Comparison of Rotation-based and Non-rotation-based Projectors for Image reconstruction in Mini-EXPLORER II

Objectives: Mini-Explorer II is a high-resolution, high sensitivity PET/CT scanner which is designed for human brain imaging and companion animal imaging. It is the prototype of the EXPLORER total-body PET but with much smaller ring diameter (52cm) and axial field of view (48.3cm). The projector, as well as the point spread function (PSF) modelling, is the most important component for high resolution image reconstruction in Mini-EXPLORER II. Rotation-based projector is efficient and suitable for parallel computing. Non-rotation-based projector is considered to be more accurate in point spread (PSF) modelling, but its computational efficiency is low. The purpose of this work was to investigate the performance of different projectors on the image resolution.

Methods: We implemented both rotation-based and non-rotation-based projectors for OP-OSEM. The workflow of rotation-based projector was shown in Fig.1. The PSF was obtained from real measurement. The measured data were fitted using a one-dimensional Gaussian function along the axial axis and a two-dimensional Gaussian function in the transaxial plane, corresponding to the axial and transaxial components of the PSF. Moreover, the axial PSF was symmetric and the width depended on the oblique angle of the LOR. For non-rotation-based projector, the system matrix was calculated by an analytic method proposed by Lecomte and Schmitt. It was later convolved with other physical factors such as non-collinearity, inter-crystal scatter and positron range, which was modelled as a symmetric Gaussian function. The mini-Derenzo phantom was used to compare the performance of these two projectors. The rod diameter of the mini-Derenzo phantom ranges from 1.6mm to 3.6mm. The phantom was filled with ~11.1 MBq of ¹⁸F-FDG, and placed at the center of the PET field of view transaxially and axially. Then it was moved to different positions in radial direction (30mm, 60mm, 90mm and 120mm off-center). About 500M prompts were collected at each position. The listmode data were sorted and rebinned into a raw sinogram of 280 angles, 559 radial bins, and 168x168 slices. The sinogram was further rebinned into 335 slices through single slice rebinning algorithm, and then summed into one slice. Image reconstruction was performed using the rotation-/non-rotation-based projectors with varied iterations and counts. The pixel size was 0.6x0.6 mm². Results: The reconstructed images are shown in Fig.2. No post-reconstruction filtering was applied. As we can see, for rotation-based projector the 2.4 mm rods were clearly resolved and the 2.0mm rods were partially resolved, while for non-rotation-based projector the 2.0 mm rods were clearly resolved and the minimum rods were partially resolved.

Conclusions: Non-Rotation-based projector shows better resolution performance than rotation-based projector if adequate counts are obtained. It could be used for ultra-high resolution PET imaging in Mini-EXPLORER II. Rotation based projector provides comparable results if the counts are not adequate. It is more suitable for clinical PET imaging because routine clinical PET imaging has limited acquisition time and counts.

• Download figure
• Open in new tab
• Download powerpoint