Volume 2 Supplement 1

Proceedings of the 4th PSMR Conference on PET/MR and SPECT/MR

Open Access

Attenuation correction for hybrid MR/PET scanners: a comparison study

  • Elena Rota Kops1,
  • Andre Santos Ribeiro2,
  • Liliana Caldeira1,
  • Hubertus Hautzel3,
  • Mathias Lukas4,
  • Gerald Antoch3,
  • Christoph Lerche1 and
  • Jon Shah1
EJNMMI Physics20152(Suppl 1):A38

https://doi.org/10.1186/2197-7364-2-S1-A38

Published: 18 May 2015

Attenuation correction of PET data acquired in hybrid MR/PET scanners is still a challenge. Different methods have been adopted by several groups to obtain reliable attenuation maps (mu-maps). In this study we compare three methods: MGH, UCL, Neural-Network. The MGH method is based on an MR/CT template obtained with the SPM8 software. The UCL method uses a database of MR/CT pairs. Both generate mu-maps from MP-RAGE images. The feed-forward neural-network from Juelich (NN-Juelich) requires two UTE images; it generates segmented mu-maps. Data from eight subjects (S1-S8) measured in the Siemens 3T MR-BrainPET scanner were used. Corresponding CT images were acquired. The resulting mu-maps were compared against the CT-based mu-maps for each subject and method. Overlapped voxels and Dice similarity coefficients, D, for bone, soft-tissue and air regions, and relative differences images were calculated. The true positive (TP) recognized voxels for the whole head were 79.9% (NN-Juelich, S7) to 92.1% (UCL method, S1). D values of the bone were D=0.65 (NN-Juelich, S1) to D=0.87 (UCL method, S1). For S8 the MHG method failed (TP=76.4%; D=0.46 for bone). D values shared a common tendency in all subjects and methods to recognize soft-tissue as bone. The relative difference images showed a variation of -10.9% - +10.1%; for S8 and MHG method the values were -24.5% and +14.2%. A preliminary comparison of three methods for generation of mu-maps for MR/PET scanners is presented. The continuous methods (MGH, UCL) seem to generate reliable mu-maps, whilst the binary method seems to need further improvement. Future work will include more subjects, the reconstruction of corresponding PET data and their comparison.

Authors’ Affiliations

(1)
Forschungszentrum Jülich GmbH
(2)
Imperial College London
(3)
Heinrich-Heine-University Düsseldorf
(4)
Technische Universitaet Muenchen

Copyright

© Rota Kops et al; licensee Springer. 2015

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.