Multiple particle tracking in PEPT using Voronoi tessellations. (2nd November 2019)

• Record Type: Journal Article
• Authors: Blakemore, D.M.; Govender, I.; McBride, A.T.; Mainza, A.N.
• Is Part Of: Chemical engineering science. Volume 207(2019)
• Page Start: 780
• Page End: 789
• Publication Date: 2019-11-02
• Subjects: PEPT -- Nuclear imaging -- Voronoi -- Clustering; Chemical engineering -- Periodicals; Génie chimique -- Périodiques; Chemical engineering; Periodicals; Electronic journals; 660
• DOI: 10.1016/j.ces.2019.06.057
• Languages: English
• ISSNs: 0009-2509

Abstract

Highlights: In-situ tracking of multiple particles in complex flow systems using PEPT. Tracer particle labelling can have any levels of radioactivity. Only tracking constraint is saturation limit of the PET scanner. Algorithm is fully automated—no user intervention required. Abstract: Many complex flow phenomena encountered in chemical engineering lack fundamental understanding. Central to this problem is the lack of non-invasive, in situ measurement tools that can simultaneously track the motion and dynamics of different particle species across the entire phase space of the flow. To this end, a novel algorithm was developed for simultaneously tracking multiple particles using a Positron Emission Tomography scanner. The algorithm discretizes the back-to-back gamma rays emanating from multiple radio-labelled particles, and uses a Voronoi tessellation to create a density map of the points. The locations of the tracers are determined using a clustering technique. A series of experiments was performed to test the precision, robustness, and performance of the algorithm. Twenty tracers were successfully tracked, limited only by the amount of available sodium-22 isotope. The presented algorithm resulted in a RMSE of 1.26 mm. The precision was found to be dependent on the level of discretization, and is robust in that the loss of a tracer (due to exiting the field of view, high acceleration, or tracer collision) is handled internally, with no need for human correction. The proposed algorithm can also be easily parallelized.