Machine learning is used to conduct positron emission particle tracking

Abstract

Positron emission particle tracking using machine learning offers crucial biomedical applications, such as the study of blood flow and the study of gastrointestinal circulation.

Recent advances in positron emission particle tracking: a comparative review

“…In many ways the most straightforward irradiation technique, in this method activity is induced directly in the desired tracer particle by bombarding it with a high-energy beam of accelerated ions (typically protons or helium-3 nuclei 18 ) in order to generate a suitable positronemitting radionuclide from a stable element within the tracer material. One of the most commonly-employed routes is to irradiate a material containing natural oxygen ( 16 O) with a 33 MeV 3 He beam to produce 18 F from 16 O atoms via the nuclear reactions: 16 O( 3 He, p) 18 F ( 3 ) and 16 O( 3 He, n) 18 Ne → 18 F. (4) The popularity of this particular reaction stems from two main factors-firstly, oxygen atoms can be found in many materials that are widely used in industry, such as glass, sand, alumina and zirconia; secondly, by directly irradiating water (H 2 O) one can then go on to label a wide array of other materials using the indirect activation techniques described below. Other commonly used isotopes include cobalt-55, produced via the bombardment of iron atoms with protons, copper-61, produced from nickel using a beam of deuterons, and gallium-66, produced by bombarding nickel with protons, with a variety of other processes also possible [60][61][62].…”

“…As well as producing 18 F, the process described in equation ( 4) also produces a number of short-lived 'side radionuclides' such as 10 C (half-life 19.3 s), 12 N (half-life 11 ms), 27 Si (half-life 1.16 s), 29 P (half-life 4.1 s), and 26 Al (half-life 6.4 s) [60]. To ensure a high RCP-i.e.…”

“…In the former, instead of a stable atom within the material of interest being converted into a positron emitter, an atom is instead removed from the material and replaced with a suitable positron emitter. This is typically performed by placing a suitable material such as an ion exchange resin [63] in a liquid containing a high concentration of free 18 F ions. Under suitable conditions [56], the free fluorine ions will preferentially attach to the resin, thereby completing the labelling process.…”

“…By exploiting different chemistries, through the careful choice of material and the radionuclide, it is possible to attach positronemitting nuclei to the surface of a particle by strong interactions. In the field of nanomaterials, this method has been applied for the radiolabelling of different inorganic particles such as those made of hydroxyapatite, silica, or iron oxide with radionuclides such as 18 F, 68 Ga, 64 Cu or 89 Zr, among others [64].…”

“…PEPT was originally developed at the University of Birmingham in the late 1980s [15,16] and, despite extensive use in diverse applications, for the first three decades of its existence operated using largely the same fundamental algorithms originally implemented. The last decade, however, has seen an upsurge in the development of novel approaches to PEPT, including algorithms utilising cutting-edge machinelearning approaches [17,18] and highly-efficient trajectory reconstruction algorithms with potential applications in the study of bloodflow, cancer metastasis and other biomedical applications [11]. Recent years have also seen the advent of the numerical modelling and simulation of PEPT systems [17,[19][20][21], a new capability with the potential to revolutionise both the development of PEPT systems and algorithms, as well as their application to scientific and industrial processes.…”