WO2022120589A1 - 基于em算法的动态pet参数图像分部重建算法 - Google Patents

基于em算法的动态pet参数图像分部重建算法 Download PDF

Info

Publication number
WO2022120589A1
WO2022120589A1 PCT/CN2020/134630 CN2020134630W WO2022120589A1 WO 2022120589 A1 WO2022120589 A1 WO 2022120589A1 CN 2020134630 W CN2020134630 W CN 2020134630W WO 2022120589 A1 WO2022120589 A1 WO 2022120589A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
algorithm
dynamic pet
iterative
reconstructed
Prior art date
Application number
PCT/CN2020/134630
Other languages
English (en)
French (fr)
Inventor
陈子翔
胡战利
郑海荣
梁栋
刘新
杨永峰
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Priority to PCT/CN2020/134630 priority Critical patent/WO2022120589A1/zh
Publication of WO2022120589A1 publication Critical patent/WO2022120589A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment

Definitions

  • the invention relates to the technical field of medical imaging, in particular to a dynamic PET parameter image segmentation reconstruction algorithm based on an EM algorithm.
  • the key to the application of dynamic PET imaging technology is the quantitative analysis of the patient's physiological and metabolic conditions.
  • Post-processing of image sequences based on the Patlak physiology model is the current mainstream dynamic PET data analysis method.
  • Reconstruction of parametric images by indirect method means that all dynamic PET image sequences are first reconstructed using the projection data acquired by image acquisition, and the time activity curve (TAC) of each tissue of the patient is obtained from the dynamic image.
  • TAC time activity curve
  • the blood input function based on the appropriate mode model and the Patlak physiology model, calculates the imaging agent influx rate Ki of each tissue of the patient, that is, the slope of the effective section of the Patlak curve.
  • Each pixel point ⁇ constitutes a dynamic PET parameter image.
  • frontier scientists in the dynamic PET discipline proposed a method to directly calculate the parametric image from the projection data acquired from the image acquisition, that is, the direct method of dynamic PET parametric image reconstruction. Since this method does not go through two image estimations, compared with the indirect method parametric image reconstruction technology, the error and interference information in the parametric image calculation results are greatly reduced.
  • the indirect method reconstructs the parametric image through two image estimation processes, namely the dynamic PET image reconstruction and the parametric image reconstruction, there are large errors and interference information in the parametric image reconstruction result.
  • the method of directly reconstructing the parametric image using the maximum a posteriori proposed in the related art greatly reduces the errors and interference signals in the reconstructed parametric image.
  • this method integrates the slope image ⁇ of the Patlak curve and the intercept image b into a target image and solves it at one time, the number of unknowns to be solved in the iteration is large, resulting in the limitation of the image quality of the reconstruction parameters.
  • the present invention proposes a dynamic PET parameter image reconstruction algorithm based on the EM algorithm.
  • the mutual interference between the slope image and the intercept image in the iterative calculation is reduced, so that the reconstructed image has higher image quality, which is more conducive to subsequent diagnosis and analysis.
  • the present invention provides a dynamic PET parameter image segmentation reconstruction algorithm based on the EM algorithm, including: disassembling the parametric image imaging linear equation obtained by coupling the blood flow physiology model and the imaging linear equation, and decomposing the reconstruction target image For the two target images, the slope image ⁇ and the intercept image b are reconstructed by alternating EM iterative image reconstruction algorithm.
  • n represents a certain time frame
  • t s,n is the start time of time frame n
  • t e,n is the end time of time frame n
  • ⁇ is the integration time variable
  • ⁇ is the inner function integration time variable
  • ⁇ is the decay constant of the labeled isotope.
  • the EM iterative image reconstruction algorithm includes the following slope image ⁇ iterative formula:
  • N is the number of pixels of the target parameter image
  • 1 N is a column vector whose element value is all 1 of length N
  • T represents the matrix transposition.
  • the EM iterative image reconstruction algorithm includes the following intercept image b iteration formula:
  • the EM iterative image reconstruction algorithm adopts an iterative formula alternate iteration method, and after the iterative calculation reaches stable convergence, the directly reconstructed Patlak curve reconstruction slope image ⁇ and intercept image b are obtained.
  • the algorithm includes dynamic PET parametric images of human brain, chest, liver, kidney or whole body.
  • the PSNR value of the slope image ⁇ reconstructed by the algorithm reaches 28.0636
  • the PSNR value of the intercept image b reaches 22.2039.
  • the present invention disassembles the parametric image imaging linear equation obtained by coupling the blood flow physiology model and the imaging linear equation, and decomposes the reconstructed target image ⁇ ,b ⁇ into two target images ⁇ and ⁇ b ⁇ , and the slope image ⁇ and the intercept image b are reconstructed by alternately performing EM iterative image reconstruction algorithm.
  • the present invention is based on the EM iterative image reconstruction algorithm and the partial direct dynamic PET parameter image reconstruction method of the Patlak mathematical model.
  • the slope image and intercept image are iteratively reconstructed in parts in the linear equation, which greatly reduces the mutual interference between the slope image and the intercept image in the iterative calculation, so that the reconstructed image has higher image quality and greatly reduces the parameters
  • the introduction of errors in the image reconstruction process makes the obtained parametric images more conducive to subsequent diagnosis and analysis.
  • Fig. 1 is the experimental verification result diagram of the present invention.
  • the embodiment of the present invention provides a dynamic PET parameter image segmentation reconstruction algorithm based on the EM algorithm, which includes: disassembling the parametric image imaging linear equation obtained by coupling the blood flow physiology model and the imaging linear equation, and reconstructing the target image ⁇ , b ⁇ is decomposed into two target images ⁇ and ⁇ b ⁇ , and the slope image ⁇ and the intercept image b are reconstructed by an alternate EM iterative image reconstruction algorithm.
  • n represents a certain time frame
  • t s,n is the start time of time frame n
  • t e,n is the end time of time frame n
  • ⁇ is the integration time variable
  • ⁇ is the inner function integration time variable
  • ⁇ is the decay constant of the labeled isotope.
  • the EM iterative image reconstruction algorithm includes the following slope image ⁇ iterative formula:
  • N is the number of pixels of the target parameter image
  • 1 N is a column vector whose element value is all 1 of length N
  • T represents the matrix transposition.
  • the EM iterative image reconstruction algorithm includes the following intercept image b iteration formula:
  • the EM iterative image reconstruction algorithm adopts the iterative formula alternately iterative method. After the iterative calculation achieves stable convergence, the directly reconstructed Patlak curve reconstruction slope image ⁇ and intercept image b are obtained.
  • the present invention includes human brain, chest, liver, kidney or whole body dynamic PET parameter images, the PSNR value of the reconstructed slope image ⁇ reaches 28.0636, and the PSNR value of the intercept image b reaches 22.2039.
  • the reconstructed parameter image undergoes two image estimation processes, namely dynamic PET image reconstruction and parametric image reconstruction.
  • dynamic PET image reconstruction the time of each tissue of the patient is obtained from the dynamic image.
  • the activity curve (TAC), which combines the time activity curve and the blood input function during the imaging process, calculates the imaging agent inflow rate Ki of each tissue of the patient based on the appropriate mode model and the Patlak physiological model, that is, the effective section of the Patlak curve
  • the slope of each pixel ⁇ constitutes a dynamic PET parameter image.
  • Direct method The method of directly reconstructing the parametric image using the maximum a posteriori, based on the formula: Perform iterative solution, P is the imaging system matrix, A is the parameter matrix, ⁇ is the image vector with estimated parameters, r is the scattering and random coincidence time in the projection data, and the operation between the matrices A and P For the Kronecker product operation, the matrix A is expressed as:
  • the vector ⁇ can be expressed as:
  • ⁇ and b are respectively the slope image and intercept image of the Patlak curve to be reconstructed, and the symbol T represents the matrix transposition. and are the decay integral function of the time integral of the blood input function C p (t) and the decay integral function of the C p (t) corresponding to the current dynamic PET imaging task, which can be expressed as
  • the Patlak slope image ⁇ and the outcome image b are integrated into a target image to solve at one time.
  • the PSNR value of the slope image ⁇ obtained by the indirect method is 26.6289, and the PSNR value of the intercept image b is 24.5197. There are large errors in the results and the image quality is low.
  • the PSNR value of the slope image ⁇ obtained by the direct method is 25.6658, the PSNR value of the intercept image b is 20.0284, and the image quality cannot meet the needs.
  • the PSNR value of the slope image ⁇ reconstructed by the method of the present invention reaches 28.0636, the PSNR value of the intercept image b reaches 22.2039, and the image quality is high and has superior performance.
  • the present invention re-disassembles the parametric image imaging linear equation obtained by coupling the blood flow physiology model and the imaging linear equation.
  • the result is that the right side of the equal sign of the linear equation becomes three terms: a slope image term, an intercept image term and a random event term ;
  • the dynamic PET parameter image reconstructed by the method of the present invention is significantly improved in image quality compared with the parameter image calculated by the indirect method and the parameter image reconstructed directly by the coupled linear equation; and the method of the present invention disassembles the coupled linear equation during the implementation process.
  • the equivalent imaging system matrix used in subsequent iterations and Compare the equivalent imaging system matrix in the original linear equation The size of the matrix is reduced by half, which effectively reduces the computing load of the computer when the data volume of the imaging task is large.
  • the method of the present invention proposes a method of alternately iteratively reconstructing the Patlak slope image ⁇ and the intercept image b in a linear equation, which greatly reduces the mutual interference between the ⁇ image and the b image in the iterative calculation, so that the reconstructed image has more High image quality is more conducive to subsequent diagnosis and analysis.
  • the method of the present invention is not limited to the dynamic PET imaging of the human brain, and is also applicable to the dynamic PET imaging analysis of the chest, liver or kidney; the method of the present invention can also be extended to the dynamic PET imaging analysis of the whole body.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Engineering & Computer Science (AREA)
  • Radiology & Medical Imaging (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Optics & Photonics (AREA)
  • Pathology (AREA)
  • Physics & Mathematics (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

一种基于EM算法的动态PET参数图像分部重建算法，涉及医学成像技术领域，将耦合血流生理学模型与成像线性方程得到的参数图像成像线性方程进行拆解，将重建目标图像分解为两个目标图像，并通过交替进行的EM迭代图像重建算法重建斜率图像κ与截距图像b；基于EM迭代图像重建算法和Patlak数学模型的分部直接动态PET参数图像重建方法，将Patlak曲线的斜率图像与截距图像在线性方程中分部分交替迭代重建，极大程度减小斜率图像与截距图像在迭代计算中的互相干扰，使重建图像具有更高的图像质量，极大程度减少参数图像重建过程中的误差引入，使所得参数图像更加有利于后续的诊断与分析。

Description

基于EM算法的动态PET参数图像分部重建算法 技术领域
本发明涉及医学成像技术领域，具体涉及基于EM算法的动态PET参数图像分部重建算法。
背景技术
动态PET成像技术应用的关键是对病人生理代谢情况的定量分析。基于Patlak生理学模型对图像序列进行后处理是目前主流的动态PET数据分析方法。间接法重建参数图像指，使用图像采集得到的投影数据首先重建出所有动态PET图像序列，从动态图像中得到病人各组织的时间活度曲线(TAC)，结合时间活度曲线与成像过程中的血液输入函数，基于合适的方式模型和Patlak生理学模型，计算出病人各个组织的显像剂流入率Ki，即Patlak曲线有效区段的斜率。各像素点κ即构成动态PET参数图像。在间接法之外，动态PET学科前沿科学家提出从图像采集所得投影数据直接计算参数图像的方法，即直接法动态PET参数图像重建。这种方法由于不经过两次图像估算，相比较间接法参数图像重建技术，极大减少了参数图像计算结果中的误差和干扰信息。
间接法重建参数图像由于经过两次图像估算的过程，即动态PET图像重建与参数图像重建，参数图像重建结果中存在较大的误差与干扰信息。相关技术提出的使用最大后验直接重建参数图像的方法很大程度减少了重建参数图像中的误差与干扰信号。但是，由于这种方法将Patlak曲线的斜率图像κ与截距图像b整合为一个目标图像一次性求解，迭代中所要求解的未知数数量较大，导致重建参数图像质量仍然存在限制。
发明内容
为了解决现有技术中的问题，本发明提出基于EM算法的动态PET参数图像分部重建算法，将Patlak曲线的斜率图像与截距图像在线性方程中分部分交替迭代重建的方法，极大程度减小斜率图像与截距图像在迭代计算中的互相干扰，使重建图像具有更高的图像质量，更加有利于后续的诊断与分析。
为了实现以上目的，本发明提供了基于EM算法的动态PET参数图像分部重建算法，包括：将耦合血流生理学模型与成像线性方程得到的参数图像成像线性方程进行拆解，将重建目标图像分解为两个目标图像，并通过交替进行的EM迭代图像重建算法重建斜率图像κ与截距图像b。
进一步地，所述参数图像成像线性方程拆解为下式：
Figure PCTCN2020134630-appb-000001
其中，
Figure PCTCN2020134630-appb-000002
与
Figure PCTCN2020134630-appb-000003
分别为当前动态PET成像任务对应的血液输入函数C p(t)的时间积分函数和C p(t)的衰减积分函数；P为成像系统矩阵；r为投影数据中的散射与随机符合时间；
Figure PCTCN2020134630-appb-000004
为克罗内克积运算；κ和b分别为待重建的Patlak曲线斜率图像与截距图像。
进一步地，所述时间积分函数
Figure PCTCN2020134630-appb-000005
表示为：
Figure PCTCN2020134630-appb-000006
其中，n表征某个时间帧，t s,n为时间帧n的起始时间；t e,n为时间帧n的结束时间；τ为积分时间变量；ξ为内层函数积分时间变量；λ为标记同位素的衰减常数。
进一步地，所述衰减积分函数
Figure PCTCN2020134630-appb-000007
表示为：
Figure PCTCN2020134630-appb-000008
进一步地，所述EM迭代图像重建算法包括如下斜率图像κ迭代公式：
Figure PCTCN2020134630-appb-000009
其中，N为目标参数图像像素数，1 N为长度为N的元素值全为1的列向量；T表示矩阵转置。
进一步地，所述EM迭代图像重建算法包括如下截距图像b迭代公式：
Figure PCTCN2020134630-appb-000010
进一步地，所述EM迭代图像重建算法采用迭代公式交替迭代方式，迭代计算达到稳定收敛之后，即获得直接重建的Patlak曲线重建斜率图像κ和截距图像b。
进一步地，所述算法包括人体脑部、胸部、肝部、肾脏或全身动态PET参数图像。
进一步地，所述算法重建的斜率图像κ的PSNR值达到28.0636，截距图像b的PSNR值达到22.2039。
与现有技术相比，本发明将耦合血流生理学模型与成像线性方程得到的参数图像成像线性方程进行拆解，将重建目标图像{κ,b}分解为两个目标图像{κ}和{b}，并通过交替进行的EM迭代图像重建算法重建斜率图像κ与截距图像b，本发明基于EM迭代图像重建算法和Patlak数学模型的分部直接动态PET参数图像重建方法，将Patlak曲线的斜率图像与截距图像在线性方程中分部分交替迭代重建，极大程度减小斜率图像与截距图像在迭代计算中的互相干扰，使重建图像具有更高的图像质量，极大程度减少参数图像重建过程中的误差引入，使所得参数图像更加有利于后续的诊断与分析。
附图说明
图1为本发明的实验验证结果图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。通常在此处附图中描述和示出的本发明实施例的组件可以以各种不同的配置来布置和设计。因此，以下对本发明的实施例的详细描述并非旨在限制要求保护的本发明的范围，而是仅仅表示本发明的选定实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明的实施例提供了基于EM算法的动态PET参数图像分部重建算法，包括：将耦合血流生理学模型与成像线性方程得到的参数图像成像线性方程进行拆解，将重建目标图像{κ,b}分解为两个目标图像{κ}和{b}，并通过交替进行的EM迭代图像重建算法重建斜率图像κ与截距图像b。
参数图像成像线性方程拆解为下式：
Figure PCTCN2020134630-appb-000011
其中，
Figure PCTCN2020134630-appb-000012
与
Figure PCTCN2020134630-appb-000013
分别为当前动态PET成像任务对应的血液输入函数C p(t)的时间积分函数和C p(t)的衰减积分函数；P为成像系统矩阵；r为投影数据中的散射与随机符合时间；
Figure PCTCN2020134630-appb-000014
为克罗内克积运算；κ和b分别为待重建的Patlak曲线斜率图像与截距图像。
时间积分函数
Figure PCTCN2020134630-appb-000015
表示为：
Figure PCTCN2020134630-appb-000016
其中，n表征某个时间帧，t s,n为时间帧n的起始时间；t e,n为时间帧n的结束时间；τ为积分时间变量；ξ为内层函数积分时间变量；λ为标记同位素的衰减常数。
衰减积分函数
Figure PCTCN2020134630-appb-000017
表示为：
Figure PCTCN2020134630-appb-000018
EM迭代图像重建算法包括如下斜率图像κ迭代公式：
Figure PCTCN2020134630-appb-000019
其中，N为目标参数图像像素数，1 N为长度为N的元素值全为1的列向量；T表示矩阵转置。
EM迭代图像重建算法包括如下截距图像b迭代公式：
Figure PCTCN2020134630-appb-000020
EM迭代图像重建算法采用迭代公式交替迭代方式，迭代计算达到稳定收敛之后，即获得直接重建的Patlak曲线重建斜率图像κ和截距图像b。
本发明包括人体脑部、胸部、肝部、肾脏或全身动态PET参数图像，重建的斜率图像κ的PSNR值达到28.0636，截距图像b的PSNR值达到22.2039。
为了验证本发明的优势，选择参考图像，分别利用间接法、直接法和本发明方法进行仿真实验验证，验证结果参见图1。
间接法：重建参数图像经过两次图像估算的过程，即动态PET图像重建与参数图像重建，使用图像采集得到的投影数据首先重建出所有动态PET图像序列，从动态图像中得到病人各组织的时间活度曲线(TAC)，结合时间活度曲线与成像过程中的血液输入函数，基于合适的方式模型和Patlak生理学模型，计算出病人各个组织的显像剂流入率Ki，即Patlak曲线有效区段的斜率，各像素点κ即构成动态PET参数图像。
直接法：采用最大后验直接重建参数图像的方法，基于公式：
Figure PCTCN2020134630-appb-000021
进行迭代求解，P为成像系统矩阵，A为参数矩阵，θ为带估计参数图像向量，r为投影数据中的散射与随机符合时间，矩阵A与P之间的运算
Figure PCTCN2020134630-appb-000022
为克罗内克积(Kronecker product)运算，矩阵A表示为：
Figure PCTCN2020134630-appb-000023
向量θ可分别表示为：
θ＝[κ T,b T] T
其中κ，b即分别为待重建的Patlak曲线斜率图像与截距图像，符号T表示矩阵转置。
Figure PCTCN2020134630-appb-000024
与
Figure PCTCN2020134630-appb-000025
分别为当前动态PET成像任务对应的血液输入函数C p(t)的时间积分的衰减积分函数和C p(t)的衰减积分函数，可分别表示为
Figure PCTCN2020134630-appb-000026
与
Figure PCTCN2020134630-appb-000027
基于公式
Figure PCTCN2020134630-appb-000028
使用最大似然期望值最大化(ML-EM)算法求解参数图像θ的迭代公式可写为：
Figure PCTCN2020134630-appb-000029
将Patlak斜率图像κ与结局图像b整合为一个目标图像一次性求解。
参见图1，间接法得到的斜率图像κ的PSNR值为26.6289，截距图像b的PSNR值为24.5197，结果中存在较大误差，图像质量较低。直接法得到的斜率图像κ的PSNR值为25.6658，截距图像b的PSNR值为20.0284，图像质量亦不能满足需要。本发明的方法重建的斜率图像κ的PSNR值达到28.0636，截距图像b的PSNR值达到22.2039，且图像质量较高，具有优越性能。
本发明将耦合血流生理学模型与成像线性方程得到的参数图像成像线性方程进行重新拆解的方法，结果是线性方程等号右边变为三项：斜率图像项，截距图像项和随机事件项；使用期望值最大化算法EM的迭代公式，交替更新斜率图像与截距图像。使用本发明方法重建得到的动态PET参数图像相比较间接法计算和耦合线性方程直接重建得到的参数图像在图像质量上有明显的提升；且本发明方法在实现过程中对耦合线性方程进行拆解，在后续迭代中所使用的等效成像系统矩阵
Figure PCTCN2020134630-appb-000030
与
Figure PCTCN2020134630-appb-000031
相比较原线性方程中的等效成像系统矩阵
Figure PCTCN2020134630-appb-000032
Figure PCTCN2020134630-appb-000033
矩阵尺寸减小了一半，在成像任务数据量较大的情况下，有效降低了计算机的运算负荷。
本发明方法提出将Patlak斜率图像κ与截距图像b在一个线性方程中分部分交替迭代重建的方法，极大程度减小κ图像与b图像在迭代计算中的互相干扰，使重建图像具有更高的图像质量，更加有利于后续的诊断与分析。
本发明方法在使用上不局限于针对人体脑部进行的动态PET成像，对于胸部，肝部或肾脏的动态PET成像分析同样适用；本发明方法同样可推广于全身动态PET成像分析中。
以上所述仅为本发明的优选实施例而已，并不用于限制本发明，对于本领域的技术人员来说，本发明可以有各种更改和变化。凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (9)

  1. 基于EM算法的动态PET参数图像分部重建算法，其特征在于，包括：将耦合血流生理学模型与成像线性方程得到的参数图像成像线性方程进行拆解，将重建目标图像分解为两个目标图像，并通过交替进行的EM迭代图像重建算法重建斜率图像κ与截距图像b。
  2. 根据权利要求1所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述参数图像成像线性方程拆解为下式：
    Figure PCTCN2020134630-appb-100001
    其中，
    Figure PCTCN2020134630-appb-100002
    与
    Figure PCTCN2020134630-appb-100003
    分别为当前动态PET成像任务对应的血液输入函数C p(t)的时间积分函数和C p(t)的衰减积分函数；P为成像系统矩阵；r为投影数据中的散射与随机符合时间；
    Figure PCTCN2020134630-appb-100004
    为克罗内克积运算；κ和b分别为待重建的Patlak曲线斜率图像与截距图像。
  3. 根据权利要求2所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述时间积分函数
    Figure PCTCN2020134630-appb-100005
    表示为：
    Figure PCTCN2020134630-appb-100006
    其中，n表征某个时间帧，t s,n为时间帧n的起始时间；t e,n为时间帧n的结束时间；τ为积分时间变量；ξ为内层函数积分时间变量；λ为标记同位素的衰减常数。
  4. 根据权利要求2所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述衰减积分函数
    Figure PCTCN2020134630-appb-100007
    表示为：
    Figure PCTCN2020134630-appb-100008
  5. 根据权利要求1所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述EM迭代图像重建算法包括如下斜率图像κ迭代公式：
    Figure PCTCN2020134630-appb-100009
    其中，N为目标参数图像像素数，1 N为长度为N的元素值全为1的列向量；T表示矩阵转置。
  6. 根据权利要求5所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述EM迭代图像重建算法包括如下截距图像b迭代公式：
    Figure PCTCN2020134630-appb-100010
  7. 根据权利要求6所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述EM迭代图像重建算法采用迭代公式交替迭代方式，迭代计算达到稳定收敛之后，即获得直接重建的Patlak曲线重建斜率图像κ和截距图像b。
  8. 根据权利要求1所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述算法包括人体脑部、胸部、肝部、肾脏或全身动态PET参数图像。
  9. 根据权利要求1所述的基于EM算法的动态PET参数图像分部重建算法，其特征在于，所述算法重建的斜率图像κ的PSNR值达到28.0636，截距图像b的PSNR值达到22.2039。
PCT/CN2020/134630 2020-12-08 2020-12-08 基于em算法的动态pet参数图像分部重建算法 WO2022120589A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/134630 WO2022120589A1 (zh) 2020-12-08 2020-12-08 基于em算法的动态pet参数图像分部重建算法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/134630 WO2022120589A1 (zh) 2020-12-08 2020-12-08 基于em算法的动态pet参数图像分部重建算法

Publications (1)

Publication Number Publication Date
WO2022120589A1 true WO2022120589A1 (zh) 2022-06-16

Family

ID=81974064

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/134630 WO2022120589A1 (zh) 2020-12-08 2020-12-08 基于em算法的动态pet参数图像分部重建算法

Country Status (1)

Country Link
WO (1) WO2022120589A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030190065A1 (en) * 2002-03-26 2003-10-09 Cti Pet Systems, Inc. Fast iterative image reconstruction from linograms
CN104050631A (zh) * 2013-11-25 2014-09-17 中国科学院上海应用物理研究所 一种低剂量ct图像重建方法
CN106510744A (zh) * 2016-04-27 2017-03-22 上海联影医疗科技有限公司 Pet扫描中多示踪剂动态参数的估计方法
CN110996800A (zh) * 2018-08-01 2020-04-10 联影美国公司 用于确定pet成像动力学参数的系统、方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030190065A1 (en) * 2002-03-26 2003-10-09 Cti Pet Systems, Inc. Fast iterative image reconstruction from linograms
CN104050631A (zh) * 2013-11-25 2014-09-17 中国科学院上海应用物理研究所 一种低剂量ct图像重建方法
CN106510744A (zh) * 2016-04-27 2017-03-22 上海联影医疗科技有限公司 Pet扫描中多示踪剂动态参数的估计方法
CN110996800A (zh) * 2018-08-01 2020-04-10 联影美国公司 用于确定pet成像动力学参数的系统、方法

Similar Documents

Publication Publication Date Title
Salehi et al. Auto-context convolutional neural network (auto-net) for brain extraction in magnetic resonance imaging
Rousseau et al. A supervised patch-based approach for human brain labeling
Wu et al. Self-supervised dynamic CT perfusion image denoising with deep neural networks
Fang et al. Towards robust deconvolution of low-dose perfusion CT: Sparse perfusion deconvolution using online dictionary learning
Huang et al. Dynamic MRI reconstruction with end-to-end motion-guided network
US20230301542A1 (en) Brain atlas individualization method and system based on magnetic resonance and twin graph neural network
US20220012890A1 (en) Model-Based Deep Learning for Globally Optimal Surface Segmentation
Fang et al. Robust low-dose CT perfusion deconvolution via tensor total-variation regularization
Ko et al. Rigid and non-rigid motion artifact reduction in X-ray CT using attention module
Hodneland et al. Segmentation-driven image registration-application to 4D DCE-MRI recordings of the moving kidneys
Groun et al. Higher order dynamic mode decomposition: From fluid dynamics to heart disease analysis
Feng et al. Liver DCE-MRI registration in manifold space based on robust principal component analysis
Zhou et al. Federated transfer learning for low-dose PET denoising: a pilot study with simulated heterogeneous data
He et al. Dynamic PET image denoising with deep learning-based joint filtering
He et al. Iterative reconstruction for low-dose CT using deep gradient priors of generative model
Joshi et al. R2Net: Efficient and flexible diffeomorphic image registration using Lipschitz continuous residual networks
Wang et al. Improving generalizability in limited-angle CT reconstruction with sinogram extrapolation
Fan et al. SGUNet: Style-guided UNet for adversely conditioned fundus image super-resolution
Fu et al. MDRANet: A multiscale dense residual attention network for magnetic resonance and nuclear medicine image fusion
WO2022120589A1 (zh) 基于em算法的动态pet参数图像分部重建算法
Fang et al. Tensor total-variation regularized deconvolution for efficient low-dose CT perfusion
Aghajani et al. Intensity based image registration by minimizing the complexity of weighted subtraction under illumination changes
CN114612583A (zh) 基于em算法的动态pet参数图像分部重建算法
Fang et al. TENDER: Tensor non-local deconvolution enabled radiation reduction in CT perfusion
Gigengack et al. Motion correction in thoracic positron emission tomography

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20964534

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20964534

Country of ref document: EP

Kind code of ref document: A1