monte Carlo modelling of the system matrix has been previously proposed by other authors, where a Monte Carlo simulation is used in the forward projector and a simpler analytic model in the backprojector. In this work, we propose a different approach, where a composite system matrix is used, with a complete Monte Carlo model computed with GATE for a small subregion of the field of view and a simpler analytic model for the voxels outside that region. We evaluated the feasibility of the method using 2D simulations of a striatum phantom and a brain phantom. For each case, a Monte Carlo system matrix was generated with GATE for a subregion centred in the striatum. The brain simulations were reconstructed using the proposed method and compared with the standard reconstruction used clinically, with and without resolution modelling. For the striatum phantom, the use of a GATE system matrix showed an improvement of the reconstructed image, where a better definition of the structures in the striatum region was observed. For the case of the brain phantom, where the composite system matrix is used, an improvement was also observed but more limited compared with the pure GATE system matrix.

Download statistics

No data available

View graph of relations

© 2015 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454