Program Information
Algorithm-Enabled Image-Quality Improvement in List-Mode TOF-PET Reconstruction
Z Zhang1*, S Rose2 , J Ye3 , A Perkins4 , C Tung5 , C Kao6 , E Sidky7 , X Pan8 , (1) The University of Chicago, Chicago, IL, (2) ,,,(3) Philips Healthcare, Cleveland, OH, (4) Philips Healthcare, Cleveland, OH, (5) Philips Healthcare, Cleveland, OH, (6) The University of Chicago, Chicago, IL, (7) University of Chicago, Chicago, IL, (8) Univ Chicago, Chicago, IL
Presentations
WE-F-201-2 (Wednesday, August 2, 2017) 1:45 PM - 3:45 PM Room: 201
Purpose: We investigate optimization-based image reconstructions in digital time-of-flight (TOF) positron-emission tomography (PET) and demonstrate their potential for improved spatial resolution, enhanced signal-to-noise ratio, and increased axial coverage.
Methods: A clinical digital PET scanner was used to collect list-mode TOF-PET data from Jaszczak and IEC phantoms. The Jaszczak phantom contains six types of cold rods of diameters 4.8, 6.4, 7.9, 9.5, 11.1, and 12.7 mm. The IEC phantom contains 6 spheres of 10, 13, 17, 22, 28, and 37 mm diameter; the activity in the smallest four spheres is 4 times the background and zero in the other two. We formulate an optimization problem for minimizing the Kullback-Leibler (KL) data divergence with an image total variation (TV) constraint. We develop an algorithm based upon the Chambolle-Pock (CP) method for image reconstruction through solving the problem. We also reconstructed images by using an un-regularized algorithm as references. In addition to visual inspection, quantitative metrics were used for quantitatively analyzing of the image quality.
Results: Reconstructions by use of the CP algorithm exhibit lower background noise, increased axial coverage, and higher contrast compared to the un-regularized images. For the Jaszczak phantom, the overall image quality of the CP reconstruction appears to outperform the clinical counterpart. E.g., for a transverse slice near the axial edge, the discernable cold rods in the former and latter reconstructions are of spatial resolution of 7.9 mm and 12.7 mm. For the IEC phantom, the CP algorithm yields 7%~20% and 4%~6% higher contrast for hot and cold spheres, respectively, and the background-noise is 27%~46% lower.
Conclusion: An optimization-based algorithm is developed for reconstruction with improved spatial and contrast resolution and signal-to-noise ratio over the standard, un-regularized algorithm in TOF-PET imaging. The improved quality of images can be of high significance for improving lesion detectability at low count statistics.
Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by NIH grants CA158446, CA182264, and EB018102, and a Philips research grant.
Contact Email: