Deep learning estimations of the production cross sections of ⁷⁷Br medical radionuclide

Bromine-77 has a half-life of 56 h and decays nearly exclusively (99.3 %) by electron capture, with prominent gamma rays at 239.0 and 520.7 keV. Once considered primarily for SPECT imaging, this nuclide is increasingly being evaluated for its potential in Auger electron therapy. In this study, deep learning algorithms with Python programming language are improved to predict the production cross sections of bromine-77 radionuclide. Experimental cross sections data used in artificial neural network were taken from the EXFOR nuclear reactions database. The deep learning results obtained for the ⁷⁷Se(p,n)⁷⁷Br, ⁷⁸Se(p,2n)⁷⁷Br, ⁸⁰Se(p,4n)⁷⁷Br and ⁷⁵As(α,2n)⁷⁷Br reactions were compared with the calculation results obtained from the TALYS code. It was observed that the results obtained with deep learning obey the experimental values much better.