Abstract
2013
Objectives A novel method improving the estimation of the attenuation map for future PET/MR systems using the anatomical reference from MRI and gamma attenuation from PET emission data is presented. Besides the unique benefits of this new hybrid modality, MRI unfortunately does not provide a direct measurement of attenuation coefficients for 511keV photons as CT, which is one of the key prerequisites for quantitative PET. Thus, several research groups are addressing this topic with segmentation, atlas-based registration and computer learning techniques using data bases with MR and corresponding CT images.
Methods We propose a new iterative reconstruction approach to simultaneously compute the activity and the attenuation distribution using the segmented MR image as geometrical reference. The main advantage over other mentioned approaches is the use of the true physical attenuation of the 511keV photons provided by the PET list mode data. The novel approach uses a discrete consistency condition from which a maximum likelihood estimation of the activity and attenuation is derived.
Results First results from simulated and measured clinical data, compared to reference data achieved using CT attenuation maps, indicate an excellent agreement between both correction techniques. High impact of scattered and accidental coincidences is discussed in detail. Moreover, it was observed that performing TOF reconstruction and previous region segmentation minimizes inherent crosstalk between the activity and attenuation estimation.
Conclusions The proposed hybrid method using PET and MR data leads to more robust and accurate attenuation correction which is a significant step towards quantitative PET/MR.
- © 2009 by Society of Nuclear Medicine