| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
|
|||||||||
|
|
Instrumentation & Data Analysis: InstrumentationInstrumentation Posters |
1 Institute of Nuclear Medicine, UCL, London, United Kingdom
1707
Objectives: The desirability of multiplexing in SPECT acquisition remains controversial. Our aim was to demonstrate the noise/bias tradeoff when multiplexed (MX) data are used and to determine if reconstruction of MX data could be improved by addition of non-MX.
Methods: Simulations were performed for an idealised SPECT system consisting of a cylindrical detector-system of 200mm radius with a slit-slat collimator. The number of slits was varied from 10 to 30 to allow increasing percentage of overlap (quantified as %data overlap). MX and non-MX projections were generated with identical geometry for a brain phantom and reconstructed using OSEM. Quality was evaluated in terms of recovery coefficient and noise for 10 realisations. In addition a starting image based on non-MX reconstruction (at different noise levels) was introduced for MX reconstruction.
Results: With even small degree of MX reconstruction quality could at best equal that of non-MX data. However inclusion of a non-MX start image, even with <5% MX counts, improved the bias/noise tradeoff to exceed that of non-MX data.
Conclusions: Our results confirm work of others that suggests that MX can only deteriorate reconstruction. However we have demonstrated that even a small degree of non-MX data appears sufficient to constrain the MX reconstruction. This has important implications for system design.
Research Support: GE Healthcare, EPSRC, EC FP6
| ||||||||||||||||||||||||||||||||||||||
