Abstract: The online measurement system of 10 MW high temperature gas cooled test reactor can obtain the activity of some nuclides by analyzing γ spectra excluding the vast majority of the transuranic isotopes. For the purpose of nuclear security, a deep learning with method of transuranic calculation was proposed to determine the activity of transuranic isotopes. The method employed a deep neural network model with back-propagation of error that took in the activity of easy-to-measure nuclides and output the activity of those difficult to measure. In this paper, the historical working capacity of 10 MW high temperature gas cooled test reactor was tracked with the nuclide generation and depletion program of the reactor, from which data samples of nuclide activity were produced and the neural network model was trained and tested. According to the research results, the deep neural network model was of higher accuracy in nuclide content estimating and therefore promising in the online estimation of transuranic elements in fuels after irradiation of pebble bed high temperature gas-cooled reactor.