Radioisotope Identification Using CLYC Detectors

In recent years, machine learning methods have become very popular. Their exceptional performance has led to their adaptation in many different domains. One of the possible applications is to enhance the radioisotope identification capabilities of radiation detectors for safety and security applications. Due to its ability to detect both neutrons and gamma rays at the same time and to distinguish between them, the Cs
2
LiYCl
6
(CLYC) scintillator is a good candidate for the application of this method. We present here an analysis of the performance of different machine learning methods available, namely multilayer perceptron with backpropagation, k-nearest neighbors, and random forest. Here, feature selection is an important scientific requirement for the classifier, because the number of features, p, is much larger than the number of observations, N, or p » N. The minimum redundancy maximum relevance algorithm was used for feature selection in solving this problem. Whereas all of the methods perform very well, results show that random forest is the best one for this type of task.