WO2019153443A1 - Magnetic resonance diffusion weighted imaging self-adaptive correction method - Google Patents

Magnetic resonance diffusion weighted imaging self-adaptive correction method Download PDF

Info

Publication number
WO2019153443A1
WO2019153443A1 PCT/CN2018/080113 CN2018080113W WO2019153443A1 WO 2019153443 A1 WO2019153443 A1 WO 2019153443A1 CN 2018080113 W CN2018080113 W CN 2018080113W WO 2019153443 A1 WO2019153443 A1 WO 2019153443A1
Authority
WO
WIPO (PCT)
Prior art keywords
diffusion weighted
magnetic resonance
correction method
adaptive correction
vector
Prior art date
Application number
PCT/CN2018/080113
Other languages
French (fr)
Chinese (zh)
Inventor
罗海
王世杰
朱高杰
周翔
陈梅泞
王超
刘霞
吴子岳
Original Assignee
奥泰医疗系统有限责任公司
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 奥泰医疗系统有限责任公司 filed Critical 奥泰医疗系统有限责任公司
Publication of WO2019153443A1 publication Critical patent/WO2019153443A1/en

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/003—Reconstruction from projections, e.g. tomography
    • G06T11/005—Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06T2207/10088—Magnetic resonance imaging [MRI]

Definitions

  • the present invention relates to the field of magnetic resonance imaging, and more particularly to an adaptive correction method for magnetic resonance diffusion weighted imaging.
  • Diffusion Weighted Imaging is an imaging method that non-invasively reflects the irregular thermal motion of living water molecules at the molecular level. Imaging depends mainly on the motion of water molecules rather than the proton density of tissue, T1 or T2 relaxation time. Diffusion-weighted imaging is suitable for detecting the micro-dynamic and micro-structural changes of biological tissues at the level of living cells, and plays an important role in the benign and malignant identification, therapeutic evaluation and prediction of tumors.
  • the applied diffusion gradient is extremely sensitive to motion.
  • Exercise mainly includes the following four aspects: (1) diffuse movement of water molecules; (2) unconscious physiological movements of patients, such as respiratory movements, gastrointestinal motility, blood flow, etc.; (3) conscious or unconscious overall movement of patients; (4) dispersion System vibration caused by gradients.
  • the diffusion of water molecules under the influence of the diffusion gradient will cause a phase difference to reduce the tissue signal with a large diffusion coefficient, which is the principle of diffusion-weighted imaging.
  • the latter three movements will cause motion artifacts. Even sub-pixel motions will produce great phase differences, causing signal loss and serious artifacts.
  • the applied diffusion gradient is very large, which can cause the system to vibrate severely, which may cause radio frequency interference caused by loose coil interface or static electricity accumulation/release, and form strip-shaped artifacts in the image, usually called RF ignition. Artifacts.
  • the invention aims to provide an adaptive correction method for magnetic resonance diffusion weighted imaging. Based on the multiple acquisition averaging technique, the principal component analysis method is used to adaptively detect and correct motion artifacts and radio frequency ignition from redundant data. Shadows, etc., to better improve image quality without the need to add hardware devices.
  • An adaptive correction method for magnetic resonance diffusion weighted imaging comprising the following steps:
  • Step 1 repeatedly collecting the diffusion weighted image N times, N is a natural number, N ⁇ 3;
  • Step 2 construct a correlation matrix point by point based on the original image or the compressed image
  • Step 3 principal component analysis; obtaining a feature vector corresponding to a maximum eigenvalue of each correlation matrix;
  • Step 4 calculating a weight according to the feature vector
  • Step 5 Perform weighted synthesis on the original image collected in step 1 according to the weight obtained in step 4, to obtain a corrected diffusion weighted image.
  • step 2 all the collected original images are compressed by using an interpolation algorithm.
  • the benefits are that the first can reduce the amount of computation, and the second can increase the signal-to-noise ratio of the input data of the subsequent algorithm.
  • step 2 includes the following steps:
  • Step 2.1 for any pixel point (x, y) in the image acquired in the nth time, take the neighboring K points to form a neighborhood vector Xn;
  • Step 2.2 For N times of repeatedly acquired images, each pixel point corresponds to N neighborhood vectors, and the correlation between the nth vector Xn and the mth vector Xm is calculated according to formula (1);
  • x i is the i-th element in the vector Xn
  • y i is the i-th element in the vector Xm. Is the mean of the vector Xn, Is the mean of the vector Xm.
  • Step 2.3 any pixel point (x, y) corresponds to an N*N correlation matrix R(x, y);
  • r 1,1 ... r 1,N are the correlation coefficients between the two vectors calculated according to formula (1).
  • step 3 includes the following steps
  • Step 3.1 calculating the eigenvalues of the matrix R(x, y) to find the largest eigenvalue
  • step 3.2 the feature vector ⁇ corresponding to the maximum eigenvalue of the matrix R(x, y) is calculated.
  • the correlation matrix is first subjected to smoothing filtering processing.
  • the smoothing filtering process comprises the following steps;
  • Step a taking the i-th correlation coefficient from the correlation matrix R(x, y) corresponding to each pixel point (x, y), to form a matrix Ri of the same size as the image matrix;
  • Step b performing two-dimensional low-pass filtering on the matrix Ri;
  • Step c replacing the filtered result with the corresponding element in R(x, y);
  • step d repeat a-c until all elements in R(x, y) have been processed.
  • step 4 the weight is calculated by formula (2);
  • ⁇ n is the nth element of the feature vector ⁇
  • ⁇ min is the smallest element of the feature vector ⁇
  • ⁇ max is the largest element of the feature vector ⁇
  • a and p are parameter control factors.
  • a 0.2
  • step 5 the original image is weighted and synthesized by the formula (3);
  • Equation (3) M n is the n-shot acquisition diffusion weighted original image, w n is the weight.
  • step 1 the diffusion weighted image is repeatedly acquired N times with the same scanning parameter.
  • the invention adopts a principal component analysis method on the basis of multiple acquisition averaging techniques, adaptively detects and corrects data from redundant data, suppresses motion artifacts, RF ignition artifacts, etc., and improves image quality; Hardware device, and image quality is better than multiple acquisition direct averaging techniques.
  • Figure 1 is a flow chart of the present invention
  • Figure 2 is the same scan parameters, 4 abdominal diffusion weighted images obtained in 4 acquisitions;
  • FIG. 3 is a diffusion-weighted image obtained by directly averaging the data collected four times in FIG. 2;
  • Figure 4 is a diffusion weighted image of the four acquisition data of Figure 2 corrected according to the method of the present invention
  • Figure 5 is a direct average technique synthesis, a diffusion-weighted image of the abdomen containing radio frequency ignition artifacts
  • Fig. 6 is a graph showing the abdominal diffusion weighted image corrected by the method of the present invention corresponding to the data in Fig. 5.
  • Step 1 the same scanning parameters, repeated acquisition of the diffusion weighted image N times, N is a natural number, N ⁇ 3;
  • Step 2 construct a correlation matrix point by point based on the original image or the compressed image: specifically including the following steps;
  • Step 2.1 for any pixel point (x, y) in the image acquired in the nth time, take the neighboring K points to form a neighborhood vector Xn;
  • Step 2.2 For N times of repeatedly acquired images, each pixel point corresponds to N neighborhood vectors, and the correlation between the nth vector Xn and the mth vector Xm is calculated according to formula (1);
  • x i is the i-th element in the vector Xn
  • y i is the i-th element in the vector Xm. Is the mean of the vector Xn, Is the mean of the vector Xm.
  • Step 2.3 any pixel point (x, y) corresponds to an N*N correlation matrix R(x, y);
  • r 1,1 ... r 1,N are the correlation coefficients between the two vectors calculated according to formula (1).
  • Step 3 Principal component analysis: obtaining a feature vector corresponding to a maximum eigenvalue of each correlation matrix; specifically comprising the following steps;
  • Step 3.1 calculating the eigenvalues of the matrix R(x, y) to find the largest eigenvalue
  • step 3.2 the feature vector ⁇ corresponding to the maximum eigenvalue of the matrix R(x, y) is calculated.
  • Step 4 calculating a weight according to formula (2);
  • ⁇ n is the nth element of the feature vector ⁇
  • ⁇ min is the smallest element of the feature vector ⁇
  • ⁇ max is the largest element of the feature vector ⁇
  • a and p are parameter control factors.
  • Step 5 Perform weighted synthesis on the original image collected in step 1 according to the weight obtained in step 4, to obtain a corrected diffusion weighted image. Specifically, the original image is weighted and synthesized by the formula (3);
  • Equation (3) M n is the n-shot acquisition diffusion weighted original image, w n is the weight.
  • Embodiment 1 The difference between this embodiment and Embodiment 1 is that all the collected original images are compressed by the interpolation algorithm before step 2.
  • the benefits are that the first can reduce the amount of computation, and the second can increase the signal-to-noise ratio of the input data of the subsequent algorithm.
  • the correlation matrix is subjected to smoothing processing before step 3.
  • the smoothing filtering process comprises the following steps;
  • Step a taking the i-th correlation coefficient from the correlation matrix R(x, y) corresponding to each pixel point (x, y), to form a matrix Ri of the same size as the image matrix;
  • Step b performing two-dimensional low-pass filtering on the matrix Ri;
  • Step c replacing the filtered result with the corresponding element in R(x, y);
  • step d repeat a-c until all elements in R(x, y) have been processed.
  • the diffusion-weighted image synthesized by direct averaging has a limited degree of artifact suppression and poor picture quality; as shown in FIG. 4, the image corrected by the method of the present invention is more accurate. As shown in Figures 5 and 6, the RF ignition artifacts in the image corrected by the method of the present invention are significantly reduced.
  • the invention adaptively calculates the weights of each scan data based on the principal component analysis method, performs weighted synthesis according to the obtained weights, suppresses motion artifacts, RF ignition artifacts, and improves image quality without requiring Add hardware devices.

Landscapes

  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Engineering & Computer Science (AREA)
  • Pathology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Radiology & Medical Imaging (AREA)
  • Theoretical Computer Science (AREA)
  • Biophysics (AREA)
  • General Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Medical Informatics (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

A magnetic resonance diffusion weighted imaging self-adaptive correction method, comprising the following step: step 1, repeatedly acquiring diffusion weighted images N times according to same scanning parameters, N being greater than or equal to 3; step 2, constructing correlation matrices point by point on the basis of original images or compressed images; step 3, performing principal component analysis after performing smooth filtering processing on the correlation matrices, so as to obtain a feature vector corresponding to the maximum feature value of each correlation matrix; step 4, calculating a weight according to the feature vector; and step 5, performing weighted synthesis on the original images according to the weight, so as to obtain the corrected diffusion weighted images. According to the method, on the basis of a technique of averaging multiple acquisitions, a principal component analysis method is used, data is self-adaptively detected and corrected from redundant data, motion artifact, radio frequency striking artifact, etc. are restrained, and the image quality is improved; no hardware device needs to be added, and the image quality is better than that on the basis of a technique of directly averaging multiple acquisitions.

Description

磁共振弥散加权成像自适应修正方法Magnetic resonance diffusion weighted imaging adaptive correction method 技术领域Technical field
本发明涉及磁共振成像领域，尤其涉及磁共振弥散加权成像自适应修正方法。The present invention relates to the field of magnetic resonance imaging, and more particularly to an adaptive correction method for magnetic resonance diffusion weighted imaging.
背景技术Background technique
弥散加权成像(Diffusion Weighted Imaging,DWI)是一种在分子水平上无创地反映活体水分子的无规则热运动状况的成像方法，成像主要依赖于水分子的运动而非组织的质子密度、T1或T2弛豫时间。弥散加权成像适用于活体细胞水平探测生物组织的微动态和微结构变化，在肿瘤的良恶性鉴别、疗效评估和预测起着举足轻重的作用。Diffusion Weighted Imaging (DWI) is an imaging method that non-invasively reflects the irregular thermal motion of living water molecules at the molecular level. Imaging depends mainly on the motion of water molecules rather than the proton density of tissue, T1 or T2 relaxation time. Diffusion-weighted imaging is suitable for detecting the micro-dynamic and micro-structural changes of biological tissues at the level of living cells, and plays an important role in the benign and malignant identification, therapeutic evaluation and prediction of tumors.
在弥散加权成像中，施加的弥散梯度对运动极为敏感。运动主要包括以下四方面：(1)水分子弥散运动；(2)患者无意识的生理运动，如呼吸运动，肠胃蠕动，血液流动等；(3)患者有意识或无意识的整体运动；(4)弥散梯度导致的系统振动。水分子弥散运动在弥散梯度的作用下会产生相位差使得弥散系数较大的组织信号降低，这是弥散加权成像的原理。而后三种运动都会导致运动伪影，即使是亚像素级别的运动，也会产生极大的相位差，使得信号丢失，形成严重的伪影。In diffusion-weighted imaging, the applied diffusion gradient is extremely sensitive to motion. Exercise mainly includes the following four aspects: (1) diffuse movement of water molecules; (2) unconscious physiological movements of patients, such as respiratory movements, gastrointestinal motility, blood flow, etc.; (3) conscious or unconscious overall movement of patients; (4) dispersion System vibration caused by gradients. The diffusion of water molecules under the influence of the diffusion gradient will cause a phase difference to reduce the tissue signal with a large diffusion coefficient, which is the principle of diffusion-weighted imaging. The latter three movements will cause motion artifacts. Even sub-pixel motions will produce great phase differences, causing signal loss and serious artifacts.
在弥散加权成像中，施加的弥散梯度非常大，会导致系统振动剧烈，进而可能出现线圈接口松动或静电积累/释放等导致射频干扰， 在图像中形成条状伪影，通常称为射频打火伪影。In diffusion-weighted imaging, the applied diffusion gradient is very large, which can cause the system to vibrate severely, which may cause radio frequency interference caused by loose coil interface or static electricity accumulation/release, and form strip-shaped artifacts in the image, usually called RF ignition. Artifacts.
上述两类伪影，在弥散加权成像中非常常见，除了在弥散加权合成图像上出现伪影，还会影响基于弥散加权成像的后续处理结果，例如会导致ADC值误差，弥散张量成像误差等，影响医生诊断。为了改善上述伪影，一方面可以通过运动检测和校正技术、射频打火检测和校正技术，减少伪影，但这种方法需要增加专用的硬件检测装置或者算法复杂、可靠性差；另一方面，通常采用多次采集平均技术，降低伪影的影响，但这种方法通过直接平均，伪影抑制的程度有限。The above two types of artifacts are very common in diffusion-weighted imaging, except for artifacts on diffusion-weighted composite images, which also affect subsequent processing results based on diffusion-weighted imaging, such as ADC value errors, diffusion tensor imaging errors, etc. , affecting the doctor's diagnosis. In order to improve the above artifacts, on the one hand, motion detection and correction techniques, radio frequency ignition detection and correction techniques can be used to reduce artifacts, but this method requires the addition of dedicated hardware detection devices or complicated algorithms and poor reliability; Multiple acquisition averaging techniques are often used to reduce the effects of artifacts, but this approach is limited by direct averaging.
发明内容Summary of the invention
本发明旨在提供磁共振弥散加权成像自适应修正方法，在多次采集平均技术的基础上，采用主成分分析方法，从冗余数据中自适应地检测并校正运动伪影、射频打火伪影等，从而更好地改善图像质量，且不需要增加硬件装置。The invention aims to provide an adaptive correction method for magnetic resonance diffusion weighted imaging. Based on the multiple acquisition averaging technique, the principal component analysis method is used to adaptively detect and correct motion artifacts and radio frequency ignition from redundant data. Shadows, etc., to better improve image quality without the need to add hardware devices.
为达到上述目的，本发明采用的技术方案如下：In order to achieve the above object, the technical solution adopted by the present invention is as follows:
磁共振弥散加权成像自适应修正方法，包括以下步骤：An adaptive correction method for magnetic resonance diffusion weighted imaging, comprising the following steps:
步骤1，重复采集弥散加权图像N次，N为自然数，N≥3；Step 1, repeatedly collecting the diffusion weighted image N times, N is a natural number, N ≥ 3;
步骤2，基于原始图像或者压缩后的图像逐点构造相关性矩阵；Step 2: construct a correlation matrix point by point based on the original image or the compressed image;
步骤3，主成分分析；获得每个相关性矩阵的最大特征值对应的特征向量；Step 3: principal component analysis; obtaining a feature vector corresponding to a maximum eigenvalue of each correlation matrix;
步骤4，根据特征向量计算权重；Step 4, calculating a weight according to the feature vector;
步骤5，根据步骤4获得的权重对步骤1中采集到的原始图像进 行加权合成，获得修正后的弥散加权图像。Step 5: Perform weighted synthesis on the original image collected in step 1 according to the weight obtained in step 4, to obtain a corrected diffusion weighted image.
进一步的，步骤2之前利用插值算法对所有采集的原始图像进行压缩。其好处有，第一可以减少运算量，第二可以增加后续算法输入数据的信噪比。Further, before the step 2, all the collected original images are compressed by using an interpolation algorithm. The benefits are that the first can reduce the amount of computation, and the second can increase the signal-to-noise ratio of the input data of the subsequent algorithm.
其中，步骤2包括以下步骤：Wherein step 2 includes the following steps:
步骤2.1，对于第n次采集的图像中任意一个像素点(x,y)，取周围相邻的K个点，构成一个邻域向量Xn；Step 2.1, for any pixel point (x, y) in the image acquired in the nth time, take the neighboring K points to form a neighborhood vector Xn;
步骤2.2，对于N次重复采集的图像，每个像素点对应了N个邻域向量，按公式(1)计算其中第n个向量Xn和第m个向量Xm之间的相关性；Step 2.2: For N times of repeatedly acquired images, each pixel point corresponds to N neighborhood vectors, and the correlation between the nth vector Xn and the mth vector Xm is calculated according to formula (1);
Figure PCTCN2018080113-appb-000001
Figure PCTCN2018080113-appb-000001
公式(1)中，x i为向量Xn中的第i个元素，y i为向量Xm中的第i个元素，
Figure PCTCN2018080113-appb-000002
为向量Xn的均值，
Figure PCTCN2018080113-appb-000003
为向量Xm的均值。 In the formula (1), x i is the i-th element in the vector Xn, and y i is the i-th element in the vector Xm.
Figure PCTCN2018080113-appb-000002
Is the mean of the vector Xn,
Figure PCTCN2018080113-appb-000003
Is the mean of the vector Xm.
步骤2.3，任一像素点(x,y)对应一个N*N的相关性矩阵R(x,y)；Step 2.3, any pixel point (x, y) corresponds to an N*N correlation matrix R(x, y);
Figure PCTCN2018080113-appb-000004
Figure PCTCN2018080113-appb-000004
其中，r 1，1...r 1，N为按公式(1)计算得到的两两向量之间的相关性系数。 Where r 1,1 ... r 1,N are the correlation coefficients between the two vectors calculated according to formula (1).
其中，步骤3包括以下步骤；Wherein step 3 includes the following steps;
步骤3.1，计算矩阵R(x,y)的特征值，找出最大的特征值；Step 3.1, calculating the eigenvalues of the matrix R(x, y) to find the largest eigenvalue;
步骤3.2，计算矩阵R(x,y)的最大特征值对应的特征向量γ。In step 3.2, the feature vector γ corresponding to the maximum eigenvalue of the matrix R(x, y) is calculated.
进一步的，在步骤3之前先对相关性矩阵做平滑滤波处理。Further, before the step 3, the correlation matrix is first subjected to smoothing filtering processing.
其中，平滑滤波处理包括以下步骤；Wherein, the smoothing filtering process comprises the following steps;
步骤a，从每个像素点(x,y)对应的相关性矩阵R(x,y)中，取出第i个相关性系数，构成一个和图像矩阵大小相同的矩阵Ri；Step a, taking the i-th correlation coefficient from the correlation matrix R(x, y) corresponding to each pixel point (x, y), to form a matrix Ri of the same size as the image matrix;
步骤b，对矩阵Ri进行二维低通滤波；Step b, performing two-dimensional low-pass filtering on the matrix Ri;
步骤c，将滤波后的结果取代R(x,y)中对应的元素；Step c, replacing the filtered result with the corresponding element in R(x, y);
步骤d，重复a-c，直到R(x,y)中所有元素均处理完毕。In step d, repeat a-c until all elements in R(x, y) have been processed.
进一步的，步骤4中通过公式(2)计算权重；Further, in step 4, the weight is calculated by formula (2);
Figure PCTCN2018080113-appb-000005
Figure PCTCN2018080113-appb-000005
公式(2)中，γ n是特征向量γ的第n个元素，γ min是特征向量γ的最小元素，γ max是特征向量γ的最大元素,a和p为参数控制因子。 In the formula (2), γ n is the nth element of the feature vector γ, γ min is the smallest element of the feature vector γ, γ max is the largest element of the feature vector γ, and a and p are parameter control factors.
优选地，其中，a＝0.2，p＝1，但不限于此，可为其他取值。Preferably, wherein a=0.2, p=1, but is not limited thereto, and may be other values.
进一步的，步骤5中通过公式(3)对原始图像进行加权合成；Further, in step 5, the original image is weighted and synthesized by the formula (3);
Figure PCTCN2018080113-appb-000006
Figure PCTCN2018080113-appb-000006
公式(3)中，M n为第n次采集得到弥散加权原始图像，w n为权重。 Equation (3), M n is the n-shot acquisition diffusion weighted original image, w n is the weight.
其中，步骤1中以相同的扫描参数重复采集弥散加权图像N次。Wherein, in step 1, the diffusion weighted image is repeatedly acquired N times with the same scanning parameter.
本发明具有以下有益效果：The invention has the following beneficial effects:
本发明在多次采集平均技术的基础上，采用主成分分析方法，从冗余数据中自适应的检测并修正数据，抑制运动伪影、射频打火伪影等，改善图像质量；不需要增加硬件装置，且图像质量优于多次采集直接平均技术。The invention adopts a principal component analysis method on the basis of multiple acquisition averaging techniques, adaptively detects and corrects data from redundant data, suppresses motion artifacts, RF ignition artifacts, etc., and improves image quality; Hardware device, and image quality is better than multiple acquisition direct averaging techniques.
附图说明DRAWINGS
图1是本发明的流程图；Figure 1 is a flow chart of the present invention;
图2是相同扫描参数，4次采集得到的4张腹部弥散加权图像；Figure 2 is the same scan parameters, 4 abdominal diffusion weighted images obtained in 4 acquisitions;
图3是图2中4次采集数据进行直接平均合成后的弥散加权图像；3 is a diffusion-weighted image obtained by directly averaging the data collected four times in FIG. 2;
图4是图2中4次采集数据按本发明方法进行修正后的弥散加权图像；Figure 4 is a diffusion weighted image of the four acquisition data of Figure 2 corrected according to the method of the present invention;
图5是直接平均技术合成，含射频打火伪影的腹部弥散加权图像；Figure 5 is a direct average technique synthesis, a diffusion-weighted image of the abdomen containing radio frequency ignition artifacts;
图6是对应于图5中的数据，利用本发明方法修正后的腹部弥散加权图像。Fig. 6 is a graph showing the abdominal diffusion weighted image corrected by the method of the present invention corresponding to the data in Fig. 5.
具体实施方式Detailed ways
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，对本发明作进一步详细说明。In order to make the objects, technical solutions and advantages of the present invention more comprehensible, the present invention will be further described in detail with reference to the accompanying drawings.
实施例1Example 1
本实施例公开的磁共振弥散加权成像自适应修正方法，包括以下 步骤：The adaptive correction method for magnetic resonance diffusion weighted imaging disclosed in this embodiment includes the following steps:
步骤1，相同的扫描参数，重复采集弥散加权图像N次，N为自然数，N≥3；Step 1, the same scanning parameters, repeated acquisition of the diffusion weighted image N times, N is a natural number, N ≥ 3;
步骤2，基于原始图像或者压缩后的图像逐点构造相关性矩阵：具体包括以下步骤；Step 2: construct a correlation matrix point by point based on the original image or the compressed image: specifically including the following steps;
步骤2.1，对于第n次采集的图像中任意一个像素点(x,y)，取周围相邻的K个点，构成一个邻域向量Xn；Step 2.1, for any pixel point (x, y) in the image acquired in the nth time, take the neighboring K points to form a neighborhood vector Xn;
步骤2.2，对于N次重复采集的图像，每个像素点对应了N个邻域向量，按公式(1)计算其中第n个向量Xn和第m个向量Xm之间的相关性；Step 2.2: For N times of repeatedly acquired images, each pixel point corresponds to N neighborhood vectors, and the correlation between the nth vector Xn and the mth vector Xm is calculated according to formula (1);
Figure PCTCN2018080113-appb-000007
Figure PCTCN2018080113-appb-000007
公式(1)中，x i为向量Xn中的第i个元素，y i为向量Xm中的第i个元素，
Figure PCTCN2018080113-appb-000008
为向量Xn的均值，
Figure PCTCN2018080113-appb-000009
为向量Xm的均值。 In the formula (1), x i is the i-th element in the vector Xn, and y i is the i-th element in the vector Xm.
Figure PCTCN2018080113-appb-000008
Is the mean of the vector Xn,
Figure PCTCN2018080113-appb-000009
Is the mean of the vector Xm.
步骤2.3，任一像素点(x,y)对应一个N*N的相关性矩阵R(x,y)；Step 2.3, any pixel point (x, y) corresponds to an N*N correlation matrix R(x, y);
Figure PCTCN2018080113-appb-000010
Figure PCTCN2018080113-appb-000010
其中，r 1，1...r 1，N为按公式(1)计算得到的两两向量之间的相关性系数。 Where r 1,1 ... r 1,N are the correlation coefficients between the two vectors calculated according to formula (1).
步骤3，主成分分析：获得每个相关性矩阵的最大特征值对应的 特征向量；具体包括以下步骤；Step 3: Principal component analysis: obtaining a feature vector corresponding to a maximum eigenvalue of each correlation matrix; specifically comprising the following steps;
步骤3.1，计算矩阵R(x,y)的特征值，找出最大的特征值；Step 3.1, calculating the eigenvalues of the matrix R(x, y) to find the largest eigenvalue;
步骤3.2，计算矩阵R(x,y)的最大特征值对应的特征向量γ。In step 3.2, the feature vector γ corresponding to the maximum eigenvalue of the matrix R(x, y) is calculated.
步骤4，根据公式(2)计算权重；Step 4, calculating a weight according to formula (2);
Figure PCTCN2018080113-appb-000011
Figure PCTCN2018080113-appb-000011
公式(2)中，γ n是特征向量γ的第n个元素，γ min是特征向量γ的最小元素，γ max是特征向量γ的最大元素,a和p为参数控制因子。其中，参数控制因子一般但不限于a＝0.2，p＝1。 In the formula (2), γ n is the nth element of the feature vector γ, γ min is the smallest element of the feature vector γ, γ max is the largest element of the feature vector γ, and a and p are parameter control factors. Among them, the parameter control factor is generally but not limited to a=0.2, p=1.
步骤5，根据步骤4获得的权重对步骤1中采集到的原始图像进行加权合成，获得修正后的弥散加权图像。具体通过公式(3)对原始图像进行加权合成；Step 5: Perform weighted synthesis on the original image collected in step 1 according to the weight obtained in step 4, to obtain a corrected diffusion weighted image. Specifically, the original image is weighted and synthesized by the formula (3);
Figure PCTCN2018080113-appb-000012
Figure PCTCN2018080113-appb-000012
公式(3)中，M n为第n次采集得到弥散加权原始图像，w n为权重。 Equation (3), M n is the n-shot acquisition diffusion weighted original image, w n is the weight.
实施例2Example 2
本实施例与实施例1的区别在于：步骤2之前利用插值算法对所有采集的原始图像进行压缩。其好处有，第一可以减少运算量，第二可以增加后续算法输入数据的信噪比。The difference between this embodiment and Embodiment 1 is that all the collected original images are compressed by the interpolation algorithm before step 2. The benefits are that the first can reduce the amount of computation, and the second can increase the signal-to-noise ratio of the input data of the subsequent algorithm.
实施例3Example 3
本实施例与实施例1或2的区别在于：如图1所示，在步骤3之 前先对相关性矩阵做平滑滤波处理。其中，平滑滤波处理包括以下步骤；The difference between this embodiment and Embodiment 1 or 2 is that, as shown in Fig. 1, the correlation matrix is subjected to smoothing processing before step 3. Wherein, the smoothing filtering process comprises the following steps;
步骤a，从每个像素点(x,y)对应的相关性矩阵R(x,y)中，取出第i个相关性系数，构成一个和图像矩阵大小相同的矩阵Ri；Step a, taking the i-th correlation coefficient from the correlation matrix R(x, y) corresponding to each pixel point (x, y), to form a matrix Ri of the same size as the image matrix;
步骤b，对矩阵Ri进行二维低通滤波；Step b, performing two-dimensional low-pass filtering on the matrix Ri;
步骤c，将滤波后的结果取代R(x,y)中对应的元素；Step c, replacing the filtered result with the corresponding element in R(x, y);
步骤d，重复a-c，直到R(x,y)中所有元素均处理完毕。In step d, repeat a-c until all elements in R(x, y) have been processed.
如图2所示，如箭头所指，第1张图像中可见明显的运动伪影，导致部分信号完全丢失。如图3所示，采用直接平均进行合成后的弥散加权图像，伪影抑制的程度有限，图片质量差；如图4所示，按本发明方法修正后的图像更准确。如图5、6所示，按本发明方法修正后的图像中射频打火伪影明显降低。As shown in FIG. 2, as indicated by the arrow, significant motion artifacts are visible in the first image, resulting in partial loss of partial signals. As shown in FIG. 3, the diffusion-weighted image synthesized by direct averaging has a limited degree of artifact suppression and poor picture quality; as shown in FIG. 4, the image corrected by the method of the present invention is more accurate. As shown in Figures 5 and 6, the RF ignition artifacts in the image corrected by the method of the present invention are significantly reduced.
本发明在多次采集平均技术的基础上，基于主成分分析方法自适应计算各次扫描数据权重，按所得权重进行加权合成，抑制运动伪影、射频打火伪影，改善图像质量且不需要增加硬件装置。Based on the multiple acquisition averaging technique, the invention adaptively calculates the weights of each scan data based on the principal component analysis method, performs weighted synthesis according to the obtained weights, suppresses motion artifacts, RF ignition artifacts, and improves image quality without requiring Add hardware devices.
当然，本发明还可有其它多种实施方式，在不背离本发明精神及其实质的情况下，熟悉本领域的技术人员可根据本发明作出各种相应的改变和变形，但这些相应的改变和变形都应属于本发明所附的权利要求的保护范围。There are a variety of other modifications and variations that can be made by those skilled in the art without departing from the spirit and scope of the invention. And modifications are intended to fall within the scope of the appended claims.

Claims (10)

  1. 磁共振弥散加权成像自适应修正方法，其特征在于：包括以下步骤：An adaptive correction method for magnetic resonance diffusion weighted imaging, comprising: the following steps:
    步骤1，重复采集弥散加权图像N次，N为自然数，N≥3；Step 1, repeatedly collecting the diffusion weighted image N times, N is a natural number, N ≥ 3;
    步骤2，基于原始图像或者压缩后的图像逐点构造相关性矩阵；Step 2: construct a correlation matrix point by point based on the original image or the compressed image;
    步骤3，主成分分析；获得每个相关性矩阵的最大特征值对应的特征向量；Step 3: principal component analysis; obtaining a feature vector corresponding to a maximum eigenvalue of each correlation matrix;
    步骤4，根据特征向量计算权重；Step 4, calculating a weight according to the feature vector;
    步骤5，根据步骤4获得的权重对步骤1中采集到的原始图像进行加权合成，获得修正后的弥散加权图像。Step 5: Perform weighted synthesis on the original image collected in step 1 according to the weight obtained in step 4, to obtain a corrected diffusion weighted image.
  2. 如权利要求1所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤2之前利用插值算法对所有采集的原始图像进行压缩。The magnetic resonance diffusion weighted imaging adaptive correction method according to claim 1, wherein all the collected original images are compressed by the interpolation algorithm before step 2.
  3. 如权利要求1或2所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤2包括以下步骤：The magnetic resonance diffusion weighted imaging adaptive correction method according to claim 1 or 2, wherein the step 2 comprises the following steps:
    步骤2.1，对于第n次采集的图像中任意一个像素点(x,y)，取周围相邻的K个点，构成一个邻域向量Xn；Step 2.1, for any pixel point (x, y) in the image acquired in the nth time, take the neighboring K points to form a neighborhood vector Xn;
    步骤2.2，对于N次重复采集的图像，每个像素点对应了N个邻域向量，按公式(1)计算其中第n个向量Xn和第m个向量Xm之间的相关性；Step 2.2: For N times of repeatedly acquired images, each pixel point corresponds to N neighborhood vectors, and the correlation between the nth vector Xn and the mth vector Xm is calculated according to formula (1);
    Figure PCTCN2018080113-appb-100001
    Figure PCTCN2018080113-appb-100001
    公式(1)中，x i为向量Xn中的第i个元素，y i为向量Xm中的第i个元素，
    Figure PCTCN2018080113-appb-100002
    为向量Xn的均值，
    Figure PCTCN2018080113-appb-100003
    为向量Xm的均值。     In the formula (1), x i is the i-th element in the vector Xn, and y i is the i-th element in the vector Xm.
    Figure PCTCN2018080113-appb-100002
    Is the mean of the vector Xn,
    Figure PCTCN2018080113-appb-100003
    Is the mean of the vector Xm.
    步骤2.3，任一像素点(x,y)对应一个N*N的相关性矩阵R(x,y)；Step 2.3, any pixel point (x, y) corresponds to an N*N correlation matrix R(x, y);
    Figure PCTCN2018080113-appb-100004
    Figure PCTCN2018080113-appb-100004
    其中，r 1，1…r 1，N为按公式(1)计算得到的两两向量之间的相关性系数。 Where r 1,1 ... r 1,N are the correlation coefficients between the two vectors calculated according to formula (1).
  4. 如权利要求3所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤3包括以下步骤；The magnetic resonance diffusion weighted imaging adaptive correction method according to claim 3, wherein the step 3 comprises the following steps;
    步骤3.1，计算矩阵R(x,y)的特征值，找出最大的特征值；Step 3.1, calculating the eigenvalues of the matrix R(x, y) to find the largest eigenvalue;
    步骤3.2，计算矩阵R(x,y)的最大特征值对应的特征向量γ。In step 3.2, the feature vector γ corresponding to the maximum eigenvalue of the matrix R(x, y) is calculated.
  5. 如权利要求1或4所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤3之前先对相关性矩阵做平滑滤波处理。The adaptive correction method for magnetic resonance diffusion weighted imaging according to claim 1 or 4, characterized in that before the step 3, the correlation matrix is first subjected to smoothing filtering processing.
  6. 如权利要求5所述的磁共振弥散加权成像自适应修正方法，其特征在于：平滑滤波处理包括以下步骤；The adaptive correction method for magnetic resonance diffusion weighted imaging according to claim 5, wherein the smoothing filtering process comprises the following steps;
    步骤a，从每个像素点(x,y)对应的相关性矩阵R(x,y)中，取出第i个相关性系数，构成一个和图像矩阵大小相同的矩阵Ri；Step a, taking the i-th correlation coefficient from the correlation matrix R(x, y) corresponding to each pixel point (x, y), to form a matrix Ri of the same size as the image matrix;
    步骤b，对矩阵Ri进行二维低通滤波；Step b, performing two-dimensional low-pass filtering on the matrix Ri;
    步骤c，将滤波后的结果取代R(x,y)中对应的元素；Step c, replacing the filtered result with the corresponding element in R(x, y);
    步骤d，重复a-c，直到R(x,y)中所有元素均处理完毕。In step d, repeat a-c until all elements in R(x, y) have been processed.
  7. 如权利要求4或6所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤4中通过公式(2)计算权重；The adaptive correction method for magnetic resonance diffusion weighted imaging according to claim 4 or 6, wherein the weight is calculated by the formula (2) in the step 4;
    Figure PCTCN2018080113-appb-100005
    Figure PCTCN2018080113-appb-100005
    公式(2)中，γ n是特征向量γ的第n个元素，γ min是特征向量γ的最小元素，γ max是特征向量γ的最大元素,a和p为参数控制因子。 In the formula (2), γ n is the nth element of the feature vector γ, γ min is the smallest element of the feature vector γ, γ max is the largest element of the feature vector γ, and a and p are parameter control factors.
  8. 如权利要求7所述的磁共振弥散加权成像自适应修正方法，其特征在于：其中，a＝0.2，p＝1。The adaptive correction method for magnetic resonance diffusion weighted imaging according to claim 7, wherein a=0.2 and p=1.
  9. 如权利要求7或8所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤5中通过公式(3)对原始图像进行加权合成；The adaptive correction method for magnetic resonance diffusion weighted imaging according to claim 7 or 8, wherein in step 5, the original image is weighted and synthesized by the formula (3);
    Figure PCTCN2018080113-appb-100006
    Figure PCTCN2018080113-appb-100006
    公式(3)中，M n为第n次采集得到弥散加权原始图像，w n为权重。 Equation (3), M n is the n-shot acquisition diffusion weighted original image, w n is the weight.
  10. 如权利要求1所述的磁共振弥散加权成像自适应修正方法，其特征在于：步骤1中以相同的扫描参数重复采集弥散加权图像N次。The adaptive correction method for magnetic resonance diffusion weighted imaging according to claim 1, wherein in step 1, the diffusion weighted image is repeatedly acquired N times with the same scanning parameter.
PCT/CN2018/080113 2018-02-09 2018-03-23 Magnetic resonance diffusion weighted imaging self-adaptive correction method WO2019153443A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810136588.1 2018-02-09
CN201810136588.1A CN108090937B (en) 2018-02-09 2018-02-09 Self-adaptive correction method for magnetic resonance diffusion weighted imaging

Publications (1)

Publication Number Publication Date
WO2019153443A1 true WO2019153443A1 (en) 2019-08-15

Family

ID=62194414

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/080113 WO2019153443A1 (en) 2018-02-09 2018-03-23 Magnetic resonance diffusion weighted imaging self-adaptive correction method

Country Status (2)

Country Link
CN (1) CN108090937B (en)
WO (1) WO2019153443A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080157764A1 (en) * 2006-12-19 2008-07-03 Hiroyuki Kabasawa Magnetic resonance imaging apparatus, magnetic resonance imaging method, and diffusion tensor color map image generating apparatus
US20080205733A1 (en) * 2006-10-19 2008-08-28 Brown Universisty Quantitative tract-of-interest metrics for white matter integrity based on diffusion tensor MRI data
US20090016590A1 (en) * 2007-07-10 2009-01-15 Isaac Wen-Yih Tseng Method of Analyzing Diffusion-Weighted Magnetic Resonance Image
CN102136071A (en) * 2011-03-09 2011-07-27 北京航空航天大学 Spatial correlation matrix-based image characteristic analysis model
CN104323777A (en) * 2014-10-30 2015-02-04 西安交通大学医学院第一附属医院 Diffusion magnetic resonance imaging motion artifact eliminating method
CN104323775A (en) * 2013-07-22 2015-02-04 株式会社日立制作所 Magnetic resonance imaging apparatus, image processing apparatus, and image processing method

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6160398A (en) * 1999-07-02 2000-12-12 Vista Clara, Inc. Adaptive reconstruction of phased array NMR imagery
CN102217934A (en) * 2011-04-08 2011-10-19 中国科学院深圳先进技术研究院 Magnetic resonance imaging method and system
CN102749600B (en) * 2012-05-30 2014-12-24 苏州朗润医疗系统有限公司 Synthetic method of magnetic resonance multi-channel image
CN104599244B (en) * 2014-12-23 2018-06-15 中国科学院深圳先进技术研究院 The denoising method and system of diffusion tensor imaging
EP3322997A1 (en) * 2015-07-15 2018-05-23 Koninklijke Philips N.V. Mr imaging with motion detection
CN107240125B (en) * 2016-03-28 2020-02-07 上海联影医疗科技有限公司 Diffusion weighted imaging method
CN106780643B (en) * 2016-11-21 2019-07-26 清华大学 Magnetic resonance repeatedly excites diffusion imaging to move antidote

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080205733A1 (en) * 2006-10-19 2008-08-28 Brown Universisty Quantitative tract-of-interest metrics for white matter integrity based on diffusion tensor MRI data
US20080157764A1 (en) * 2006-12-19 2008-07-03 Hiroyuki Kabasawa Magnetic resonance imaging apparatus, magnetic resonance imaging method, and diffusion tensor color map image generating apparatus
US20090016590A1 (en) * 2007-07-10 2009-01-15 Isaac Wen-Yih Tseng Method of Analyzing Diffusion-Weighted Magnetic Resonance Image
CN102136071A (en) * 2011-03-09 2011-07-27 北京航空航天大学 Spatial correlation matrix-based image characteristic analysis model
CN104323775A (en) * 2013-07-22 2015-02-04 株式会社日立制作所 Magnetic resonance imaging apparatus, image processing apparatus, and image processing method
CN104323777A (en) * 2014-10-30 2015-02-04 西安交通大学医学院第一附属医院 Diffusion magnetic resonance imaging motion artifact eliminating method

Also Published As

Publication number Publication date
CN108090937A (en) 2018-05-29
CN108090937B (en) 2021-03-19

Similar Documents

Publication Publication Date Title
van Amerom et al. Fetal cardiac cine imaging using highly accelerated dynamic MRI with retrospective motion correction and outlier rejection
US9797974B2 (en) Nonrigid motion correction in 3D using autofocusing with localized linear translations
US8649585B2 (en) Method and system for retrospective image combination under minimal total deformation constrain for free-breathing cardiac magnetic resonance imaging with motion correction
US9684979B2 (en) MRI 3D cine imaging based on intersecting source and anchor slice data
US10123720B2 (en) Retrospectively correlated turbo spin echo imaging
US20080205730A1 (en) Independent Motion Correction In Respective Signal Channels Of A Magnetic Resonance Imaging System
US20180217217A1 (en) System and method for dynamic, cardiac phase-resolved quantitative longitudinal relaxation parameter mapping
CN101915901A (en) Magnetic resonance imaging method and device
Cheng et al. Learning data consistency and its application to dynamic MR imaging
WO2005106534A2 (en) Improving inter-subject coherence in dt-mri
US20140232395A1 (en) Method and system for multi-shot spiral magnetic resonance elastography pulse sequence
US11835612B2 (en) System and method for motion correction of magnetic resonance image
Vuissoz et al. Free‐breathing imaging of the heart using 2D cine‐GRICS (generalized reconstruction by inversion of coupled systems) with assessment of ventricular volumes and function
Ferreira et al. Accelerating Cardiac Diffusion Tensor Imaging With a U‐Net Based Model: Toward Single Breath‐Hold
US9857446B2 (en) Robust self-navigating MRI using large coil arrays
Kafali et al. Phase‐correcting non‐local means filtering for diffusion‐weighted imaging of the spinal cord
WO2013033558A1 (en) System and method for joint degradation estimation and image reconstruction in magnetic resonance imaging
WO2019153443A1 (en) Magnetic resonance diffusion weighted imaging self-adaptive correction method
WO2006109550A1 (en) Magnetic resonance imaging device and method
EP4028787A1 (en) Systems and methods of deep learning for large-scale dynamic magnetic resonance image reconstruction
WO2020034674A1 (en) Quantitative myocardial magnetic resonance imaging method and device, and storage medium
JP6356412B2 (en) Magnetic resonance apparatus and program
Andreychenko et al. Respiratory motion model based on the noise covariance matrix of a receive array
JPH06327649A (en) Method for recomposing images in nuclear magnetic resonance imaging apparatus
Chen et al. Automatic extraction and sign determination of respiratory signal in real-time cardiac magnetic resonance imaging

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: 18904578

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: 18904578

Country of ref document: EP

Kind code of ref document: A1