量研学術機関リポジトリ「QST-Repository」は、国立研究開発法人 量子科学技術研究開発機構に所属する職員等が生み出した学術成果（学会誌発表論文、学会発表、研究開発報告書、特許等）を集積しインターネット上で広く公開するサービスです。 Welcome to QST-Repository where we accumulates and discloses the academic research results(Journal Publications, Conference presentation, Research and Development Report, Patent, etc.) of the members of National Institutes for Quantum and Radiological Science and Technology.
Thank you very much for using our website. On the 11th of March 2019, this site was moved from our own network server to the JAIRO Cloud network server. If you previously bookmarked this site, that bookmark will no longer work. We would be grateful if you could bookmark the website again. Thank you very much for your understanding and cooperation.
The OpenPET, which has a physical gap between two detector rings, is our new geometry. In order to realize future radiation therapy guided by OpenPET, real-time imaging is required. Therefore we developed a list-mode image reconstruction method using general purpose graphic processing units (GPUs). For GPU implementation, the efficiency of acceleration depends on the implementation method which is required to avoid conditional statements. In this paper, therefore, we developed a new system model suitable for GPU implementation. In the proposed system model, each element of system matrix was calculated as the value of detector response function (DRF) of the length between the center of a voxel and a line of response (LOR). The DRF, which was calculated analytically to represent the probability distribution of each LOR, was modeled by a sixth-order polynomial function. The system model enabled us to calculate each element of the system matrix with reduced number of the conditional statements. We used the list-mode dynamic row-action maximum likelihood algorithm (DRAMA) which could reduce the number of iterations to only one. We applied the developed method to a small OpenPET prototype, which was developed for a proof-of-concept. The results showed that high quality reconstructed images were obtained using the proposed system model with 14.8 times faster than using the conventional system model.
GPU-based image reconstruction method including geometrical detector response functions for OpenPET
