WEKO3
アイテム
Parallel Implementation of 3-D Dynamic RAMLA with Intra-node Image Update for the jPET-D4.
https://repo.qst.go.jp/records/69144
https://repo.qst.go.jp/records/69144240fa37f-31ef-442d-9515-658111ef166e
| Item type | 会議発表用資料 / Presentation(1) | |||||
|---|---|---|---|---|---|---|
| 公開日 | 2007-11-07 | |||||
| タイトル | ||||||
| タイトル | Parallel Implementation of 3-D Dynamic RAMLA with Intra-node Image Update for the jPET-D4. | |||||
| 言語 | ||||||
| 言語 | eng | |||||
| 資源タイプ | ||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_c94f | |||||
| 資源タイプ | conference object | |||||
| アクセス権 | ||||||
| アクセス権 | metadata only access | |||||
| アクセス権URI | http://purl.org/coar/access_right/c_14cb | |||||
| 著者 |
Lam, ChihFung
× Lam, ChihFung× Yamaya, Taiga× Obi, Takashi× Yoshida, Eiji× Inadama, Naoko× Shibuya, Kengo× Murayama, Hideo× Ｌａｍ ＣｈｉｈＦｕｎｇ× 山谷 泰賀× 小尾 高史× 吉田 英治× 稲玉 直子× 澁谷 憲悟× 村山 秀雄 |
|||||
| 抄録 | ||||||
| 内容記述タイプ | Abstract | |||||
| 内容記述 | The use of small crystals in new PET scanners results in large number of line-of-response (LOR). Thus, fully 3D image reconstruction becomes time consuming especially for iterative algorithms. One of the methods to speed-up image reconstruction is by master-slave parallel computing. However, as number of computing nodes increases, the parallel efficiency decreases due to network transfer delay and other overheads. Network delay becomes significant particularly for high converging algorithm such as dynamic RAMLA, where image reconstruction is possible with large subset size. In this paper, we propose a way to reduce data transfer between nodes by allowing intra-node image update. In details, few intra-node updates for L-1 number of sub-iterations are performed in each slave thread before a global update is done at the master thread. We implemented a normal parallel version and the proposed intra-node update version of dynamic RAMLA for the jPET-D4 scanner by using MPICH2 library. We evaluated it with four servers, each contain two dual-core opteron 2.8GHz processors. Firstly, we compare the parallel efficiency (PE) gain by the proposed method with several L parameters: 1, 4, 8, and 12 for subset size of 192. PE gain is significant for L=4, after which no much gain is obtained as other factors such as load balancing become more prominent. Next, we accessed image quality of the proposed method with a hotspot phantom generated by GATE simulator, and a human brain experiment data. In both cases, no significant image artifact was found by using the proposed method. We showed that, by distributing the calculation load in a way where LORs in each slave thread covers the FOV substantially, it is possible to achieve almost the same image quality by allowing intra-node update. With proposed method, image reconstruction time is reduced by more than 22%. | |||||
| 会議概要（会議名, 開催地, 会期, 主催者等） | ||||||
| 内容記述タイプ | Other | |||||
| 内容記述 | IEEE 2007 Nuclear Science Symposium and Medical Imaging Conference | |||||
| 発表年月日 | ||||||
| 日付 | 2007-11-03 | |||||
| 日付タイプ | Issued | |||||
