3D PET Reconstruction Using Generalized Natural Pixels

描述（由申请人提供）：在过去的十年中，临床 PET 在肿瘤成像应用中的使用大幅增加，这主要是由于 18F-FDG 的可用性增加所致。为了最大限度地提高灵敏度，PET 扫描仪设计的最新趋势是使用更快、更亮的闪烁体、更大的轴向尺寸以及在不使用隔垫的情况下采集全三维 (3D) 数据。这些趋势导致重建过程的复杂性增加。该提案的总体目标是开发 PET 3D 图像重建方法，以提高诊断准确性。临床3D PET重建最常见的方法是使用重新分箱方法结合2D重建。虽然这种方法可以以较低的计算成本实现，但它将数据建模为穿过物体的线积分，因此无法准确地解释 PET 中探测器响应的空间变化。实际上，成像过程中存在许多物理效应，使该线积分假设无效。这些影响包括：正电子范围、非共线性、空间变化的几何效率、晶体间穿透、晶体散射以及准确定位探测器块内相互作用位置的不确定性。在这个项目中，我们建议开发和研究一种 3D PET 方法，该方法使用替代基函数（而不是体素基函数）来描述感兴趣的对象。这些基函数充分利用 PET 几何结构中存在的对称性，产生具有块循环特性的系统响应矩阵。这些特性使得可以通过将整个 3D 系统响应矩阵存储在内存中来实现重建算法，并且计算时间非常快。精确的蒙特卡罗模拟代码 (GATE) 将用于计算 3D 系统响应矩阵。将使用临床图像进行心理物理学观察研究，以评估不同重建方法在 1 厘米以下肿瘤检测方面的改进。如果所提出的重建方法能够成功实现其目的，那么疑似或已知癌症患者的护理将会得到改善。公共健康相关性：该提案的总体目标是开发 PET 3D 图像重建方法，以提高诊断准确性。将使用临床图像进行心理物理学观察研究，以评估不同重建方法在 1 厘米以下肿瘤检测方面的改进。如果所提出的重建方法能够成功实现其目的，那么疑似或已知癌症患者的护理将会得到改善。

DESCRIPTION (provided by applicant): In the past ten years, there has been a substantial increase in the use of clinical PET for oncological imaging applications, which has primarily been driven by the increased availability of 18F-FDG. In order to maximize sensitivity, the recent trend in PET scanner design is for faster and brighter scintillators, larger axial dimensions, and acquisition of fully three-dimensional (3D) data, without the use of septa. These trends result in an increased complexity of the reconstruction process. The overall goal of this proposal is to develop 3D image reconstruction methods for PET that can provide improved diagnostic accuracy. The most common approach for clinical 3D PET reconstruction is to use a re-binning method combined with 2D reconstruction. While this approach can be implemented with low computational cost, it models the data as line integrals through the object, and thus cannot accurately account for the spatially variant detector response in PET. In actuality, there are a number of physical effects in the imaging process that invalidate this line integral assumption. These include such effects as: positron range, non-collinearity, spatially variant geometric efficiency, inter-crystal penetration, crystal scatter, and uncertainties in accurately locating the position of interaction within the detector block. In this project, we propose to develop and investigate an approach for 3D PET that use alternative basis functions (as opposed to voxel basis functions) to describe the object of interest. These basis functions takes full advantage of the symmetries present in the PET geometry resulting in a system response matrix with block circulant properties. These properties make it possible to implement the reconstruction algorithm with storage of the entire 3D system response matrix in memory, and with very fast computation time. An accurate Monte Carlo simulation code (GATE) will be used to compute the 3D system response matrix. Psychophysical observer studies, using clinical images, will be conducted to evaluate improvements in sub 1 cm tumor detection with different reconstruction methods. If the proposed reconstruction methods are successful in their intent, the care of patients with suspected or known cancer will be improved. PUBLIC HEALTH RELEVANCE: The overall goal of this proposal is to develop 3D image reconstruction methods for PET that can provide improved diagnostic accuracy. Psychophysical observer studies, using clinical images, will be conducted to evaluate improvements in sub 1 cm tumor detection with different reconstruction methods. If the proposed reconstruction methods are successful in their intent, the care of patients with suspected or known cancer will be improved.