WO2020118844A1 - Pet图像的重建方法、计算机存储介质、计算机设备 - Google Patents
Pet图像的重建方法、计算机存储介质、计算机设备 Download PDFInfo
- Publication number
- WO2020118844A1 WO2020118844A1 PCT/CN2019/071746 CN2019071746W WO2020118844A1 WO 2020118844 A1 WO2020118844 A1 WO 2020118844A1 CN 2019071746 W CN2019071746 W CN 2019071746W WO 2020118844 A1 WO2020118844 A1 WO 2020118844A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- pet
- reconstructed image
- reconstructed
- block
- Prior art date
- Legal status (The legal status 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 status listed.)
- Ceased
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
- A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
- A61B6/03—Computed tomography [CT]
- A61B6/037—Emission tomography
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T12/00—Tomographic reconstruction from projections
- G06T12/20—Inverse problem, i.e. transformations from projection space into object space
-
- 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
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5205—Devices using data or image processing specially adapted for radiation diagnosis involving processing of raw data to produce diagnostic data
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10104—Positron emission tomography [PET]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2211/00—Image generation
- G06T2211/40—Computed tomography
- G06T2211/424—Iterative
Definitions
- the invention belongs to the field of information technology, and particularly relates to a PET image reconstruction method, a computer storage medium, and a computer device.
- PET Positron emission tomography
- positron tomography is imaging by first injecting a radiotracer into the patient and then measuring the distribution of radioisotopes in the patient.
- PET reconstruction algorithms are mainly divided into analytical reconstruction algorithms and iterative reconstruction algorithms.
- Analytic reconstruction algorithms mainly include back projection, filtered back projection and Fourier reconstruction.
- the most widely used algorithm is Filtered Back Projection (FBP).
- FBP Filtered Back Projection
- the FBP method is based on the Radon transform, but FBP does not consider the spatial and temporal inhomogeneity of the system response, nor does it consider the noise of the instrument during measurement, so the reconstructed image contains a lot of noise.
- Iterative reconstruction algorithms include algebraic reconstruction and statistical reconstruction.
- the existing PET reconstruction methods have the following problems: 1) Because only a single pixel difference is used to distinguish the true edge and noise fluctuations in the image reconstruction with noise, the reconstructed image does not retain the correct edge; 2) Because There is no good trade-off between removing noise and preserving detailed information, resulting in a lot of detailed information being lost in the reconstructed image; 3) For under-sampled images and noisy images, due to lack of good constraints, it will cause loss of details and blockiness Artifacts.
- the technical problem to be solved by the present invention is: how to obtain a PET image with better reconstruction effect.
- a PET image reconstruction method includes the following steps:
- Step 1 Obtain the projection data Y of the PET image and the system matrix P;
- Step 3 Acquire the initial reconstructed image X, and iteratively update the initial reconstructed image X according to the first objective function to obtain the first reconstructed image.
- the first objective function is:
- Q L (X; X n ) is a likelihood agent function constructed based on Poisson random distribution variables, Is a penalty agent function constructed based on the prior of the neighborhood block, X n is the reconstructed image obtained after the nth iteration, and ⁇ is the regularization parameter;
- Step 4 Iteratively update the first reconstructed image according to the second objective function to obtain the second reconstructed image, where the second objective function is a function constructed based on dictionary learning;
- Step 5 Determine whether the iteration conditions are satisfied. If yes, output the second reconstruction image obtained by this iteration as the final PET reconstruction image; if not, return to step 3, and use the second reconstruction image of this iteration as the next round The initial reconstructed image at iteration.
- the expression of the likelihood agent function is:
- n j represents the total number of pixels
- p ij represents the probability that the jth pixel is detected by the i-th detector
- n i represents the total number of detectors
- p j represents the total probability of the j-th pixel detected by the n i detectors Value
- X j represents the value of the jth pixel of the reconstructed image X
- y i represents the projection data detected by the ith detector
- w jk represents the weight between the jth pixel and the kth pixel of the reconstructed image X n
- N j represents the neighborhood block centered on the jth pixel
- j l is the lth pixel in the neighborhood block f j (X)
- k l Is the lth pixel in the neighborhood block f k (X)
- h l is a positive weight vector.
- the method for iteratively updating and initializing the reconstructed image X according to the first objective function to obtain the first reconstructed image includes:
- a first reconstructed image is generated based on the desired maximized image and the intermediate image.
- the expression of the second objective function is:
- X represents the first reconstructed image obtained in step three reconstruction
- R ij is the operation to obtain the image block from X
- D is a dictionary based on the image block
- ⁇ ij is a sparse representation of X ij about the dictionary D
- T 0 indicates the sparse level to be achieved.
- the method for iteratively updating the first reconstructed image according to the second objective function to obtain the second reconstructed image includes:
- the method of generating the sparse coefficient of each image block according to each image block, a pre-trained low-resolution dictionary and a high-resolution dictionary :
- a preset coefficient calculation formula calculate a sparse coefficient of the image block that satisfies the first coefficient constraint condition and the second coefficient constraint condition.
- the reconstruction method before performing image segmentation on the first reconstructed image, the reconstruction method further includes:
- the preset low-resolution PET training image set the preset high-resolution PET training image set, the size of the image block in the low-resolution PET training image set, and the size of the image block in the high-resolution PET image set
- the low-resolution dictionary and the high-resolution dictionary For joint training of the low-resolution dictionary and the high-resolution dictionary.
- the invention also discloses a computer storage medium in which a PET image reconstruction program is stored.
- a PET image reconstruction program is executed by a processor, the above-mentioned PET image reconstruction method is realized.
- the invention also discloses a computer device.
- the computer device includes a memory, a processor, and a PET image reconstruction program stored in the memory.
- the PET image reconstruction program is implemented by the processor to implement any A reconstruction method of the aforementioned PET image.
- a PET image reconstruction method disclosed by the present invention overcomes the problem of a large amount of noise in the reconstructed image by adding Poisson random noise variables in the reconstruction process, and performs image reconstruction through the neighbor block prior, which overcomes the current In the prior art, a single pixel difference is used to distinguish between true edges and noise fluctuations, resulting in the problem that the reconstructed image does not retain the correct edge.
- the first reconstructed image is further iteratively updated based on dictionary learning to remove image noise and pseudo Therefore, the reconstruction algorithm of the present invention does not depend on the degree of conformity of the anatomical structure information and the functional information. Whether or not noise interferes with the image edge, the image edge can be well distinguished.
- FIG. 1 is a flowchart of a PET image reconstruction method according to an embodiment of the present invention
- FIG. 2 is a flowchart of iteratively updating the initial reconstructed image X according to the first objective function to obtain the first reconstructed image according to an embodiment of the present invention
- FIG. 3 is a flowchart of iteratively updating the first reconstructed image according to the second objective function to obtain the second reconstructed image according to an embodiment of the present invention
- 4a to 4d are PET images obtained by different reconstruction methods, respectively.
- the method of PET image reconstruction according to an embodiment of the present invention includes the following steps:
- Step 1 S10 Obtain the projection data Y of the PET image and the system matrix P.
- the projection data of the PET image may be simulated data, that is, the simulated projection data corresponding to the existing simulated PET image may be used, and the existing system matrix may be obtained.
- the projection data can be obtained by scanning with the PET scanning system, and then the inherent system matrix of the PET system is calculated based on the geometric structure information of the PET scanning system.
- Step 3 S30 Acquire the initial reconstructed image X, and iteratively update the initial reconstructed image X according to the first objective function to obtain the first reconstructed image.
- the first objective function is:
- Q L (X; X n ) is a likelihood agent function constructed based on Poisson random distribution variables
- X n is the reconstructed image obtained after the nth iteration
- ⁇ is the regularization parameter
- n j represents the total number of pixels
- p ij represents the probability that the jth pixel is detected by the i-th detector
- n i represents the total number of detectors
- p j represents the total probability of the j-th pixel detected by the n i detectors
- X j represents the value of the jth pixel of the reconstructed image X
- y i represents the projection data detected by the ith detector
- Represents the desired projection data which is obtained by affine transformation of the initial reconstructed image X at each iteration
- w jk represents the weight between the jth pixel and the kth pixel of the reconstructed image X n
- N j represents the neighborhood block centered on the jth pixel
- j l is the lth pixel in the neighborhood block f j (X)
- k l Is the lth pixel in the neighborhood block f k (X)
- h l is a positive weight vector.
- step three specifically includes the following steps:
- Step S31 Acquire the desired maximized image according to the initialized reconstructed image X, the projection data Y, and the system matrix P.
- Step S32 Perform image smoothing according to the initial reconstructed image X to obtain an intermediate image.
- each pixel to obtain an intermediate image.
- Step S33 Generate a first reconstructed image according to the desired maximized image and the intermediate image.
- Step 4 S40 Iteratively update the first reconstructed image according to the second objective function to obtain a second reconstructed image, where the second objective function is a function constructed based on dictionary learning.
- the expression of the second objective function is: st.
- X represents the first reconstructed image obtained in step three reconstruction
- R ij is the operation to obtain the image block from X
- D is a dictionary based on the image block
- ⁇ ij is a sparse representation of X ij about the dictionary D
- T 0 indicates the sparse level to be achieved.
- this step includes the following steps:
- Step S41 Perform image segmentation on the first reconstructed image to generate multiple image blocks.
- the first reconstructed image may be segmented by a preset image segmentation algorithm to generate multiple image blocks of the first reconstructed image.
- the size of each image block is the same when dividing, and there is an overlapping area between the adjacent image blocks.
- Step S42 Generate a sparse coefficient for each image block according to each image block, and the pre-trained low-resolution dictionary and high-resolution dictionary.
- this step includes the following steps:
- a preset coefficient calculation formula calculate a sparse coefficient of the image block that satisfies the first coefficient constraint condition and the second coefficient constraint condition.
- Step S43 Generate a high-resolution image corresponding to the first reconstructed image according to the sparse coefficient of each image block and the high-resolution dictionary;
- Step S44 Generate and output a second reconstructed image according to the first reconstructed image, the high-resolution image, the preset blur matrix, and the preset down-sampling matrix. This completes an iterative process.
- the reconstruction method before performing image segmentation on the first reconstructed image, the reconstruction method further includes:
- the preset low resolution PET training image set the preset high resolution PET training image set, the size of the image block in the low resolution PET training image set, and the size of the image block in the high resolution PET image set, the low resolution Rate dictionary and the high-resolution dictionary for joint training.
- FIG. 4a is the original simulated PET emission image, which is used as a comparison image in this embodiment
- FIG. 4b is obtained based on the regularization of a single pixel
- Figure 4c is a PET reconstruction map obtained based on neighborhood block regularization
- Figure 4d is a PET reconstruction map obtained by the reconstruction method of the present invention. It can be seen from FIGS. 4b and 4c that although the method based on the regularization of single pixels and domain blocks can reconstruct the edges and tumor areas of the image, the reconstructed image contains a lot of noise and the tumor edges are more blurred. The tumor areas contain artifacts. It can be seen from FIG. 4d that the image reconstructed by the method of the present invention suppresses noise and artifacts, and at the same time the edge of the tumor area and some detailed information are also well preserved.
- a PET image reconstruction method disclosed by the present invention overcomes the problem of a large amount of noise in the reconstructed image by adding Poisson random noise variables in the reconstruction process, and performs image reconstruction through the neighbor block prior, which overcomes the current In the prior art, a single pixel difference is used to distinguish between true edges and noise fluctuations, resulting in the problem that the reconstructed image does not retain the correct edge.
- the first reconstructed image is further iteratively updated based on dictionary learning to remove image noise and pseudo Therefore, the reconstruction algorithm of the present invention does not depend on the degree of conformity of the anatomical structure information and the functional information. Whether or not noise interferes with the image edge, the image edge can be well distinguished.
- an embodiment of the present invention also discloses a computer storage medium in which a PET image reconstruction program is stored, and the PET image reconstruction program implements the aforementioned PET image reconstruction when executed by the processor method.
- an embodiment of the present invention also discloses a computer device, the computer device includes a memory, a processor, and a PET image reconstruction program stored in the memory, the PET image reconstruction program is executed by the processor To realize the above-mentioned PET image reconstruction method.
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Biomedical Technology (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Optics & Photonics (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Veterinary Medicine (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- High Energy & Nuclear Physics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
一种PET图像的重建方法、计算机存储介质、计算机设备。该方法包括:步骤一:获取PET图像的投影数据Y和系统矩阵P(S10);步骤二:构建成像模型方程Y=PX,X为重建的PET图像(S20);步骤三:获取初始重建图像X,并根据第一目标函数迭代更新初始重建图像X以获得第一重建图像(S30);步骤四:根据第二目标函数迭代更新第一重建图像以获得第二重建图像(S40);步骤五:判断是否满足迭代条件,若是,则输出本轮迭代得到第二重建图像作为最终的PET重建图像;若否,则返回步骤三,并将本轮迭代的第二重建图像作为下一轮迭代时的初始重建图像(S50)。上述重建方法不依赖解剖结构的信息和功能信息的符合程度,无论是否有噪声干扰图像边缘,都能很好的将图像边缘区分出来。
Description
本发明属于信息技术领域,特别涉及PET图像的重建方法、计算机存储介质、计算机设备。
正电子发射型断层(positron emission tomography,PET),即正电子断层成像,是首先通过向病人体内注射放射性示踪剂,然后测量病人体内放射性同位素的分布来进行成像的。PET重建算法主要分为解析重建算法和迭代重建算法两类。解析重建算法主要包括反投影,滤波反投影以及傅里叶重建。其中最广泛应用的算法是滤波反投影(Filteredback-projection,FBP)。FBP的方法基于莱登(Radon)变换,但FBP没有考虑系统响应的时空不均一性,也没有考虑到仪器在测量时的噪声,所以重建出的图像含有大量噪声。迭代重建算法又包括代数重建和统计重建,其中代数重建中主要有代数重建算法(Algebra Reconstruction Technique,ART)及在此基础上进一步扩展得到的一些新的算法。统计重建中极大似然-期望最大化(Maximum likelihood-expectation maximization,ML-EM)方法因其在病变检测方面比传统的算法有更好的性能,目前在临床和实践中被广泛应用,但该方法随迭代次数的增加图像质量退化,会产生“棋盘格伪影”。提前终止迭代和在似然函数中集成惩罚项或某种先验知识,在一定程度上克服了ML-EM方法存在的问题。
综上,现有的PET重建方法存在如下问题:1)由于在有噪声影响的图像重建中仅用单个像素差来区分真实边缘和噪声波动,导致重建的图像没有保留正确的边缘;2)由于在去除噪声和保留细节信息之间没有很好的权衡,导致重建出的图像丢失很多细节信息;3)对于欠采样图像和噪声图像,由于没有好的约束条件,将导致细节丢失和产生块状的伪影。
发明内容
(一)要解决的技术问题
本发明要解决的技术问题是:如何得到重建效果更好的PET图像。
(二)技术方案
为了实现上述的目的,本发明采用了如下的技术方案:
一种PET图像的重建方法,包括如下步骤:
步骤一:获取PET图像的投影数据Y和系统矩阵P;
步骤二:构建成像模型方程Y=PX,X为重建的PET图像。
步骤三:获取初始重建图像X,并根据第一目标函数迭代更新初始重建图像X以获得第一重建图像,第一目标函数为:
步骤四:根据第二目标函数迭代更新第一重建图像以获得第二重建图像,其中所述第二目标函数为基于字典学习构建的函数;
步骤五:判断是否满足迭代条件,若是,则输出本轮迭代得到第二重建图像作为最终的PET重建图像;若否,则返回步骤三,并将本轮迭代的第二重建图像作为下一轮迭代时的初始重建图像。
可选地,所述似然代理函数的表达式为:
其中,其中,
n
j表示像素的总量,p
ij表示第j个像素被第i个探测器探测到概率,n
i表示探测器的总数,p
j表示第j个像素被n
i个探测器探测到总概率值,X
j表示重建图像X的第j个像 素的值,
表示经过第n次迭代后的重建图像X
n的第j个像素的值,y
i表示第i个探测器探测到的投影数据,
表示期望投影数据,
表示期望最大化图像的第j个像素的值。
其中,
表示第j个像素的权重,w
jk表示重建图像X
n的第j个像素与第k个像素之间的权重,
表示中间图像的第j个像素的值,N
j表示以第j个像素为中心的邻域块,
表示经过第n次迭代后的重建图像X
n的第j个像素的邻域块中的第k个像素的值,j
l为邻域块f
j(X)中的第l个像素,k
l为邻域块f
k(X)中的第l个像素,h
l为正权重向量。
可选地,所述根据第一目标函数迭代更新初始化重建图像X以获得第一重建图像的方法包括:
根据所述初始化重建图像X、投影数据Y和系统矩阵P获取期望最大化图像;
对所述初始化重建图像X进行图像平滑处理以获取中间图像;
根据所述期望最大化图像和所述中间图像生成第一重建图像。
可选地,所述第二目标函数的表达式为:
其中,X表示步骤三重建得到的第一重建图像,R
ij是从X中获得图像块的操作,D是以图像块为基的字典,α
ij是关于字典D的X
ij的稀疏表示,T
0表示所要达到的稀疏水平。
可选地,所述根据第二目标函数迭代更新第一重建图像以获得第二重建图像的方法包括:
对第一重建图像进行图像分割,生成多个图像块;
根据每个所述图像块、和预先训练好的低分辨率字典和高分辨率字典生成每个图像块的稀疏系数;
根据所述每个图像块的稀疏系数、和所述高分辨率字典生成所述第一重建图像对应的高分辨率图像;
根据所述第一重建图像、所述高分辨率图像、预设的模糊矩阵和预设的下采样矩阵,生成并输出第二重建图像。
可选地,所述根据每个所述图像块、预先训练好的低分辨率字典和高分辨率字典生成每个图像块的稀疏系数的方法:
根据所述图像块、所述低分辨率字典、预设的特征提取函数和预设第一阈值,构建第一系数约束条件;
根据所述图像块、所述图像块的与上一个图像块的重叠区域、所述高分辨率字典和预设的第二阈值,构建第二系数约束条件;
根据预设的系数计算公式,计算满足所述第一系数约束条件和所述第二系数约束条件的所述图像块的稀疏系数。
可选地,在对第一重建图像进行图像分割之前,所述重建方法还包括:
对所述低分辨率字典和所述高分辨字典进行随机初始化;
根据预设的低分辨率PET训练图像集、预设的高分辨率PET训练图像集、所述低分辨率PET训练图像集中图像块的尺寸、以及所述高分辨率PET图像集中图像块的尺寸,对所述低分辨率字典和所述高分辨字典进行联合训练。
本发明还公开了一种计算机存储介质,所述计算机存储介质中存储有PET图像的重建程序,所述PET图像的重建程序被处理器执行时实现上述的PET图像的重建方法。
本发明还公开了一种计算机设备,所述计算机设备包括存储器、处理器、存储在所述存储器中的PET图像的重建程序,所述PET图像的重建程序被所述处理器执行时实现任一种上述的PET图像的重建方法。
(三)有益效果
本发明公开的一种PET图像的重建方法,通过在重建过程中加入泊松随机噪声变量来克服重建的图像中含有大量噪声的问题,并通过邻域块先验来进行图像重建,克服了现有技术中利用单个像素差来区分真实边缘和噪声波动,导致重建的图像没有保留正确的边缘的问题,另外通过对第一重建图像进一步地进行基于字典学习的更新迭代,以去除图像噪声和伪影,因此本发明的重建算法不依赖解剖结构的信息和功能信息的符合程度,无论是否有噪声干扰图像边缘,都能很好的将图像边缘区分出来。
图1是本发明的实施例的PET图像的重建方法的流程图;
图2是本发明的实施例的根据第一目标函数迭代更新初始重建图像X以获得第一重建图像的流程图;
图3是本发明的实施例的根据第二目标函数迭代更新第一重建图像以获得第二重建图像的流程图;
图4a至图4d分别为通过不同的重建方法得到的PET图像。
为了使本发明的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本发明进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。
如图1所示,根据本发明的实施例的PET图像的重建方法包括如下步骤:
步骤一S10:获取PET图像的投影数据Y和系统矩阵P。
一方面,从仿真实验角度考虑,PET图像的投影数据可为模拟数据,即利用现有模拟PET图像对应的模拟的投影数据,另外可获取现有的系统矩阵。另一方面,从实际测量角度考虑,投影数据可通过PET扫描系统进行扫描来获 得,接着根据PET扫描系统的几何结构信息计算出PET系统固有的系统矩阵。
步骤二S20:构建成像模型方程Y=PX,X为重建的PET图像。
步骤三S30:获取初始重建图像X,并根据第一目标函数迭代更新初始重建图像X以获得第一重建图像,第一目标函数为:
具体地,初始化参数,设置最大迭代次数为Maxlter=100,正则化参数β=2
-7,高参数(hyper-parameter)δ=1e
-9。初始化图像,将初始重建图像X赋值为
其中,j表示第j个像素。
进一步地,似然代理函数的表达式为:
其中,
n
j表示像素的总量,p
ij表示第j个像素被第i个探测器探测到概率,n
i表示探测器的总数,p
j表示第j个像素被n
i个探测器探测到总概率值,X
j表示重建图像X的第j个像素的值,
表示经过第n次迭代后的重建图像X
n的第j个像素的值,y
i表示第i个探测器探测到的投影数据,
表示期望投影数据,其由每次迭代时的初始重建图像X通过仿射变换得到的,
表示期望最大化图像的第j个像素的值。
进一步地,惩罚代理函数的表达式为:
其中,
表示第j个像素的权重,w
jk表示重建图像X
n的第j个像素与第k个像素之间的权重,
表示平滑图像的第j个像素的值,N
j表示以第j个像素为中心的邻域块,
表示经过第n次迭代后的重建图像X
n的第j个像素的邻域块中的第k个像素的值,j
l为邻域块f
j(X)中的第l个像素,k
l为邻域块f
k(X)中的第l个像素,h
l为正权重向量。
进一步地,如图2所示,步骤三具体包括如下步骤:
步骤S31:根据所述初始化重建图像X、投影数据Y和系统矩阵P获取期望最大化图像。
步骤S32:根据初始化重建图像X进行图像平滑以获取中间图像。
步骤S33:根据期望最大化图像和中间图像生成第一重建图像。
步骤四S40:根据第二目标函数迭代更新第一重建图像以获得第二重建图像,其中所述第二目标函数为基于字典学习构建的函数。
其中,X表示步骤三重建得到的第一重建图像,R
ij是从X中获得图像块的 操作,D是以图像块为基的字典,α
ij是关于字典D的X
ij的稀疏表示,T
0表示所要达到的稀疏水平。
进一步地,如图3所示,该步骤包括如下步骤:
步骤S41:对第一重建图像进行图像分割,生成多个图像块。
具体地,可通过预设的图像分割算法对第一重建图像进行分割,生成多个第一重建图像的图像块。其中,在分割时每个图像块的尺寸相同,且前后相邻的图像块之间存在重叠区域。
步骤S42:根据每个图像块、和预先训练好的低分辨率字典和高分辨率字典生成每个图像块的稀疏系数。
具体来说,该步骤包括如下步骤:
根据所述图像块、所述低分辨率字典、预设的特征提取函数和预设第一阈值,构建第一系数约束条件;
根据所述图像块、所述图像块的与上一个图像块的重叠区域、所述高分辨率字典和预设的第二阈值,构建第二系数约束条件;
根据预设的系数计算公式,计算满足所述第一系数约束条件和所述第二系数约束条件的所述图像块的稀疏系数。
步骤S43:根据每个图像块的稀疏系数和高分辨率字典生成所述第一重建图像对应的高分辨率图像;
步骤S44:根据第一重建图像、高分辨率图像、预设的模糊矩阵和预设的下采样矩阵,生成并输出第二重建图像。这样即完成了一次迭代过程。
进一步地,在对第一重建图像进行图像分割之前,重建方法还包括:
对所述低分辨率字典和所述高分辨字典进行随机初始化;
根据预设的低分辨率PET训练图像集、预设的高分辨率PET训练图像集、低分辨率PET训练图像集中图像块的尺寸、以及高分辨率PET图像集中 图像块的尺寸,对低分辨率字典和所述高分辨字典进行联合训练。
步骤五S50:判断是否满足迭代条件,若是,则输出本轮迭代得到第二重建图像作为最终的PET重建图像;若否,则返回步骤三,并将本轮迭代的第二重建图像作为下一轮迭代时的初始重建图像,对于本实施例来说,迭代条件为迭代次数达到最大迭代次数Maxlter=100。
下面对现有技术的重建方法和本发明的重建方法得到的各个图像进行对比,图4a为原始的模拟PET发射图,在本实施例中作为对比图像,图4b为基于单个像素正则化得到的PET重建图,图4c为基于邻域块正则化得到PET重建图,图4d为本发明的重建方法得到的PET重建图。从图4b和图4c可以看出,虽然基于单个像素和领域块正则化的方法可以重建出图像的边缘和肿瘤区域,但重建出的图像含有大量噪声及肿瘤边缘较模糊肿瘤区域含有伪影,由图4d可以看到本发明方法重建得到的图像抑制了噪声和伪影同时肿瘤区域的边缘和一些细节信息也被很好的保留。
本发明公开的一种PET图像的重建方法,通过在重建过程中加入泊松随机噪声变量来克服重建的图像中含有大量噪声的问题,并通过邻域块先验来进行图像重建,克服了现有技术中利用单个像素差来区分真实边缘和噪声波动,导致重建的图像没有保留正确的边缘的问题,另外通过对第一重建图像进一步地进行基于字典学习的更新迭代,以去除图像噪声和伪影,因此本发明的重建算法不依赖解剖结构的信息和功能信息的符合程度,无论是否有噪声干扰图像边缘,都能很好的将图像边缘区分出来。
进一步地,本发明的实施例还公开了一种计算机存储介质,所述计算机存储介质中存储有PET图像的重建程序,所述PET图像的重建程序被处理器执行时实现上述的PET图像的重建方法。
进一步地,本发明的实施例还公开了计算机设备,所述计算机设备包括存储器、处理器、存储在所述存储器中的PET图像的重建程序,所述PET图像的重建程序被所述处理器执行时实现上述的PET图像的重建方法。
上面对本发明的具体实施方式进行了详细描述,虽然已表示和描述了一些实施例,但本领域技术人员应该理解,在不脱离由权利要求及其等同物限 定其范围的本发明的原理和精神的情况下,可以对这些实施例进行修改和完善,这些修改和完善也应在本发明的保护范围内。
Claims (17)
- 一种PET图像的重建方法,其中,包括如下步骤:步骤一:获取PET图像的投影数据Y和系统矩阵P;步骤二:构建成像模型方程Y=PX,X为重建的PET图像。步骤三:获取初始重建图像X,并根据第一目标函数迭代更新初始重建图像X以获得第一重建图像,第一目标函数为:步骤四:根据第二目标函数迭代更新第一重建图像以获得第二重建图像,其中所述第二目标函数为基于字典学习构建的函数;步骤五:判断是否满足迭代条件,若是,则输出本轮迭代得到第二重建图像作为最终的PET重建图像;若否,则返回步骤三,并将本轮迭代的第二重建图像作为下一轮迭代时的初始重建图像。
- 根据权利要求1所述的PET图像的重建方法,其中,所述根据第一目标函数迭代更新初始化重建图像X以获得第一重建图像的方法包括:根据所述初始化重建图像X、投影数据Y和系统矩阵P获取期望最大化图像;对所述初始化重建图像X进行图像平滑处理以获取中间图像;根据所述期望最大化图像和所述中间图像生成第一重建图像。
- 根据权利要求1所述的PET图像的重建方法,其中,所述根据第二目标函数迭代更新第一重建图像以获得第二重建图像的方法包括:对第一重建图像进行图像分割,生成多个图像块;根据每个所述图像块、和预先训练好的低分辨率字典和高分辨率字典生 成每个图像块的稀疏系数;根据所述每个图像块的稀疏系数、和所述高分辨率字典生成所述第一重建图像对应的高分辨率图像;根据所述第一重建图像、所述高分辨率图像、预设的模糊矩阵和预设的下采样矩阵,生成并输出第二重建图像。
- 根据权利要求6所述的PET图像的重建方法,其中,所述根据每个所述图像块、预先训练好的低分辨率字典和高分辨率字典生成每个图像块的稀疏系数的方法:根据所述图像块、所述低分辨率字典、预设的特征提取函数和预设第一阈值,构建第一系数约束条件;根据所述图像块、所述图像块的与上一个图像块的重叠区域、所述高分辨率字典和预设的第二阈值,构建第二系数约束条件;根据预设的系数计算公式,计算满足所述第一系数约束条件和所述第二系数约束条件的所述图像块的稀疏系数。
- 根据权利要求6所述的PET图像的重建方法,其中,在对第一重建图像进行图像分割之前,所述重建方法还包括:对所述低分辨率字典和所述高分辨字典进行随机初始化;根据预设的低分辨率PET训练图像集、预设的高分辨率PET训练图像集、所述低分辨率PET训练图像集中图像块的尺寸、以及所述高分辨率PET图像集中图像块的尺寸,对所述低分辨率字典和所述高分辨字典进行联合训练。
- 一种计算机存储介质,其特征在于,所述计算机存储介质中存储有PET图像的重建程序,所述PET图像的重建程序被处理器执行时实现如权利要求1所述的PET图像的重建方法。
- 一种计算机设备,其中,所述计算机设备包括存储器、处理器、存 储在所述存储器中的PET图像的重建程序,所述PET图像的重建程序被所述处理器执行时实现PET图像的重建方法,所述PET图像的重建方法包括:步骤一:获取PET图像的投影数据Y和系统矩阵P;步骤二:构建成像模型方程Y=PX,X为重建的PET图像。步骤三:获取初始重建图像X,并根据第一目标函数迭代更新初始重建图像X以获得第一重建图像,第一目标函数为:步骤四:根据第二目标函数迭代更新第一重建图像以获得第二重建图像,其中所述第二目标函数为基于字典学习构建的函数;步骤五:判断是否满足迭代条件,若是,则输出本轮迭代得到第二重建图像作为最终的PET重建图像;若否,则返回步骤三,并将本轮迭代的第二重建图像作为下一轮迭代时的初始重建图像。
- 根据权利要求10所述的计算机设备,其中,所述根据第一目标函数迭代更新初始化重建图像X以获得第一重建图像的方法包括:根据所述初始化重建图像X、投影数据Y和系统矩阵P获取期望最大化图像;对所述初始化重建图像X进行图像平滑处理以获取中间图像;根据所述期望最大化图像和所述中间图像生成第一重建图像。
- 根据权利要求10所述的计算机设备,其中,所述根据第二目标函数迭代更新第一重建图像以获得第二重建图像的方法包括:对第一重建图像进行图像分割,生成多个图像块;根据每个所述图像块、和预先训练好的低分辨率字典和高分辨率字典生 成每个图像块的稀疏系数;根据所述每个图像块的稀疏系数、和所述高分辨率字典生成所述第一重建图像对应的高分辨率图像;根据所述第一重建图像、所述高分辨率图像、预设的模糊矩阵和预设的下采样矩阵,生成并输出第二重建图像。
- 根据权利要求15所述的计算机设备,其中,所述根据每个所述图像块、预先训练好的低分辨率字典和高分辨率字典生成每个图像块的稀疏系数的方法:根据所述图像块、所述低分辨率字典、预设的特征提取函数和预设第一阈值,构建第一系数约束条件;根据所述图像块、所述图像块的与上一个图像块的重叠区域、所述高分辨率字典和预设的第二阈值,构建第二系数约束条件;根据预设的系数计算公式,计算满足所述第一系数约束条件和所述第二系数约束条件的所述图像块的稀疏系数。
- 根据权利要求15所述的计算机设备,其中,在对第一重建图像进行图像分割之前,所述重建方法还包括:对所述低分辨率字典和所述高分辨字典进行随机初始化;根据预设的低分辨率PET训练图像集、预设的高分辨率PET训练图像集、所述低分辨率PET训练图像集中图像块的尺寸、以及所述高分辨率PET图像集中图像块的尺寸,对所述低分辨率字典和所述高分辨字典进行联合训练。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US17/287,040 US20210366168A1 (en) | 2018-12-14 | 2019-01-15 | Pet image reconstruction method, computer storage medium, and computer device |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811536162.1 | 2018-12-14 | ||
| CN201811536162.1A CN109712209B (zh) | 2018-12-14 | 2018-12-14 | Pet图像的重建方法、计算机存储介质、计算机设备 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020118844A1 true WO2020118844A1 (zh) | 2020-06-18 |
Family
ID=66256570
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2019/071746 Ceased WO2020118844A1 (zh) | 2018-12-14 | 2019-01-15 | Pet图像的重建方法、计算机存储介质、计算机设备 |
Country Status (3)
| Country | Link |
|---|---|
| US (1) | US20210366168A1 (zh) |
| CN (1) | CN109712209B (zh) |
| WO (1) | WO2020118844A1 (zh) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508813A (zh) * | 2020-12-04 | 2021-03-16 | 上海交通大学 | 一种基于改进Kernel方法结合稀疏约束的PET图像重建方法 |
| CN112634388A (zh) * | 2020-11-30 | 2021-04-09 | 明峰医疗系统股份有限公司 | 一种ct迭代重建代价函数的优化方法及ct图像重建方法、系统及ct |
| CN115423890A (zh) * | 2022-09-15 | 2022-12-02 | 京心禾(北京)医疗科技有限公司 | 一种断层图像迭代重建方法 |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110276813B (zh) * | 2019-05-06 | 2023-01-24 | 深圳先进技术研究院 | Ct图像重建方法、装置、存储介质和计算机设备 |
| CN110580689A (zh) * | 2019-08-19 | 2019-12-17 | 深圳先进技术研究院 | 一种图像重建方法及装置 |
| CN111161182B (zh) * | 2019-12-27 | 2021-03-09 | 南方医科大学 | Mr结构信息约束的非局部均值引导的pet图像部分容积校正方法 |
| EP4136611B1 (en) * | 2020-05-18 | 2025-10-15 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image reconstruction |
| CA3124052C (en) | 2020-06-08 | 2022-05-10 | Guangzhou Computational Super-Resolution Biotech Co., Ltd. | Systems and methods for image processing |
| CN112270643B (zh) * | 2020-09-04 | 2024-06-14 | 深圳市菲森科技有限公司 | 一种三维成像数据拼接方法、装置、电子设备及存储介质 |
| CN112085808B (zh) * | 2020-09-14 | 2024-04-05 | 深圳先进技术研究院 | 一种ct图像重建方法、系统、设备和介质 |
| CN112365593B (zh) * | 2020-11-12 | 2024-03-29 | 江苏赛诺格兰医疗科技有限公司 | 一种pet图像重建的方法和系统 |
| CN112652029A (zh) * | 2020-12-04 | 2021-04-13 | 深圳先进技术研究院 | Pet成像方法、装置与设备 |
| WO2022116143A1 (zh) * | 2020-12-04 | 2022-06-09 | 深圳先进技术研究院 | Pet成像方法、装置与设备 |
| CN112488952B (zh) * | 2020-12-07 | 2024-09-13 | 深圳先进技术研究院 | Pet图像的重建方法及重建终端、计算机可读存储介质 |
| CN114494294B (zh) * | 2022-01-25 | 2022-10-14 | 北京市测绘设计研究院 | 地表覆盖数据处理方法及装置、电子设备和存储介质 |
| CN114782570B (zh) * | 2022-03-25 | 2025-10-03 | 中国科学院深圳先进技术研究院 | 双示踪剂ect成像方法、装置、设备及存储介质 |
| CN114926559B (zh) * | 2022-05-09 | 2025-06-03 | 浙江大学 | 一种基于字典学习思想无衰减校正的pet重建方法 |
| CN114998224B (zh) * | 2022-05-16 | 2025-03-18 | 浙江理工大学 | 一种工业图像的压缩重构方法 |
| CN116452425B (zh) * | 2023-06-08 | 2023-09-22 | 常州星宇车灯股份有限公司 | 图像超分辨率重建方法、设备及介质 |
| CN116739933B (zh) * | 2023-06-20 | 2026-04-17 | 安徽大学 | 一种融合系统响应矩阵和神经网络的pet图像生成方法 |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130188856A1 (en) * | 2011-12-01 | 2013-07-25 | Varian Medical Systems, Inc. | Systems and methods for real-time target validation for image-guided radiation therapy |
| CN104077763A (zh) * | 2014-07-08 | 2014-10-01 | 中国科学院高能物理研究所 | 基于压缩感知理论的tof-pet图像重建方法 |
| CN104574459A (zh) * | 2014-12-29 | 2015-04-29 | 沈阳东软医疗系统有限公司 | 一种pet图像的重建方法和装置 |
| CN105374060A (zh) * | 2015-10-15 | 2016-03-02 | 浙江大学 | 一种基于结构字典约束的pet图像重建方法 |
| CN107680146A (zh) * | 2017-09-13 | 2018-02-09 | 深圳先进技术研究院 | Pet图像的重建方法、装置、设备及存储介质 |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102010029281A1 (de) * | 2010-05-25 | 2011-12-01 | Siemens Aktiengesellschaft | Verfahren und Bildrekonstruktionseinrichtung zur Rekonstruktion von Bilddaten |
| EP2601639A1 (en) * | 2010-08-04 | 2013-06-12 | Koninklijke Philips Electronics N.V. | Method and system for iterative image reconstruction |
| WO2013106512A1 (en) * | 2012-01-10 | 2013-07-18 | The Johns Hopkins University | Methods and systems for tomographic reconstruction |
| CN103136731B (zh) * | 2013-02-05 | 2015-11-25 | 南方医科大学 | 一种动态pet图像的参数成像方法 |
| US9595120B2 (en) * | 2015-04-27 | 2017-03-14 | Siemens Healthcare Gmbh | Method and system for medical image synthesis across image domain or modality using iterative sparse representation propagation |
| CN106447610B (zh) * | 2016-08-31 | 2019-07-23 | 重庆大学 | 图像重建方法及装置 |
| CN106683146B (zh) * | 2017-01-11 | 2021-01-15 | 上海联影医疗科技股份有限公司 | 一种图像重建方法和图像重建算法的参数确定方法 |
| EP3404615A1 (en) * | 2017-05-17 | 2018-11-21 | Koninklijke Philips N.V. | Iterative image reconstruction/de-noising with artifact reduction |
-
2018
- 2018-12-14 CN CN201811536162.1A patent/CN109712209B/zh active Active
-
2019
- 2019-01-15 WO PCT/CN2019/071746 patent/WO2020118844A1/zh not_active Ceased
- 2019-01-15 US US17/287,040 patent/US20210366168A1/en not_active Abandoned
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130188856A1 (en) * | 2011-12-01 | 2013-07-25 | Varian Medical Systems, Inc. | Systems and methods for real-time target validation for image-guided radiation therapy |
| CN104077763A (zh) * | 2014-07-08 | 2014-10-01 | 中国科学院高能物理研究所 | 基于压缩感知理论的tof-pet图像重建方法 |
| CN104574459A (zh) * | 2014-12-29 | 2015-04-29 | 沈阳东软医疗系统有限公司 | 一种pet图像的重建方法和装置 |
| CN105374060A (zh) * | 2015-10-15 | 2016-03-02 | 浙江大学 | 一种基于结构字典约束的pet图像重建方法 |
| CN107680146A (zh) * | 2017-09-13 | 2018-02-09 | 深圳先进技术研究院 | Pet图像的重建方法、装置、设备及存储介质 |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112634388A (zh) * | 2020-11-30 | 2021-04-09 | 明峰医疗系统股份有限公司 | 一种ct迭代重建代价函数的优化方法及ct图像重建方法、系统及ct |
| CN112634388B (zh) * | 2020-11-30 | 2024-01-02 | 明峰医疗系统股份有限公司 | 一种ct迭代重建代价函数的优化方法及ct图像重建方法、系统及ct |
| CN112508813A (zh) * | 2020-12-04 | 2021-03-16 | 上海交通大学 | 一种基于改进Kernel方法结合稀疏约束的PET图像重建方法 |
| CN115423890A (zh) * | 2022-09-15 | 2022-12-02 | 京心禾(北京)医疗科技有限公司 | 一种断层图像迭代重建方法 |
| CN115423890B (zh) * | 2022-09-15 | 2023-09-19 | 京心禾(北京)医疗科技有限公司 | 一种断层图像迭代重建方法 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109712209A (zh) | 2019-05-03 |
| CN109712209B (zh) | 2022-09-20 |
| US20210366168A1 (en) | 2021-11-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020118844A1 (zh) | Pet图像的重建方法、计算机存储介质、计算机设备 | |
| CN113516659B (zh) | 一种基于深度学习的医学影像自动分割方法 | |
| Wang et al. | PET image reconstruction using kernel method | |
| CN111325686A (zh) | 一种基于深度学习的低剂量pet三维重建方法 | |
| CN102968762B (zh) | 一种基于稀疏化和泊松模型的pet重建方法 | |
| CN110827216A (zh) | 图像去噪的多生成器生成对抗网络学习方法 | |
| Qi | A unified noise analysis for iterative image estimation | |
| CN107527359A (zh) | 一种pet图像重建方法及pet成像设备 | |
| CN104657950B (zh) | 一种基于Poisson TV的动态PET图像重建方法 | |
| CN104700438B (zh) | 图像重建方法及装置 | |
| Xu et al. | Efficient low‐dose CT artifact mitigation using an artifact‐matched prior scan | |
| WO2023279316A1 (zh) | 一种基于去噪打分匹配网络的pet重建方法 | |
| Chan et al. | An attention-based deep convolutional neural network for ultra-sparse-view CT reconstruction | |
| Van Eyndhoven et al. | Region-based iterative reconstruction of structurally changing objects in CT | |
| CN111161182B (zh) | Mr结构信息约束的非局部均值引导的pet图像部分容积校正方法 | |
| Sureau et al. | Convergent admm plug and play pet image reconstruction | |
| CN110580689A (zh) | 一种图像重建方法及装置 | |
| CN112365593B (zh) | 一种pet图像重建的方法和系统 | |
| CN113469915A (zh) | 一种基于去噪打分匹配网络的pet重建方法 | |
| CN118644570A (zh) | 一种超稀疏角ct图像的重建方法 | |
| KR20240101340A (ko) | 두 장의 투영 영상을 이용한 삼차원 물체 복원 뉴럴 소프트 최적화 방법 및 그 시스템 | |
| CN108010093B (zh) | 一种pet图像重建方法和装置 | |
| CN117152291A (zh) | 一种基于原对偶算法的非凸加权变分金属伪影去除方法 | |
| Chen et al. | Joint-MAP tomographic reconstruction with patch similarity based mixture prior model | |
| Gong et al. | Learning personalized representation for inverse problems in medical imaging using deep neural network |
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: 19895565 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 03.11.2021) |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 19895565 Country of ref document: EP Kind code of ref document: A1 |

