Artificial neural network for the determination of neutron spectra in the bunker of a Linac of 18 MV

The neutron spectrum and equivalent of environmental dose H(10) were calculated for a radiotherapy room in 16 punctual detectors, 15 inside of and 1 outside of the same one. The calculations were carried out with the Monte Carlo method and with the code MCNP5 for a generic room model with a Linac of 18 MV, obtaining this way 16 spectra with 47 intervals of energy class, starting from these spectra the values of H(10) were calculated. On the other hand, an artificial neural network was designed and trained to determine the spectra by neutrons in 15 different locations inside the radiotherapy room starting from the value of H(10) in the detector 16 located in the exterior of the room, using as training data the spectra and calculated dose by neutrons, of which a medium quadratic error was obtained (m se) in the adjustment between the objective data and the exit data of m se=1E(-8). The results demonstrate that the use of the artificial intelligence as technique is an useful tool in the spectrometry and dosimetry of neutrons, since it simplifies the characterization process of neutron fields in radiotherapy rooms without the use of spectrometry systems, and that once the energy distribution of the neutrons produced by the Linac is known and the corresponding doses be calculated H(10), they can take the appropriate cautions for the security patient in treatment as well as for the personnel in the room. (Author)