Gamma-ray position reconstruction in large lanthanum-halide crystals with SiPM readout: analytical vs. neural-network algorithms

The main objective of this research was to achieve an excellent gamma-ray position reconstruction with very large monolithic scintillation crystals coupled to pixelated silicon photosensors. Analyzing the scintillation-light distribution measured along the SiPM-pixels one can get a sub-pixel accuracy for the position reconstruction. The quality of the latter depends on both the characteristics of the crystal-photosensor assembly, and the goodness of the algorithm used to reconstruct the gamma-ray hit location. We have carried out a systematic study for 511 keV gamma-rays using three different crystal thicknesses of 10 mm, 20 mm and 30 mm, all of them with planar geometry and a base size of 50x50mm2. To our knowledge, these are the largest monolithic crystals with SiPM readout aimed at gamma-ray imaging reported in the literature thus far. We have implemented state-of-the-art 3D position-reconstruction methods based on the fit of an analytical model for the propagation of the scintillation light distribution, as well as methods based on artificial neural-networks. In all cases the experimental data-set used as reference was a matrix of 35x35 collimated-source positions measured across the transversal xy-plane on a pitch of 1.5mm. In terms of spatial resolution, a superior performance is obtained with the analytical-fit methods for the 10 mm and 20 mm thick crystals, with results of 1-2 mm FWHM on average across the full field-of-view of nearly 25 cm2. On the other hand, NN-algorithms perform better for thick crystals (30mm), with average position resolutions of 3 mm FWHM. This research is intended for the development of a total-energy detector with gamma-ray imaging capability, so-called i-TED, which is aimed at the measurement of neutron-capture cross sections using the time-of-flight technique. The results reported here may also be of interest for medical imaging, homeland security and Compton astronomy.