Positron Emission Particle Tracking Using Machine Learning

We introduce a new approach to positron emission particle tracking based on machine learning algorithms, demonstrating novel methods for particle location, tracking, and trajectory separation. The method allows radioactively labeled particles to be located, in three-dimensional space, with high temporal and spatial resolution, requiring no prior knowledge of the number of tracers within the system and can successfully distinguish multiple particles separated by distances as small as 2 mm. The technique's spatial resolution is observed to be invariant with the number of tracers used, allowing large numbers of particles to be tracked simultaneously, with no loss of data quality.