Optimization of PET Image Reconstruction for Lesion Detection

用于病灶检测的PET图像重建的优化 摘要 PET是广泛用于肿瘤学研究的分子成像模式，这是由于其高灵敏度和高分辨率。 早期诊断的潜力。对于神经内分泌肿瘤（NET），68 Ga-DOTATATE PET已被用于 自2016年以来，最近用于成人和儿科患者的临床常规NET成像。发挥着 在NET的诊断和分期中发挥重要作用。然而，与18F-FDG PET相比， 68 Ga-DOTATATE PET的质量较低，这是由于正电子范围大得多，半衰期短， 给药剂量受发电机容量限制。所有这些都损害了 68 Ga-DOTATATE PET，尤其是小病变，可能导致NET不准确 诊断.随着68 Ga-DOTATATE PET越来越多地用于临床， 进一步优化用于NET检测的68 Ga-DOTATATE PET/CT成像。最近，数据驱动 已经开发了用于PET图像去噪的方法，其中PET系统模型不 考虑了由于68 Ga-DOTATATE PET的肿瘤与背景比大于18F-FDG PET， 68 Ga-DOTATATE PET的损伤恢复也会受到平滑效应的巨大影响 作为训练和测试数据集之间的潜在不匹配。在这项研究中，我们提出了一个新的数据- 知情和病变检测驱动的图像重建框架。PET系统模型、图像 去噪模块和损伤检测模块都将被包括在该重建框架中。 本探索性建议的两个具体目标是：（1）开发病变检测驱动的PET 图像重建框架，并验证它的基础上全面的计算机模拟，（2）， 将所提出的重建框架应用于现有的临床68 Ga-DOTATATE PET/CT数据集 并根据各种品质因数对其进行测试。我们期望，具体目标的综合成果 将是一种新颖和强大的图像重建框架，以更好地恢复68 Ga中的病变， DOTATATE PET扫描，对于NET管理至关重要。

Optimization of PET Image Reconstruction for Lesion Detection Abstract PET is a molecular imaging modality widely used in oncology studies due to its high sensitivity and the potential of early diagnosis. For neuroendocrine tumors (NETs), 68Ga-DOTATATE PET has been recently used in clinical routine for imaging NETs in adult and pediatric patients since 2016. It plays an important role in the diagnosis and staging of NETs. However, compared to 18F-FDG PET, the image quality of 68Ga-DOTATATE PET is lower due to much larger positron range, shorter half-life, and lower dose administration limited by generator capacity. All of these compromises the lesion detectability of 68Ga-DOTATATE PET, especially for small lesions, and can potentially lead to inaccurate NET diagnosis. As 68Ga-DOTATATE PET is increasingly used in clinics, there is an urgent and unmet need to further optimize 68Ga-DOTATATE PET/CT imaging for NET detection. Recently, data-driven methods have been developed for PET image denoising, where the PET system model is not considered. As the tumor-to-background ratio of 68Ga-DOTATATE PET is greater than 18F-FDG PET, the lesion recovery of 68Ga-DOTATATE PET can be hugely influenced by the smoothing effects as well as potential mismatches between training and testing datasets. In this study, we propose a novel data- informed and lesion detection-driven image reconstruction framework. The PET system model, image denoising module, and lesion-detection module will all be included in this reconstruction framework. The two specific aims of this exploratory proposal are (1) to develop a lesion detection-driven PET image reconstruction framework and validate it based on comprehensive computer simulations, (2) to apply the proposed reconstruction framework to existing clinical 68Ga-DOTATATE PET/CT datasets and test it based on various figure-of-merits. We expect that the integrated outcome of the specific aims will be a novel and robust image reconstruction framework to better recover lesions in a 68Ga- DOTATATE PET scan, which is essential for NET managements.