Combined peak searching and machine learning algorithm for identification of radionuclides

There are two main kinds of methods of radionuclides identification: peak searching (peak fitting) and machine learning (processing spectrum at once). In this work we present an identification method that combines peak searching and machine learning. As a peak searching routine convolution with the second derivative of Gauss function is used. For the machine learning part different algorithms were tested and MLPClassifier algorithm was selected. This identification method was trained using 540 real spectra and tested with 580 real spectra. As a result, all identifications were correct.