Speaker
Description
Human body receives an external and internal exposure from radioactive materials. The effective dose can be derived as the sum of the personal dose equivalent (Hp (10)) from external exposure and the committed effective dose (E (50)) from internal exposure [1]. The International Commission on Radiological Protection (ICRP) recommends that the dose limit should be expressed as an effective dose of 20 mSv per year, averaged over defined 5 years periods, with the further provision that the effective dose should not exceed 50 mSv in any single year [1]. Whole body counters (WBC) are used to evaluate the radioactive contamination of workers at domestic nuclear power plants and it is often possible to estimate external contamination as internal contamination [2]. A method to identify the location of the radioactive contamination using WBC has been developed to solve this problem [2]. However, this method has a critical problem that can not discrimination between external and internal contamination if the inside and the outside of the human body are contaminated simultaneously. The purpose of this study was to develop a method for assessing the location of the radioactive contamination of human body using an Artificial Neural Network (ANN). A humanoid ATOM phantom recommended by ICRP was used for this study. Various spectra were obtained for three cases to get a training set of the ANN: (1) 137Cs and 60Co sources were attached to the surface of the phantom, (2) The sources were inserted between slices of the phantom, (3) The sources were attached and inserted of the phantom at the same time. The hyper parameter including the number of layers, the number of neurons in each layer, the learning rate of the optimizer, and the neuron dropout rate were optimized to improve accuracy of the ANN. The accuracy of the ANN was evaluated using a test set.
