基于点扩散函数和CT解剖信息的高分辨率、低噪声的PET重建技术

PET/CT can combine functional and anatomical imaging modalities, which has attracted much attention. Two disadvantages of PET image, however, are low spatial resolution and high noise level. It is a hot topic to improve image resolution by point spread function,which, however, will lead to apparent noise amplification. It is difficult to give consideration to both resolution and noise. This project aims to resolution improvement and noise suppression simultaneously by point spread function, CT anatomical information and estimated noise-levels. We first obtain more accurate point spread function, which is basic to high-quality reconstruction. For some specific organs, respiration and heartbeat will make PET/CT images mismatched, and thus we must analyse magnitude of organ motion and response of the surrounding tissues, so that motion artifacts can be corrected. After that, a CT segmentation is implemented to obtain anatomical information, and the corresponding noise levels are estimated. Finally, we reconstruct a high-quality image to meet clinical diagnostic needs by combining point spread function, anatomical information and estimated noise-levels. Both simulation software and real PET/CT system will be used to evaluate the algorithm performance.

PET/CT设备将功能成像和结构成像两种技术融合为一体，受到医学界广泛关注，但是PET图像分辨率低、噪声水平高是制约临床应用的主要问题。利用点扩散函数提高PET图像分辨率是国际研究的热点，但是该方法大大提高噪声水平，两者此消彼长使临床应用陷入困境。本项目利用点扩散函数和CT图像解剖信息，考虑不同器官的噪声水平差异，同时提高分辨率和抑制噪声。首先研究精确的点扩散函数测量方法，为高质量重建提供基本保证。针对特定扫描部位，呼吸和心脏运动将造成PET和CT数据不匹配，本项目分析运动影响范围和周围组织响应情况，提供有针对性的运动校正。运动校正之后，通过分割算法在CT图像中提取PET显影器官，并通过扫描条件预估计它们的噪声水平。最后，我们使用点扩散函数、解剖信息及噪声水平作为先验知识构建重建模型，并设计有效的迭代算法，获得满足临床需求的图像。算法性能将通过仿真软件和真实PET/CT系统进行评估。

本项目立足于医用PET/CT重建中关键技术的突破，从国内外研究现状看，已有方法无法彻底解决PET图像分辨率和噪声此消彼长的问题。本项目研究同时提高分辨率和降低噪声的方法，即通过PSF、解剖信息和感兴趣器官噪声水平估计获得高质量PET图像，并突破PET和CT扫描数据的运动影响和运动校正，推动PET/CT成像技术发展。已有成果已经应用于东软PET-CT系统。本项目完成了关键算法及实验，建立高分辨率、低噪声重建的原型系统，作为产品化的雏形。研究成果主要以学术论文形式发表，已在国内外核心学术期刊上发表SCI、EI 论文10余 篇（其中第一作者SCI论文7篇），申请专利8项，参加国内外学术会议1 次，并培育中央高校基本科研业务费项目1项。多名青年教师、在读博士，在读硕士及本科生参与该项目。共培养硕士研究生5名，本科生7名。