学位論文

小動物用2核種SPECTシステムにおける2核種の正確な分離

Accurate separation of two radionuclidesin dual isotope SPECT scan for small animal imaging

jpn

1

57

2016-03-24

Not Applicable (or Unknown)

2016-03-24

修士(工学)

理工学研究科応用情報工学専攻; 指導教授: 尾川浩一

Multi-isotope SPECT study yields an optimization of therapeutic strategy and accurate evaluation of therapeutic effect. However, it is difficult to estimate primary photons in a multi-isotope SPECT system due to Compton scattered photons. The aim of this study is to estimate three primary counts of 99m Tc (photopeak energy 141keV) and 111 In (171, 245keV)in a simultaneous SPECT data aquisition. The target system is Nano SPECT/CT system and it has four dtectors. Each detector has nine pinholes, thus the system acquires 36 projection data simultaneously. In this paper, an accurate measurement method of 99m Tc and 111 In using an artificial neural network was proposed. In the acquisition of SPECT data, we used four energy windows: one for the 99m Tc photopeak (141keV), two for 111 In photopeaks (171, 245keV) and the last one for scattered photons. The count data of these four energy windows were input to a feed forward neural network with these layers. Using the output values we estimated two primary counts of 99m Tc and 111 In. The learning of the neural network was performed with the calculated phantom data using an Monte Carlo method. The accuracy of the proposed method was cnfirmed with the simulations. The simulation setting was assumed with the actual data acquisition of Nano SPECT/CT system.

Thesis