This work presents the experimental results for the position estimation of the interaction point of gamma photons in a thick 3”3” cylindrical scintillation crystal, when implementing machine learning algorithms. For the experimental characterization, a LaBr3 crystal is coupled to an array of 144 SiPM detectors. The machine learning algorithms implemented are: k-NN, Decision Tree, and Support Vector Machine. Principal Components Analysis is also implemented; this algorithm enhances the classification accuracy. Experimental results show a planar resolution close to 1 cm, suggesting the adoption of these algorithms and electronics architecture in nuclear physics experiments to correct the Doppler broadening effect. The algorithms can be executed both in post-processing and in real-time. Edge computing of coordinates and energy, on microcontroller or FPGA devices with sub-μs processing time per scintillation event, is in this context successfully applied, leading to a relaxation of data transmission rate constraints, and to the achievement of real-time position sensitivity of gamma photons.
Implementation of Real-Time Machine Learning Algorithms for 3D Scintillation Position Estimation in Thick Crystals
L. Buonanno;D. Di Vita;F. Canclini;M. Carminati;C. E. Fiorini
2020-01-01
Abstract
This work presents the experimental results for the position estimation of the interaction point of gamma photons in a thick 3”3” cylindrical scintillation crystal, when implementing machine learning algorithms. For the experimental characterization, a LaBr3 crystal is coupled to an array of 144 SiPM detectors. The machine learning algorithms implemented are: k-NN, Decision Tree, and Support Vector Machine. Principal Components Analysis is also implemented; this algorithm enhances the classification accuracy. Experimental results show a planar resolution close to 1 cm, suggesting the adoption of these algorithms and electronics architecture in nuclear physics experiments to correct the Doppler broadening effect. The algorithms can be executed both in post-processing and in real-time. Edge computing of coordinates and energy, on microcontroller or FPGA devices with sub-μs processing time per scintillation event, is in this context successfully applied, leading to a relaxation of data transmission rate constraints, and to the achievement of real-time position sensitivity of gamma photons.File | Dimensione | Formato | |
---|---|---|---|
Implementation of Real-Time Machine Learning Algorithms for 3D Scintillation Position Estimation in Thick Crystals - NSS 2020.pdf
Accesso riservato
:
Publisher’s version
Dimensione
6.35 MB
Formato
Adobe PDF
|
6.35 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.