基于低秩与稀疏矩阵分解的心肌灌注动态PET图像重建方法研究

Myocardial perfusion PET imaging, as a non-invasive examination method for myocardial ischemia detection, can provide improved diagnostic accuracy, risk stratification and treatment decision and has distinct advantage compared to SPECT. It requires dynamic cardiac PET scans. However, with the increased temporal sample in dynamic PET imaging, the detected counts of single frame decrease and the quality of the associated image is commonly deteriorated, which will affect the accuracy and reliability of the quantitative analysis...The object of this proposal is to develop algorithms for low-rank plus sparse matrix decomposition based dynamic myocardial perfusion PET image reconstruction, mainly including: (1)The dynamic PET images were decomposed into background and dynamic components by low-rank plus sparse decomposition, and the specific low-rank and sparse constraints were provided to the PET image reconstruction model;(2)Considering the kinetic characteristic of myocardial perfusion imaging, the sparse representation of dynamic component was studied in different transform domain to optimize the sparse constraint; (3) The analysis and evaluation metrics of myocardial perfusion imaging were studied to optimize the reconstruction model and parameters to improve the quality of reconstructed image and provide accuracy and reliability evidence for the treatment of coronary heart disease.

PET心肌灌注显像(Myocardial Perfusion Imaging，MPI)作为一种检测心肌缺血的非侵入性检查手段，可以为冠心病的早期诊断、危险度分层与后期治疗决策提供可靠的依据，较SPECT具有显著优势。其显像过程需要对注入示踪剂的心脏进行动态数据采集，然而，随时间采样增加，单帧光子计数减少，重建图像质量严重退化，直接影响后续定量分析的准确性与可靠性。鉴于此，本项目拟开展基于低秩与稀疏矩阵分解的心肌灌注动态PET图像重建方法研究，主要包括: (1)基于低秩与稀疏矩阵分解，将动态PET图像分解为背景与动态成分，为动态PET图像重建模型构建特有的低秩与稀疏约束；(2) 结合心肌灌注显像的动力学特性，分析动态成分在不同变换域中的稀疏表达，以优化稀疏约束构建；(3)开展PET心肌灌注显像分析与评价研究，优化重建模型与参数，以期提高PET图像质量，为冠心病患者的治疗提供准确可靠的依据。

PET心肌灌注显像(Myocardial perfusion imaging，MPI)作为一种检测心肌缺血的无创性诊查手段，可以为冠心病的早期诊断、危险度分层与后期治疗决策提供可靠的依据。本项目基于低秩与稀疏矩阵分解的动态心肌灌注PET图像重建方法展开研究。研究内容主要包括以下几点：首先，实现了小动物PET的蒙特卡洛仿真平台构建及其性能评价，该系统可提供高度仿真PET数据，为后续图像重建与分析提供有效的验证条件；在此基础上，我们开发了基于低秩与稀疏分解的心肌灌注动态PET图像重建方法，并研究了在不同变换域中稀疏约束对重建图像质量的影响。通过Rb82仿真心肌灌注PET数据和临床病人数据的研究，验证了提出方法可以有效地增强重建PET图像了视觉效果和定量分析的准确性。开发了心肌灌注PET定量分析系统，并提出一种基于阿尔法散度的动态PET图像因子处理方法，为动态PET成像中血输入函数自动估计提供新的方法。上述方法与技术可以有效提升PET-MPI影像质量与定量分析的可靠性，有望为冠心病患者的治疗提供准确可靠的依据。该项目合计发表SCI论文11篇，包括医学影像领域的主流期刊《European Radiology》、《Molecular Imaging and Biology》、《Computer Methods and Programs in Biomedicine》，《Medical Physics》，《Physics in Medicine and Biology》，《IEEE Transactions on Biomedical Engineering》；获中国发明专利授权2项，软件版权1项。项目负责人2018年入选广东省珠江学者青年学者，2016年获广东省科学技术二等奖1项（第七完成人），并入选广州市珠江科技新星，共同培养在读硕博连读生2名，硕士毕业生3名，其中2名研究生获国家奖学金。