Enhanced Neutron-Gamma Discrimination Using Deep Neural Networks for Precision Nuclear Medicine

Abstract

Scintillator detectors are sensitive to neutrons and gamma rays (n/γ), which is important In the field of nuclear medicine. However, It is necessary to eliminate or weaken the influence of gamma rays in neutron detection techniques. Considering that deep neural networks (DNN) can memorize train samples and classify test samples, this paper combines pulse shape discrimination techniques with DNN to achieve particle discrimination in mixed neutron and gamma rays fields. After training the DNN model and comparing it with the charge comparison algorithm, rise time algorithm, frequency domain gradient analysis algorithm, and K-means clustering algorithm. The accuracy of DNN application to n/γ pulse waveform discrimination is verified. The results show that the proposed DNN discrimination method not only provides effective discrimination of the mixed radiation fields but also improves the discrimination time compared with other discrimination methods.