Resolution Enhancement in PET Reconstruction Using Collimation.

Resolution Enhancement in PET Reconstruction Using Collimation.
复制标题

使用准直增强 PET 重建的分辨率。

DOI:
10.1109/tns.2012.2214444
复制
发表时间:
2013
影响因子:
1.8
通讯作者:
Karp,JoelS
Karp,JoelS
中科院分区:
工程技术3区
文献类型:
--
作者:
Metzler,ScottD;Matej,Samuel;Karp,JoelS

文献摘要

相似文献

与没有准直的扫描仪相比,准直可以提高空间分辨率和采样性能。空间分辨率得到提高,因为每个原始晶体可以在概念上分成两个(即,平面内晶体的数量加倍),通过用高密度衰减器(例如,钨)掩盖一半的晶体;由于包含响应线(LOR)的两种晶体都被掩盖了,这使得符合率降低了4倍,但产生了4倍的分辨率增强(RE) LORs。用不同配置的准直器扫描测量了所有新型的准直器。在本仿真研究中,假设准直器是理想的,既不允许伽马穿透也不截断视场。在2D中,我们比较了一个未准直的小动物系统,其2毫米晶体被认为是完全吸收的,而同一系统的准直将有效晶体尺寸缩小到1毫米。数字幻影包括一个热点和一个单一热点,两者都在一个统一的背景下,活动比为4:1。除了准直和非准直配置外,还模拟了2毫米情况下的角度和空间抖动采集。同样,考虑了不同组合的LORs配置,包括(i)所有LORs, (ii)仅与2mm LORs平行的LORs;(iii)仅与2mm lor不平行的交叉对。最后,以对比恢复系数和均方误差(MSE)为指标,对准直和未准直数据进行了定量研究。重建表明,对于大多数噪声水平,通过使用准直,即使在计数减少4次或在某些情况下,与无噪声的非准直重建相比,图像质量(即视觉质量、分辨率和伪影减少)也有实质性的改善。通过对不同采样配置的比较,结果表明,各采样点空间分辨率的提高和采样点数量的增加是提高重建效果的关键。此外,定量研究表明,对于低计数扫描,准直数据对小病变提供更好的MSE,而未准直数据对大病变提供更好的MSE;对于高计数的研究,校正后的数据对评估的整个病变大小范围产生更好的定量值。
Collimation can improve both the spatial resolution and sampling properties compared to the same scanner without collimation. Spatial resolution improves because each original crystal can be conceptually split into two (i.e., doubling the number of in-plane crystals) by masking half the crystal with a high-density attenuator (e.g., tungsten); this reduces coincidence efficiency by 4× since both crystals comprising the line of response (LOR) are masked, but yields 4× as many resolution-enhanced (RE) LORs. All the new RE LORs can be measured by scanning with the collimator in different configurations.In this simulation study, the collimator was assumed to be ideal, neither allowing gamma penetration nor truncating the field of view. Comparisons were made in 2D between an uncollimated small-animal system with 2-mm crystals that were assumed to be perfectly absorbing and the same system with collimation that narrowed the effective crystal size to 1 mm. Digital phantoms included a hot-rod and a single-hot-spot, both in a uniform background with activity ratio of 4:1. In addition to the collimated and uncollimated configurations, angular and spatial wobbling acquisitions of the 2-mm case were also simulated. Similarly, configurations with different combinations of the RE LORs were considered including (i) all LORs, (ii) only those parallel to the 2-mm LORs; and (iii) only cross pairs that are not parallel to the 2-mm LORs. Lastly, quantitative studies were conducted for collimated and uncollimated data using contrast recovery coefficient and mean-squared error (MSE) as metrics. The reconstructions show that for most noise levels there is a substantial improvement in image quality (i.e., visual quality, resolution, and a reduction in artifacts) by using collimation even when there are 4 fewer counts or-in some cases-comparing with the noiseless uncollimated reconstruction. By comparing various configurations of sampling, the results show that it is the matched combination of both improved spatial resolution of each LOR and the increase in the number of LORs that yields improved reconstructions. Further, the quantitative studies show that for low-count scans, the collimated data give better MSE for small lesions and the uncollimated data give better MSE for larger lesions; for highcount studies, the collimated data yield better quantitative values for the entire range of lesion sizes that were evaluated.