Abstract
1477
Objectives: The current spatial resolution of PET scanners is ~4mm for whole body imaging which depends on factors such as positron range, scatter, and detector element size. The PET image quality depends on factors such as injected radiotracer dose, its half-life and scan time. One way to control image noise and increase PET image quality and resolution is using penalized maximum likelihood algorithm based on the relative difference prior (RDP) with a block sequential regularized expectation maximization (BSREM) optimizer [1]. The RDP penalty is calculated based on the difference between each voxel activity and its neighboring voxels to control noise at higher iterations [2]. Our objective is to use a similar method using MR priors to further reduce noise and improve image quality.
Methods: Two subjects were injected with 5 mCi of 68Ga-RM2 and after 45 min uptake time, they underwent a 20 min prostate exam on SIGNA PET/MR (GE Healthcare, Waukesha). T2 CUBE and DWI were acquired simultaneously with PET. A seed PET image was reconstructed with TOF-OSEM with 28 subsets and 2 iterations. The normalized seed PET images and co-registered MR images were mapped to a 3D feature space where each voxel was represented by a point in the feature space with its PET and MR normalized values. An additional penalty, which is calculated based on the relative difference between each voxel and its neighboring voxels in the feature space was calculated and added to the BSREM reconstruction method. By using the seed PET images in the formation of the feature space, the MR images alone will not determine the penalty function and therefore the reconstruction will be immune to mismatches between the anatomical images and the true activity. The PET images were reconstructed with TOF-OSEM with 2.34×2.34×2.78 mm3 resolution (256×256 matrix on a 60 cm FOV) and using the MR guided BSREM (MRgTOF-BSREM) with isotropic 1.17 mm resolution (512×512 matrix on a 60 cm FOV).
Results: Figure 1 shows the fused PET and MR images of two subjects in axial, coronal and sagittal planes. The images reconstructed by TOF-OSEM have lower resolution and do not show the anatomical boundaries well in comparison to the MRgTOF-BSREM images, which are higher resolution with detailed anatomical boundaries.
Conclusions: When a penalized maximum likelihood algorithm based on the relative difference prior is combined with using MR priors in the feature space, it can control the image noise better and improve the signal to noise ratio (SNR) and image resolution. Because MRgTOF- BSREM uses the feature space and incorporates a seed PET image into the feature space, it is not vulnerable to mismatches between the MR images and true activity distribution. References: [1] Ahn S, Ross SG, Asma E, et al., “Quantitative comparison of OSEM and penalized likelihood image reconstruction using relative difference penalties for clinical PET,” Phys. Med. Biol. 2015; 60: 5733-5751. [2] Lantos J, Mittra ES, Levin CS, and Iagaru A, "Standard OSEM vs. regularized PET image reconstruction: qualitative and quantitative comparison using phantom data and various clinical radiopharmaceuticals," Am J Nucl Med Mol Imaging. 2018; 8(2): 110-118.