Browsing by Author "Pedemonte, S"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
- ItemFramework for the construction of a Monte Carlo simulated brain PET–MR image database(Elsevier, 2014-01-11) Thomas, BA; Erlandsson, K; Drobnjak, I; Pedemonte, S; Vunckx, K; Bousse, A; Reilhac-Laborde, A; Ourselin, S; Hutton, BFSimultaneous PET–MR acquisition reduces the possibility of registration mismatch between the two modalities. This facilitates the application of techniques, either during reconstruction or post-reconstruction, that aim to improve the PET resolution by utilising structural information provided by MR. However, in order to validate such methods for brain PET–MR studies it is desirable to evaluate the performance using data where the ground truth is known. In this work, we present a framework for the production of datasets where simulations of both the PET and MR, based on real data, are generated such that reconstruction and post-reconstruction approaches can be fairly compared. © 2013 Elsevier B.V.
- ItemWhat approach to brain partial volume correction is best for PET/MRI?(Elsevier B.V., 2013-02-21) Hutton, BF; Thomas, BA; Erlandsson, K; Bousse, A; Reilhac-Laborde, A; Kazantsev, D; Pedemonte, S; Vunckx, K; Arridge, SR; Ourselin, SMany partial volume correction approaches make use of anatomical information, readily available in PET/MRI systems but it is not clear what approach is best. Seven novel approaches to partial volume correction were evaluated, including several post-reconstruction methods and several reconstruction methods that incorporate anatomical information. These were compared with an MRI-independent approach (reblurred van Cittert ) and uncorrected data. Monte Carlo PET data were generated for activity distributions representing both 18F FDG and amyloid tracer uptake. Post-reconstruction methods provided the best recovery with ideal segmentation but were particularly sensitive to mis-registration. Alternative approaches performed better in maintaining lesion contrast (unseen in MRI) with good noise control. These were also relatively insensitive to mis-registration errors. The choice of method will depend on the specific application and reliability of segmentation and registration algorithms. (c) 2012 Elsevier Science B.V.