Joint MR regularized reconstruction of activity and attenuation for PET-MRI

Abolfazl Mehranian, Habib Zaidi, Andrew J. ReaderResearch 2016

Abstract

<b>637</b> <h3><b>Objectives</b></h3> Joint reconstruction of activity and attenuation in time-of-flight (TOF) PET-MRI is a promising approach to improve the quantitative accuracy of PET-MR imaging. The aim of this study is to maximize the benefits of MR anatomical images in constraining the maximum likelihood reconstruction of activity and attenuation (MLAA). Specifically, MR images were exploited to anatomically guide the reconstruction of activity and constrain the estimation of attenuation maps from emission data. <h3><b>Methods</b></h3> A modified Lysaker-Osher-Tai (LOT) regularization was employed for MRI-guided anatomical PET image reconstruction, which was optimized using the alternating direction method of multipliers (ADMM) with Green’s one- step-late approximation. For attenuation estimation, an MR-constrained quadratic soft-tissue preference prior was employed to suppress large deviation of attenuation coefficients of soft-tissue from their expected value. For qualitative and quantitative evaluations, 3D simulations were performed for a clinical PET scanner with TOF resolution of 580 ps using a simulated PET-MRI brain phantom. The performance of the proposed MLAA algorithm exploiting both attenuation and activity regularizations was compared with the original MLAA and the MLAA with only attenuation regularization. <h3><b>Results</b></h3> The simulation results showed that the proposed modification of MLAA effectively suppresses noise in activity maps while preserving anatomical boundaries derived from MR images. Moreover, noise was substantially reduced in the estimated attenuation maps and bone attenuation coefficients were correctly estimated from emission data. Quantitative evaluations showed that the original MLAA, MLAA with attenuation regularization and the proposed MR-constrained MLAA algorithms resulted in -13 ± 25%, -12 ± 24% and -6 ± 15% bias in grey matter, respectively. <h3><b>Conclusions</b></h3> The proposed jointly-regularized MLAA algorithm was found promising in exploiting the anatomical MRI to improve the estimated activity and attenuation maps, leading to improved quantitative accuracy of TOF PET-MR imaging.
BibTeX
@inproceedings{mehranian2016joint,
  author = {Abolfazl Mehranian and Habib Zaidi and Andrew J. Reader},
  title = {{Joint MR regularized reconstruction of activity and attenuation for PET-MRI}},
  booktitle = {Research},
  year = {2016},
}

Loading related content...

Figure