WO2014172927A1 - 一种基于rpca的pet图像动态重建方法及系统 - Google Patents

一种基于rpca的pet图像动态重建方法及系统 Download PDF

Info

Publication number
WO2014172927A1
WO2014172927A1 PCT/CN2013/075593 CN2013075593W WO2014172927A1 WO 2014172927 A1 WO2014172927 A1 WO 2014172927A1 CN 2013075593 W CN2013075593 W CN 2013075593W WO 2014172927 A1 WO2014172927 A1 WO 2014172927A1
Authority
WO
WIPO (PCT)
Prior art keywords
matrix
pet
iteration
correction coefficient
kth iteration
Prior art date
Application number
PCT/CN2013/075593
Other languages
English (en)
French (fr)
Inventor
刘华锋
于行健
Original Assignee
浙江大学
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 浙江大学 filed Critical 浙江大学
Priority to US14/786,534 priority Critical patent/US9576379B2/en
Publication of WO2014172927A1 publication Critical patent/WO2014172927A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/002D [Two Dimensional] image generation
    • G06T11/003Reconstruction from projections, e.g. tomography
    • G06T11/006Inverse problem, transformation from projection-space into object-space, e.g. transform methods, back-projection, algebraic methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10072Tomographic images
    • G06T2207/10104Positron emission tomography [PET]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2211/00Image generation
    • G06T2211/40Computed tomography
    • G06T2211/424Iterative

Definitions

  • the invention belongs to the field of PET imaging technology, and particularly relates to a PET image dynamic reconstruction method and system based on PRCA (robust principal component analysis method). Background technique
  • PET Positron emission tomography
  • PET is a medical imaging technology based on nuclear physics and molecular biology. It can observe the metabolic activity of cells at the molecular level, and detect and prevent early diseases, especially tumors. Provides a valid basis.
  • PET essentially images the concentration distribution of a drug in a patient.
  • the radioisotope labeled drug injected into the patient enters the circulatory system through the blood, and these substances form a certain concentration distribution in various tissues and organs in the human body. Since the radioactive isotope has a short half-life and is extremely unstable, it will decay rapidly. The positrons released during the decay process are quenched with nearby free electrons, resulting in a pair of directions almost opposite, equal energy, and energy.
  • PET has become more and more widely used in practical medicine, but at the same time, the requirements for PET imaging have become higher in clinical practice. More and more medical fields require PET to provide higher imaging resolution and real-time. Scanning the patient, and the corresponding expansion of image dimensions and the sharp increase in the amount of data collected pose challenges to existing reconstruction algorithms, and these requirements are also extremely demanding on the computing power and storage space of computers. Claim.
  • the reconstruction methods of PET imaging can be roughly divided into two categories: analytical methods and traditional iterative statistical methods.
  • the former category is mainly FBP (filtered back projection method), which has a fast calculation speed, but the imaging resolution is low and the artifacts are serious, and the latter one is the most commonly used ML-EM (Maximum Likelihood Maximum Expectation) algorithm.
  • ML-EM Maximum Likelihood Maximum Expectation
  • the algorithm greatly improves the resolution of the image, it still can not solve the serious artifacts very well.
  • the above two methods traditionally treat each frame as an independent individual, unable to express the temporal correlation of the PET image in the reconstruction, and the two methods cannot make the noise-filled background and target at the time of acquisition. The area is separated, so it is not possible to cope with noise interference. Summary of the invention
  • the present invention provides a PRCA PET image dynamic reconstruction method and system, which can solve the problem that the computer has low accuracy in the process of performing dynamic image reconstruction.
  • a PRCA-based PET image dynamic reconstruction method includes the following steps:
  • n groups of PET conform to the counting vector, and correct the described coincidence counting vector to construct a conforming counting matrix of PET; n is a natural number greater than 1;
  • X k+1 is the PET concentration distribution matrix after the k+1th iteration
  • Z k+1 is the PET background concentration distribution matrix after the k+1th iteration
  • G is the system matrix
  • H is the correction matrix
  • Y is the coincidence counting matrix
  • r is the correction coefficient
  • a k is the first correction coefficient matrix after the kth iteration
  • B k is the second correction coefficient matrix after the kth iteration
  • C k is the third correction coefficient matrix after the kth iteration
  • D k is the fourth correction coefficient matrix after the kth iteration
  • E k is the fifth correction coefficient matrix after the kth iteration
  • X and Z respectively For a randomized matrix corresponding to X k+1 and Z k+1 , k is a natural number;
  • G is the system matrix
  • y is the corrected coincidence count vector
  • X is the PET concentration distribution vector
  • the iterative calculation is performed according to the coincidence counting matrix by the iterative equation,
  • the PET concentration distribution matrix after the convergence is the PET concentration dynamic data.
  • the iterative convergence conditions are as follows: Where: X k is the PET concentration distribution matrix after the kth iteration, Z k is the PET background concentration distribution matrix after the kth iteration, and p is the preset convergence threshold, indicating the F norm.
  • the first correction coefficient matrix A k after the kth iteration is obtained by the following formula:
  • a k A k _ l + G(X k + Z k )-Y
  • Aw is the first correction coefficient matrix after the k-1th iteration
  • X k is the PET concentration distribution matrix after the kth iteration
  • Z k is the PET background concentration distribution matrix after the kth iteration.
  • the second correction coefficient matrix B k after the kth iteration is obtained by the following formula:
  • W k is the matrix to be decomposed after the kth iteration and X k is the PET concentration distribution matrix after the kth iteration, which is the third correction coefficient matrix after the k-1th iteration, U k , diag(i k ) and V k are the singular values of the matrix to be decomposed W k respectively
  • the time parameter correlation matrix, the singular value diagonal matrix and the mediation coefficient matrix obtained after decomposition, i k is the diagonal element in the singular value diagonal matrix diag(i k ).
  • the third correction coefficient matrix C k after the kth iteration is obtained by the following formula:
  • X k is the PET concentration distribution matrix after the kth iteration, and is the third correction coefficient matrix after the k-1th iteration.
  • - r, 0) where: Z k is the PET background concentration distribution matrix after the kth iteration, which is the fifth correction coefficient matrix after the k-1th iteration. E k E k _ l + HZ k -D k
  • Z k is the PET background concentration distribution matrix after the kth iteration, and is the fifth correction coefficient matrix after the k-1th iteration.
  • the correction coefficient r is obtained by the following formula:
  • m is the dimension that matches the count vector.
  • the dimension of the system matrix G is mxp, which characterizes the probability that the emitted photon is received by the detector, which is subjected to detector structure and detection. Efficiency, attenuation, dead time and other factors;
  • the coincidence counting matrix is a matrix consisting of n sets of coincident counting vectors with a dimension of mxn.
  • a PRCA-based PET image dynamic reconstruction system comprising a detector and the detector connected to the detector for detecting biological tissue injected with radioactive material, and dynamically collecting the n sets of PET conforming to the counting vector, n is a natural number greater than one;
  • a data receiving module configured to receive the corrected count vector and correct it, thereby constructing a conforming counting matrix of the PET;
  • a concentration estimating module configured to estimate PET concentration dynamic data by using a predetermined iterative equation group according to the matched counting matrix
  • the PET imaging module is configured to decompose the PET concentration dynamic data to obtain n consecutive frames of PET images.
  • the concentration estimation module estimates PET concentration dynamic data according to the following iterative equations:
  • a k A k _ l + G(X k +Z k )-Y
  • X k+1 is the PET concentration distribution matrix after the k+1th iteration
  • Z k+1 is the PET background concentration distribution matrix after the k+1th iteration
  • G is the system matrix
  • H is the correction matrix Y is in accordance with the counting matrix
  • r is the correction coefficient
  • a k is the first correction coefficient matrix after the kth iteration
  • B k is the second correction coefficient matrix after the kth iteration
  • C k is the kth iteration
  • D k is the fourth correction coefficient matrix after the kth iteration
  • E k is the fifth correction coefficient matrix after the kth iteration
  • X and Z are corresponding to X k+1 and Z k , respectively.
  • Aw is the first correction coefficient matrix after the k-1th iteration
  • X k is the PET concentration distribution matrix after the kth iteration
  • Z k is the PET background concentration distribution after the kth iteration Matrix
  • W k is the matrix to be decomposed after the kth iteration and
  • U k , diag(i k ) and V k are the time parameter correlation matrix obtained by the singular value decomposition of the matrix W k to be decomposed, respectively, and the singular value diagonal Matrix and median coefficient matrix
  • i k is the diagonal element in the singular value diagonal matrix diag(i k ), which is the fifth correction coefficient matrix after the k-1th iteration
  • m is the dimension of the count vector
  • k It is a natural number
  • the PET concentration distribution matrix after iteration convergence is the PET concentration dynamic data.
  • the beneficial technical effects of the present invention are: reconstructing the collected data of different frames as a whole, making full use of the temporal correlation of the PET data, so that the obtained result can more reflect that the dynamic PET can show the target.
  • the present invention uses the background and target area methods to reduce the interference of the background to the target area, and adds time and space corrections in the reconstruction, so that the reconstruction result is more accurate.
  • the contrast between the target area and the background is improved, making the reconstruction better and more medical than the traditional FBP and ML-EM algorithms.
  • 1 is a schematic structural view of an estimation system of the present invention.
  • FIG. 2 is a schematic flow chart showing the steps of the estimation method of the present invention.
  • Fig. 3(a) is a simulation image of the level 1 level using Monte Carlo simulation of the present invention.
  • Fig. 3(b) is a simulation image of the level 2 level using Monte Carlo simulation of the present invention.
  • Fig. 4(a) is a schematic diagram showing the results of the original concentration of the Levell level simulated image.
  • Fig. 4(b) is a schematic diagram showing the concentration results of ML-EM reconstruction of the Level1 level simulated image.
  • Fig. 4(c) is a schematic diagram showing the concentration results of the level 1 simulation image reconstructed by the method of the present invention.
  • Figure 5(a) is a schematic diagram showing the results of the original concentration of the Level 2 level simulated image.
  • Fig. 5(b) is a schematic diagram showing the concentration results of the ML-EM reconstruction of the Level 2 level analog image.
  • Fig. 5(c) is a schematic diagram showing the concentration results of the level 2 level simulated image reconstructed by the method of the present invention.
  • Figure 6 is a schematic view of a Phantom phantom used in the present invention.
  • Figure 7 is a schematic diagram showing the results of the original concentration of the Phantom phantom.
  • Figure 7(b) is a graphical representation of the concentration results of the Phantom phantom reconstructed using ML-EM.
  • Fig. 7(c) is a schematic diagram showing the concentration results of the Phantom phantom reconstructed by the method of the present invention.
  • Figure 8 is a comparison of pixel values in the 40th row of the Phantom phantom reconstruction results. detailed description
  • a PRCA-based PET image dynamic reconstruction system includes a detector and a computer connected to the detector; the operation execution flowchart of the system is shown in Fig. 2;
  • the detector is used for detecting the biological tissue injected with the radioactive substance, and the n-group conforming to the counting vector of the PET is dynamically collected.
  • the detector adopts the PET scanner of the model SHR74000 produced by Hamamatsu Corporation of Japan.
  • the computer is loaded with a data receiving module, a concentration estimating module and a PET imaging module; wherein: the data receiving module is configured to receive and correct the coincidence counting vector; the positron emission tomography scanner detects the radioactive signal emitted by the human body, after conforming and The acquisition system processes to form the original coincident event.
  • the compliance events recorded by the PET detector include true compliance, random compliance, and scattering compliance. Through exploration The delay window and the energy window of the detector correct the random event and the scattering event, and then perform attenuation correction to obtain the sinogram data, that is, the corrected coincidence count vector y.
  • the data receiving module constructs a PET-compliant count matrix Y consisting of these coincident count vectors y, which is a matrix of mxn, corresponding to a set of coincidence count vectors for each column in the count matrix Y.
  • the concentration estimation module is configured to estimate the PET concentration dynamic data by a predetermined iterative equation group according to the coincidence counting matrix Y.
  • y is the corrected coincidence count vector and is an m-dimensional vector
  • X is a PET concentration distribution vector and is a p-dimensional vector
  • G is a system matrix and is an mxp-dimensional matrix, which characterizes the probability that the emitted photons are received by the detector, It is affected by factors such as detector structure, detection efficiency, attenuation, and dead time.
  • a k A k _ l + G(X k +Z k )-Y
  • X k+1 is the PET concentration distribution matrix after the k+1th iteration
  • Z k+1 is the PET background concentration distribution matrix after the k+1th iteration
  • G is the system matrix
  • H is the correction matrix
  • Y In order to comply with the counting matrix
  • r is the correction coefficient
  • a k is the first correction coefficient matrix after the kth iteration
  • B k is the second correction coefficient matrix after the kth iteration
  • C k is the first after the kth iteration
  • D k is the fourth correction coefficient matrix after the kth iteration
  • E k is the fifth correction coefficient matrix after the kth iteration
  • X and Z are respectively Corresponding to the randomization matrix of X k+1 and Z k+1
  • Aw is the first correction coefficient matrix after the k-1th iteration
  • X k is the PET concentration distribution matrix after the kth iteration
  • Z k is the kth
  • the PET concentration distribution matrix after the iteration convergence is the PET concentration dynamic data;
  • the iterative convergence conditions are as follows: Where: p is the preset convergence threshold,
  • f represents the F norm; in the present embodiment, ⁇ 1( ⁇ 3 ; the maximum number of iterations is preset to 50 times.
  • the PET imaging module is used to decompose the PET concentration dynamic data to obtain n consecutive frames of PET images; since the iteratively converged PET concentration distribution matrix is a matrix of pxn, each column in the matrix corresponds to a set of PET concentration distribution vectors (ie A frame of PET image data), each column of vector elements are recombined to restore the PET image.
  • a set of PET concentration distribution vectors ie A frame of PET image data

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Algebra (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Nuclear Medicine (AREA)

Abstract

提供了一种基于PRCA的PET图像动态重建方法,包括:(1)采集、校正,得到符合计数矩阵;(2)对PET测量方程施加修正约束;(3)迭代估计PET动态浓度数据;(4)重建出各帧PET图像。本发明通过将采集到的不同帧的数据看作一个整体来进行重建,充分利用了PET数据在时间上的相关性,使所得结果更能体现出动态PET能表现出目标区域时间变化的特点;其次,本发明使用了背景和目标区域的方法,减小了背景对目标区域的干扰,加之在重建中添加了时间和空间的修正,使得重建结果的准确度更高,目标区域与背景的对比度提高,使的重建效果比传统的FBP和ML-EM算法更加优秀,更具有医学价值。

Description

一种基于 PRCA的 PET图像动态重建方法及系统
技术领域
本发明属于 PET成像技术领域, 具体涉及一种基于 PRCA (鲁棒性主元分 析法 ) 的 PET图像动态重建方法及系统。 背景技术
正电子发射断层成像 ( Positron emission tomography, PET )是一种基于核 物理学、 分子生物学的医学影像技术, 它能够从分子层面上观察细胞的新陈代 谢活动, 为早期疾病尤其是肿瘤的检测和预防提供了有效依据。 PET 本质上是 对病人体内药物的浓度分布进行成像, 被注射入病人体内的放射性同位核素标 记药物通过血液进入循环系统, 这些物质在人体内各组织器官中将形成一定的 浓度分布。 由于放射性同位核素的半衰期较短, 且极其不稳定, 将很快发生衰 变, 衰变过程中所释放的正电子与附近的自由电子发生湮灭反应, 产生一对方 向几乎相反、 能量相等, 能量大小为 511kev的伽玛光子对, 这些光子由探测器 环接收, 再经由符合采集系统对这些带有放射性药物分布信息的成对光子进行 处理生成投影数据(sinogram )。 之后, 通过相应的数学方法对投影数据进行反 演求解, 可重建出人体的放射性物质的空间浓度分布。
近几年 PET在实际医学领域的应用日趋广泛, 但与此同时, 临床上对 PET 成像的要求也随之变高, 越来越多的医学领域需要 PET能提供更高成像分辨率 和能够实时的对患者进行扫描, 而与此相应的图像维数的扩展和采集数据量的 急剧加大对现有的重建算法提出了挑战, 并且这些需求也对计算机的计算能力 和存储空间有极为苛刻的要求。
目前, PET成像的重建方法大致可分为两类: 解析法和传统迭代统计法。 前一类主要是 FBP (滤波反投影法), 计算速度快, 但成像分辨率低并且存在伪 影严重, 而后一种现在最为常用的是 ML-EM (最大似然最大期望)算法, 这种 算法虽然很大程度上提高了图像的分辨率, 但依然不能很好的解决严重的伪影 问题。 而且, 以上这两种方法传统上都是将每帧视为独立的个体, 无法在重建 中将 PET图像的时间关联性表现出来, 加之两种方法都无法将采集时的充满噪 声的背景和目标区域分离, 所以不能 4艮好的应对噪声的干扰。 发明内容
针对现有技术所存在的上述技术问题, 本发明提供了一种 PRCA的 PET图 像动态重建方法及系统, 能够解决计算机在进行动态图像重建的过程中准确率 低的问题。
一种基于 PRCA的 PET图像动态重建方法, 包括如下步骤:
( 1 )利用探测器对注入有放射性物质的生物组织进行探测, 动态采集得到
PET的 n组符合计数向量, 并对所述的符合计数向量进行校正, 进而构建 PET 的符合计数矩阵; n为大于 1的自然数;
( 2 )根据 PET成像原理, 建立 PET的测量方程;
( 3 )通过对所述的测量方程在时间和空间上施加^ ίι爹正约束得到以下迭代方 程; 根据所述的符合计数矩阵通过以下迭代方程估计出 PET浓度动态数据;
(Xk+l, Zk+l) = arg min + Ζ) - Υ - ΑΛ
Figure imgf000004_0001
+ r \\HZ - Dk + Ek 其中: Xk+1为第 k+1次迭代后的 PET浓度分布矩阵, Zk+1为第 k+1次迭代后的 PET背景浓度分布矩阵, G为系统矩阵, H为修正矩阵, Y为符合计数矩阵, r 为修正系数, Ak为第 k次迭代后的第一修正系数矩阵, Bk为第 k次迭代后的第 二修正系数矩阵, Ck为第 k次迭代后的第三修正系数矩阵, Dk为第 k次迭代后 的第四修正系数矩阵, Ek为第 k次迭代后的第五修正系数矩阵, X和 Z分别为 对应 Xk+1和 Zk+1的随机化矩阵, k为自然数;
( 4 )对所述的 PET浓度动态数据进行分解, 得到 n帧连续的 PET图像。 所述的测量方程的表达式如下:
y = Gx
其中: G为系统矩阵; y为校正后的符合计数向量; X为 PET浓度分布向量。
所述的步骤(3 ) 中, 根据符合计数矩阵通过迭代方程进行迭代计算, 则迭 代收敛后的 PET浓度分布矩阵即为 PET浓度动态数据。
所述的迭代收敛条件如下:
Figure imgf000005_0001
其中: Xk为第 k次迭代后的 PET浓度分布矩阵, Zk为第 k次迭代后的 PET背景 浓度分布矩阵, p为预设的收敛阈值, 表示 F范数。 所述的第 k次迭代后的第一修正系数矩阵 Ak通过以下算式求得:
Ak =Ak_l + G(Xk +Zk)-Y
其中: Aw为第 k-1次迭代后的第一修正系数矩阵, Xk为第 k次迭代后的 PET 浓度分布矩阵, Zk为第 k次迭代后的 PET背景浓度分布矩阵。
所述的第 k次迭代后的第二修正系数矩阵 Bk通过以下算式求得:
Bh =Uh - diag(max(ik - 1, 0)) · Vk
Wk =Uk-diag(ik)-Vk
其中: Wk为第 k次迭代后的待分解矩阵且
Figure imgf000005_0002
Xk为第 k次迭代后的 PET浓度分布矩阵, 为第 k-1次迭代后的第三修正系数矩阵, Uk、 diag(ik)和 Vk分别为待分解矩阵 Wk经奇异值分解后得到的时间参数相关矩阵、奇异值对角 矩阵和中介系数矩阵, ik为奇异值对角矩阵 diag(ik)中的对角线元素。
所述的第 k次迭代后的第三修正系数矩阵 Ck通过以下算式求得:
Ck = Ck-\ + Xk ~Bk
其中: Xk为第 k次迭代后的 PET浓度分布矩阵, 为第 k-1次迭代后的第三 修正系数矩阵。
所述的第 k次迭代后的第四修正系数矩阵 Dk通过以下算式求得: Dk = sgn(HZk + Ek_x) .
Figure imgf000005_0003
+ Ek_x | - r, 0) 其中: Zk为第 k次迭代后的 PET背景浓度分布矩阵, 为第 k-1次迭代后的第 五修正系数矩阵。 Ek =Ek_l +HZk-Dk
其中: Zk为第 k次迭代后的 PET背景浓度分布矩阵, 为第 k-1次迭代后的第 五修正系数矩阵。
所述的修正系数 r通过以下算式求得:
1
jmax(n,m)
其中: m为符合计数向量的维度。
本发明中, 若符合计数向量的维度为 m, PET浓度分布矩阵的维度为 pxn, 则系统矩阵 G的维度为 mxp, 其表征了发射光子被探测器接收的概率, 其受探 测器结构、 探测效率、 衰减、 死时间等因素的影响; 符合计数矩阵为由 n组符 合计数向量组成的维度为 mxn的矩阵。修正矩阵 H为 mxp的矩阵且为多级紧凑 型小波分解中算子, 其满 J HTH = I , I为单位矩阵。 一种基于 PRCA的 PET图像动态重建系统, 包括探测器和与探测器相连的 所述的探测器用于对注入有放射性物质的生物组织进行探测 , 动态采集得 到 PET的 n组符合计数向量, n为大于 1的自然数;
所述的计算机内加载有以下功能模块:
数据接收模块, 用于接收所述的符合计数向量并对其进行校正, 进而构建 PET的符合计数矩阵;
浓度估计模块, 用于根据所述的符合计数矩阵通过预设的迭代方程组估计 出 PET浓度动态数据;
PET成像模块, 用于对所述的 PET浓度动态数据进行分解, 得到 n帧连续 的 PET图像。
所述的浓度估计模块根据以下迭代方程组估计 PET浓度动态数据:
Figure imgf000006_0001
Ak =Ak_l + G(Xk+Zk)-Y
Bh =Uh- diag(max(ik - 1, 0)) · Vk Wk = Uk - diag(ik) - Vk
Ck = Ck-\ + Xk ~ Bk
Dk = sgn(HZk + Ek_x) . max( HZk + Ek_x ― r, 0)
Ek = Ek_l + HZk - Dt
Figure imgf000007_0001
其中: 其中: Xk+1为第 k+1次迭代后的 PET浓度分布矩阵, Zk+1为第 k+1次迭 代后的 PET背景浓度分布矩阵, G为系统矩阵, H为修正矩阵, Y为符合计数 矩阵, r为修正系数, Ak为第 k次迭代后的第一修正系数矩阵, Bk为第 k次迭 代后的第二修正系数矩阵, Ck为第 k次迭代后的第三修正系数矩阵, Dk为第 k 次迭代后的第四修正系数矩阵, Ek为第 k次迭代后的第五修正系数矩阵, X和 Z 分别为对应 Xk+1和 Zk+1的随机化矩阵, Aw为第 k- 1次迭代后的第一修正系数矩 阵, Xk为第 k次迭代后的 PET浓度分布矩阵, Zk为第 k次迭代后的 PET背景浓 度分布矩阵, Wk为第 k次迭代后的待分解矩阵且
Figure imgf000007_0002
, 为第 k-1次 迭代后的第三修正系数矩阵, Uk、 diag(ik)和 Vk分别为待分解矩阵 Wk经奇异值 分解后得到的时间参数相关矩阵、 奇异值对角矩阵和中介系数矩阵, ik为奇异值 对角矩阵 diag(ik)中的对角线元素, 为第 k-1次迭代后的第五修正系数矩阵, m为符合计数向量的维度, k为自然数; 迭代收敛后的 PET浓度分布矩阵即为 PET浓度动态数据。 本发明的有益技术效果在于: 通过将采集到的不同帧的数据看作一个整体 来进行重建, 充分利用了 PET数据在时间上的相关性, 使所得结果更能体现出 动态 PET能表现出目标区域时间变化的特点; 其次, 本发明使用了背景和目标 区域的方法, 减小了背景对目标区域的干扰, 加之在重建中添加了时间和空间 的修正, 使得重建结果的准确度更高, 目标区域与背景的对比度提高, 使的重 建效果比传统的 FBP和 ML-EM算法更加优秀, 更具有医学价值。 附图说明
图 1为本发明估计系统的结构示意图。
图 2为本发明估计方法的步骤流程示意图。
图 3(a)为本发明使用蒙特卡罗模拟在 Levell等级的模拟图像。
图 3(b)为本发明使用蒙特卡罗模拟在 Level2等级的模拟图像。
图 4(a)为 Levell等级模拟图像的原始浓度结果示意图。
图 4(b)为 Levell等级的模拟图像采用 ML-EM重建的浓度结果示意图。 图 4(c)为 Levell等级的模拟图像采用本发明方法重建的浓度结果示意图。 图 5(a)为 Level2等级模拟图像的原始浓度结果示意图。
图 5(b)为 Level2等级的模拟图像采用 ML-EM重建的浓度结果示意图。 图 5(c)为 Level2等级的模拟图像采用本发明方法重建的浓度结果示意图。 图 6为本发明使用的 Phantom体模示意图。
图 7 为 Phantom体模的原始浓度结果示意图。
图 7(b)为 Phantom体模采用 ML-EM重建的浓度结果示意图。
图 7(c)为 Phantom体模采用本发明方法重建的浓度结果示意图。
图 8为 Phantom体模重建结果中第 40行像素值的对比示意图。 具体实施方式
为了更为具体地描述本发明, 下面结合附图及具体实施方式对本发明的技 术方案进行详细说明。
如图 1所示, 一种基于 PRCA的 PET图像动态重建系统, 包括探测器和与 探测器相连的计算机; 系统的操作执行流程图如图 2所示; 其中:
探测器用于对注入有放射性物质的生物组织进行探测, 动态采集得到 PET 的 n组符合计数向量; 本实施方式中, 探测器采用日本滨松公司生产的型号为 SHR74000的 PET扫描仪。
计算机内加载有数据接收模块、 浓度估计模块和 PET成像模块; 其中: 数据接收模块用于接收符合计数向量并对其进行校正; 正电子发射断层扫 描仪探测人体内发出的放射性信号, 经过符合和采集系统处理, 形成原始符合 事件。 PET探测器记录的符合事件包括真符合、 随机符合和散射符合。 通过探 测器的延时窗口和能量窗口对随机事件和散射事件进行校正, 而后进行衰减校 正, 得到正弦图数据即校正后的符合计数向量 y。
对于 n组校正后的符合计数向量 y,数据接收模块构建由这些符合计数向量 y组成的 PET符合计数矩阵 Y, 其为 mxn的矩阵, 符合计数矩阵 Y中每一列对 应着一组符合计数向量 。
浓度估计模块用于根据符合计数矩阵 Y通过预设的迭代方程组估计出 PET 浓度动态数据。
根据 PET成像原理, 可以得到 PET的测量方程的表达式如下:
y = Gx
其中: y为校正后的符合计数向量且为 m维向量; X为 PET浓度分布向量且为 p维向量; G为系统矩阵且为 mxp维矩阵, 其表征了发射光子被探测器接收的 概率, 其受探测器结构、 探测效率、 衰减、 死时间等因素的影响。
通过对上述测量方程在时间和空间上施加爹正约束得到以下迭代方程组:
2 2
, Zk+l ) = arg min G(X + Z) - 7 - Ak + X-Bk + Ck +r HZ-Dk +Ei k
Ak =Ak_l + G(Xk+Zk)-Y
Bh =Uh- diag(max(ik - 1, 0)) · Vk
Wk =Uk-diag(ik)-Vk
Figure imgf000009_0001
Dk = sgn(HZk + E^) · max( HZk + Ek_x
Ek =Ek_l +HZk-Dk jmsLx(n,m)
其中: Xk+1为第 k+1次迭代后的 PET浓度分布矩阵, Zk+1为第 k+1次迭代后的 PET背景浓度分布矩阵, G为系统矩阵, H为修正矩阵, Y为符合计数矩阵, r 为修正系数, Ak为第 k次迭代后的第一修正系数矩阵, Bk为第 k次迭代后的第 二修正系数矩阵, Ck为第 k次迭代后的第三修正系数矩阵, Dk为第 k次迭代后 的第四修正系数矩阵, Ek为第 k次迭代后的第五修正系数矩阵, X和 Z分别为 对应 Xk+1和 Zk+1的随机化矩阵, Aw为第 k-1 次迭代后的第一修正系数矩阵, Xk为第 k次迭代后的 PET浓度分布矩阵, Zk为第 k次迭代后的 PET背景浓度分 布矩阵, Wk为第 k次迭代后的待分解矩阵且 Wk=Xk+Cw , 为第 k- 1次迭代 后的第三修正系数矩阵, Uk、 diag(ik)和 Vk分别为待分解矩阵 Wk经奇异值分解 后得到的时间参数相关矩阵、 奇异值对角矩阵和中介系数矩阵, ik为奇异值对角 矩阵 diag(ik)中的对角线元素, 为第 k-1次迭代后的第五修正系数矩阵。
本实施方式中, PET浓度分布矩阵 X的维度为 ρχη, 该矩阵中每一列对应 着一组 PET浓度分布向量 X; X。、 Z。、 A。、 B。、 C。、 D。和 E。均为初始化给定的 零矩阵, 修正矩阵 H为 mxp 的矩阵且为多级紧凑型小波分解中算子, 其满足 ΗΤΗ = Ι , I为单位矩阵。
根据符合计数矩阵 Y通过上述迭代方程组进行迭代计算, 则迭代收敛后的 PET浓度分布矩阵即为 PET浓度动态数据; 迭代收敛条件如下:
Figure imgf000010_0001
其中: p为预设的收敛阈值, | |f表示 F范数; 本实施方式中 ρ=1(Τ3; 最大迭代 次数预设为 50次。
PET成像模块用于对 PET浓度动态数据进行分解,得到 n帧连续的 PET图 像; 由于迭代收敛后的 PET浓度分布矩阵为 pxn的矩阵, 该矩阵中每一列即对 应一组 PET浓度分布向量(即一帧 PET图像数据), 将每一列向量元素进行重 组即可还原得到 PET图像。 以下我们对本实施方式进行了两组验证实验, 第一组实验采用的是蒙特卡 罗模拟, 第二组采用的是六个圆柱体的 phantom体模进行模拟。
在第一组试验中, 我们采用的是 180度内 48个采样角, 所有采样结果被分 为 20帧, 我们生成了两组不同噪声和计算率等级的模拟图像, 如图 3所示; 其 中 Levell对应的是 0.53%的随机噪声和 0.03%的散射噪声, 其每帧计数率约为 7xl06; Level2对应的是 1.28%的随机噪声和 22.9%的散射噪声, 其每帧计数率 约为 2.5χ105。 我们对比了传统 ML-EM和本实施方式的重建结果, 如图 4和图 5所示, 我们主要对比的是第 8帧的图像结果。 为了能更好地量化结果我们还 别求出了 ML-EM和本实施方式重建结果的偏差和方差, 结果如表 1所示: 表 1
Figure imgf000011_0001
在第二组中我们对实际的 phantom体模数据进行了重建, 我们采用的圆柱 型体模, 整个体模的大小为直径 200mm, 深度 290mm, 如图 6所示, 其中包含 有六个不同直径的圆形区域, 其直径分别为: 37mm, 28mm, 22mm, 17mm, 13mm, 10mm。 在其中我们注入放射性浓度为 107.92Bq/ml的 F-18示踪剂, 采 样时间为 120分钟, 其对比结果如图 7所示; 同理, 为了能更好的量化结果, 我们选取图中第 40行的像素值进行比对, 其结果如图 8所示。
通过以上的实验结果我们可以看出, 本实施方式的重建结果无论是在实际 的图像还是偏差和方差这两个方面的表现都要优于 ML-EM的重建结果,由此我 们可以看出 RPCA算法有效地提高了 PET动态重建的准确率并且大幅提高了图 像目标区域和背景的对比度。

Claims

权 利 要 求 书
1. 一种基于 PRCA的 PET图像动态重建方法, 包括如下步骤:
( 1 )利用探测器对注入有放射性物质的生物组织进行探测, 动态采集得到
PET的 n组符合计数向量, 并对所述的符合计数向量进行校正, 进而构建 PET 的符合计数矩阵; n为大于 1的自然数;
( 2 )根据 PET成像原理, 建立 PET的测量方程;
( 3 )通过对所述的测量方程在时间和空间上施加^ ίι爹正约束得到以下迭代方 程; 根据所述的符合计数矩阵通过以下迭代方程估计出 PET浓度动态数据; d , ) = arg min ||G( + Z) - 7 - A, + X - Bk + Ck + r HZ - Dk + Ek χ,ζ
其中: Xk+1为第 k+1次迭代后的 PET浓度分布矩阵, Zk+1为第 k+1次迭代后的 PET背景浓度分布矩阵, G为系统矩阵, H为修正矩阵, Y为符合计数矩阵, r 为修正系数, Ak为第 k次迭代后的第一修正系数矩阵, Bk为第 k次迭代后的第 二修正系数矩阵, Ck为第 k次迭代后的第三修正系数矩阵, Dk为第 k次迭代后 的第四修正系数矩阵, Ek为第 k次迭代后的第五修正系数矩阵, X和 Z分别为 对应 Xk+1和 Zk+1的随机化矩阵, k为自然数;
( 4 )对所述的 PET浓度动态数据进行分解, 得到 n帧连续的 PET图像。
2. 根据权利要求 1所述的 PET图像动态重建方法, 其特征在于: 所述的测 量方程的表达式如下:
y = Gx
其中: G为系统矩阵; y为校正后的符合计数向量; X为 PET浓度分布向量。
3. 根据权利要求 1所述的 PET图像动态重建方法, 其特征在于: 所述的步 骤(3 )中,根据符合计数矩阵通过迭代方程进行迭代计算,则迭代收敛后的 PET 浓度分布矩阵即为 PET浓度动态数据; 迭代收敛条件如下:
Y - G(Xk + Zk)
P
Y F 其中: Xk为第 k次迭代后的 PET浓度分布矩阵, Zk为第 k次迭代后的 PET背景 浓度分布矩阵, p为预设的收敛阈值, 表示 F范数。
4. 根据权利要求 1所述的 PET图像动态重建方法, 其特征在于: 所述的第 k次迭代后的第一修正系数矩阵 Ak通过以下算式求得:
Ak =Ak_l + G(Xk +Zk)-Y
其中: Aw为第 k-1次迭代后的第一修正系数矩阵, Xk为第 k次迭代后的 PET 浓度分布矩阵, Zk为第 k次迭代后的 PET背景浓度分布矩阵。
5. 根据权利要求 1所述的 PET图像动态重建方法, 其特征在于: 所述的第 k次迭代后的第三修正系数矩阵 Ck通过以下算式求得:
Ck = Ck-\ + Xk ~Bk
其中: Xk为第 k次迭代后的 PET浓度分布矩阵, 为第 k-1次迭代后的第三 修正系数矩阵。
6. 根据权利要求 1或 5所述的 PET图像动态重建方法, 其特征在于: 所述 的第 k次迭代后的第二修正系数矩阵 Bk通过以下算式求得:
Bh =Uh- diag(max(ik - 1, 0)) · Vk
Wk =Uk-diag(ik)-Vk
其中: Wk为第 k次迭代后的待分解矩阵且
Figure imgf000013_0001
Xk为第 k次迭代后的 PET浓度分布矩阵, 为第 k-1次迭代后的第三修正系数矩阵, Uk、 diag(ik)和 Vk分别为待分解矩阵 Wk经奇异值分解后得到的时间参数相关矩阵、奇异值对角 矩阵和中介系数矩阵, ik为奇异值对角矩阵 diag(ik)中的对角线元素。
7. 根据权利要求 1所述的 PET图像动态重建方法, 其特征在于: 所述的第 k次迭代后的第五修正系数矩阵 Ek通过以下算式求得:
Ek =Ek_l +HZk-Dk
其中: Zk为第 k次迭代后的 PET背景浓度分布矩阵, 为第 k-1次迭代后的第 五修正系数矩阵。
8. 根据权利要求 1或 7所述的 PET图像动态重建方法, 其特征在于: 所述 的第 k次迭代后的第四修正系数矩阵 Dk通过以下算式求得:
Dk = sgn(HZk + Ek_x) .
Figure imgf000013_0002
+ Ek_x | - r, 0) 其中: Zk为第 k次迭代后的 PET背景浓度分布矩阵, 为第 k-1次迭代后的第 五修正系数矩阵。
9. 一种基于 PRCA的 PET图像动态重建系统, 包括探测器和与探测器相连 的计算机; 所述的探测器用于对注入有放射性物质的生物组织进行探测, 动态 采集得到 PET的 n组符合计数向量, n为大于 1的自然数; 其特征在于:
所述的计算机内加载有以下功能模块:
数据接收模块, 用于接收所述的符合计数向量并对其进行校正, 进而构建 PET的符合计数矩阵;
浓度估计模块, 用于根据所述的符合计数矩阵通过预设的迭代方程组估计 出 PET浓度动态数据;
PET成像模块, 用于对所述的 PET浓度动态数据进行分解, 得到 n帧连续 的 PET图像。
10. 根据权利要求 9所述的 PET图像动态重建系统, 其特征在于: 所述的 浓度估计模块根据以下迭代方程组估计 PET浓度动态数据:
Figure imgf000014_0001
Ak =Ak_l + G(Xk+Zk)-Y
Bh =Uh- diag(max(ik - 1, 0)) · Vk
Wk =Uk-diag(ik)-Vk
Figure imgf000014_0002
Dk = sgn(HZk + E^) · max( HZk + Ek_x
Ek =Ek_l +HZk-Dk jmsLx(n,m)
其中: 其中: Xk+1为第 k+1次迭代后的 PET浓度分布矩阵, Zk+1为第 k+1次迭 代后的 PET背景浓度分布矩阵, G为系统矩阵, H为修正矩阵, Y为符合计数 矩阵, r为修正系数, Ak为第 k次迭代后的第一修正系数矩阵, Bk为第 k次迭 代后的第二修正系数矩阵, Ck为第 k次迭代后的第三修正系数矩阵, Dk为第 k 次迭代后的第四修正系数矩阵, Ek为第 k次迭代后的第五修正系数矩阵, X和 Z 分别为对应 Xk+1和 Zk+1的随机化矩阵, Aw为第 k- 1次迭代后的第一修正系数矩 阵, Xk为第 k次迭代后的 PET浓度分布矩阵, Zk为第 k次迭代后的 PET背景浓 度分布矩阵, Wk为第 k次迭代后的待分解矩阵且 Wk=Xk+Cw, 为第 k-1次 迭代后的第三修正系数矩阵, Uk、 diag(ik)和 Vk分别为待分解矩阵 Wk经奇异值 分解后得到的时间参数相关矩阵、 奇异值对角矩阵和中介系数矩阵, ik为奇异值 对角矩阵 diag(ik)中的对角线元素, 为第 k-1次迭代后的第五修正系数矩阵, m为符合计数向量的维度, k为自然数; 迭代收敛后的 PET浓度分布矩阵即为 PET浓度动态数据。
PCT/CN2013/075593 2013-04-23 2013-05-14 一种基于rpca的pet图像动态重建方法及系统 WO2014172927A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US14/786,534 US9576379B2 (en) 2013-04-23 2013-05-14 PRCA-based method and system for dynamically re-establishing PET image

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201310144543.6A CN103279964B (zh) 2013-04-23 2013-04-23 一种基于prca的pet图像动态重建方法及系统
CN201310144543.6 2013-04-23

Publications (1)

Publication Number Publication Date
WO2014172927A1 true WO2014172927A1 (zh) 2014-10-30

Family

ID=49062471

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2013/075593 WO2014172927A1 (zh) 2013-04-23 2013-05-14 一种基于rpca的pet图像动态重建方法及系统

Country Status (3)

Country Link
US (1) US9576379B2 (zh)
CN (1) CN103279964B (zh)
WO (1) WO2014172927A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190008468A1 (en) * 2017-01-16 2019-01-10 Zhejiang University A method for mixed tracers dynamic pet concentration image reconstruction based on stacked autoencoder

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103810731A (zh) * 2014-01-20 2014-05-21 浙江大学 一种基于tv范数的pet图像重建方法
CN106251313B (zh) 2016-08-15 2020-06-26 上海联影医疗科技有限公司 医学成像方法及系统
CN106778024B (zh) * 2017-01-04 2020-02-14 东软医疗系统股份有限公司 一种图像显示方法和装置
CN107274459B (zh) * 2017-05-29 2020-06-09 明峰医疗系统股份有限公司 一种用于加快锥形束ct图像迭代重建的预条件方法
CN107464270B (zh) * 2017-07-17 2020-08-11 东软医疗系统股份有限公司 一种图像重建方法和装置
CN110533734B (zh) * 2019-04-25 2023-06-13 南方医科大学 基于传统单能ct的多能谱分段稀疏扫描迭代重建方法
CN110599562B (zh) * 2019-09-02 2023-01-10 四川轻化工大学 基于多能量系统响应矩阵的放射源定位重建方法
US11528200B2 (en) * 2020-09-15 2022-12-13 Cisco Technology, Inc. Proactive insights for IoT using machine learning

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030014132A1 (en) * 2000-02-07 2003-01-16 Hiroyuki Ohba Positron emission tomograph
CN102184559A (zh) * 2011-05-17 2011-09-14 刘华锋 一种基于粒子滤波的静态pet图像重建方法
CN102938154A (zh) * 2012-11-13 2013-02-20 浙江大学 一种基于粒子滤波的动态pet图像重建方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6987270B2 (en) * 2003-05-07 2006-01-17 General Electric Company Method to account for event losses due to positron range in positron emission tomography and assay of positron-emitting isotopes
US20050023473A1 (en) * 2003-08-01 2005-02-03 Burr Kent Charles System and method for reducing optical crosstalk in multi-anode photomultiplier tube
US7649176B2 (en) * 2003-11-14 2010-01-19 Siemens Medical Solutions Usa, Inc. Method for improving clinical data quality in positron emission tomography
US7756310B2 (en) * 2006-09-14 2010-07-13 General Electric Company System and method for segmentation
EP3709022A1 (en) * 2011-07-08 2020-09-16 Sloan-Kettering Institute for Cancer Research Uses of labeled hsp90 inhibitors
WO2014047446A1 (en) * 2012-09-21 2014-03-27 The General Hospital Corporation System and method for single-scan rest-stress cardiac pet
US9256967B2 (en) * 2012-11-02 2016-02-09 General Electric Company Systems and methods for partial volume correction in PET penalized-likelihood image reconstruction

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030014132A1 (en) * 2000-02-07 2003-01-16 Hiroyuki Ohba Positron emission tomograph
CN102184559A (zh) * 2011-05-17 2011-09-14 刘华锋 一种基于粒子滤波的静态pet图像重建方法
CN102938154A (zh) * 2012-11-13 2013-02-20 浙江大学 一种基于粒子滤波的动态pet图像重建方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190008468A1 (en) * 2017-01-16 2019-01-10 Zhejiang University A method for mixed tracers dynamic pet concentration image reconstruction based on stacked autoencoder
US10765382B2 (en) * 2017-01-16 2020-09-08 Zhejiang University Method for mixed tracers dynamic PET concentration image reconstruction based on stacked autoencoder

Also Published As

Publication number Publication date
US9576379B2 (en) 2017-02-21
CN103279964A (zh) 2013-09-04
US20160078646A1 (en) 2016-03-17
CN103279964B (zh) 2015-10-28

Similar Documents

Publication Publication Date Title
WO2014172927A1 (zh) 一种基于rpca的pet图像动态重建方法及系统
CN111627082B (zh) 基于滤波反投影算法和神经网络的pet图像重建方法
US7127095B2 (en) Factor analysis in medical imaging
CN104657950B (zh) 一种基于Poisson TV的动态PET图像重建方法
US9507033B2 (en) Method and apparatus for compensating for scattering of emission gamma photons for PET imaging
WO2019095411A1 (zh) 一种闪烁脉冲事件的符合处理方法
CN113160347B (zh) 一种基于注意力机制的低剂量双示踪剂pet重建方法
CN106204674A (zh) 基于结构字典和动力学参数字典联合稀疏约束的动态pet图像重建方法
CA3003664A1 (en) Normalization crystal efficiencies estimation for continuous motion bed acquisition
Ruan et al. Separation of a mixture of simultaneous dual-tracer PET signals: a data-driven approach
CN109741411B (zh) 基于梯度域的低剂量pet图像重建方法、装置、设备及介质
Qi et al. Limited‐angle effect compensation for respiratory binned cardiac SPECT
JP2018044957A (ja) 医用画像処理装置、医用画像診断装置及び医用画像処理方法
Zhang et al. Deep generalized learning model for PET image reconstruction
CN115423892A (zh) 一种基于最大期望网络的无衰减校正pet重建方法
CN114943782A (zh) 一种正电子发射断层成像方法、系统和装置
CN105279777B (zh) 基于改进序贯滤波的静态pet图像重建方法
Arunprasath et al. Reconstruction of PET brain image using conjugate gradient algorithm
US11663758B2 (en) Systems and methods for motion estimation in PET imaging using AI image reconstructions
Reutter et al. Fully 4-D dynamic cardiac SPECT image reconstruction using spatiotemporal B-spline voxelization
Xiang et al. SPECT/CT scatter estimation using a deep convolutional neural network: implementation in Y-90 imaging
Ralli et al. 4D-PET reconstruction of dynamic non-small cell lung cancer [18-F]-FMISO-PET data using adaptive-knot cubic B-splines
CN107221012B (zh) 基于改进了适用范围的卡尔曼滤波的静态pet图像重建方法
WO2023241722A1 (en) Methods and systems for image reconstruction
CN105551067B (zh) 基于类序贯h∞滤波的静态pet图像重建方法

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: 13882703

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 14786534

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 13882703

Country of ref document: EP

Kind code of ref document: A1