31 March 2016 Event-by-event PET image reconstruction using list-mode origin ensembles algorithm
Author Affiliations +
Abstract
There is a great demand for real time or event-by-event (EBE) image reconstruction in emission tomography. Ideally, as soon as event has been detected by the acquisition electronics, it needs to be used in the image reconstruction software. This would greatly speed up the image reconstruction since most of the data will be processed and reconstructed while the patient is still undergoing the scan. Unfortunately, the current industry standard is that the reconstruction of the image would not start until all the data for the current image frame would be acquired. Implementing an EBE reconstruction for MLEM family of algorithms is possible, but not straightforward as multiple (computationally expensive) updates to the image estimate are required. In this work an alternative Origin Ensembles (OE) image reconstruction algorithm for PET imaging is converted to EBE mode and is investigated whether it is viable alternative for real-time image reconstruction. In OE algorithm all acquired events are seen as points that are located somewhere along the corresponding line-of-responses (LORs), together forming a point cloud. Iteratively, with a multitude of quasi-random shifts following the likelihood function the point cloud converges to a reflection of an actual radiotracer distribution with the degree of accuracy that is similar to MLEM. New data can be naturally added into the point cloud. Preliminary results with simulated data show little difference between regular reconstruction and EBE mode, proving the feasibility of the proposed approach.
© (2016) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Andriy Andreyev, Andriy Andreyev, "Event-by-event PET image reconstruction using list-mode origin ensembles algorithm", Proc. SPIE 9783, Medical Imaging 2016: Physics of Medical Imaging, 97835B (31 March 2016); doi: 10.1117/12.2217307; https://doi.org/10.1117/12.2217307
PROCEEDINGS
5 PAGES


SHARE
Back to Top