Machine learning techniques for the development of an , and radiation sources discrimination software using a 2d-imaging radiation detector

Laffolley, H.; Tsubota, Yoichi   ; Tsuji, Tomoya   ; Honda, Fumiya; Nitta, Ayako; Kikuchi, Riku

In support of the decommissioning efforts at the Fukushima Daiichi Nuclear Power Station, the Japan Atomic Energy Agency (JAEA) is developing the TRACE software for the accurate discrimination of alpha, beta, and gamma radiation sources in radioactive samples. This new technology utilizes the MiniPIX, a commercial 2D imaging semiconductor radiation sensor. The MiniPIX provides both a spatial image of radiation interactions within its pixel matrix and quantification of the energy transferred. The TRACE software analyzes the data from the MiniPIX by examining the energy and shape of pixel clusters created by radiation events. It identifies alpha particles primarily through the higher energy transferred in their interactions. For other events, it extracts shape parameters: gamma particles, having a quasi-straight path, create very small pixel clusters, while beta particles, being more deviated, generate snake-like clusters. The software can also provide an energy spectrum split by radiation type to assist in identifying radioisotopes. A core component of TRACE is a machine learning algorithm, trained using data acquired from standard sources such as Cobalt-60, Strontium-90, Cesium-137, and Americium-241, chosen for their relevance to Fukushima samples. Planned future integration with an XYZ motorized stage aims to automate data acquisition, improve efficiency, facilitate handling of radioactive and non-flat samples, and enable real-time adjustment of acquisition duration.