An Automatic Localization Method and System for Magnetic Resonance Imaging Based on Image Registration

Abstract

The present invention provides a method and system for automatic positioning of magnetic resonance imaging based on image registration, including: performing image registration on the three-dimensional magnetic resonance image of a to-be-imaged part with a plurality of pre-generated three-dimensional magnetic resonance template images respectively to obtain registration images and corresponding first registration parameters, and adding each registration image, the corresponding three-dimensional magnetic resonance template image and the registration parameters into a data set respectively; for each data set, calculating the similarity between the registration image and the three-dimensional magnetic resonance template image respectively, and processing according to each similarity and the first registration parameters in the corresponding data set to obtain a second registration parameter; processing according to the second registration parameter and a standard positioning parameter associated with the to-be-imaged part obtained in advance to obtain an automatic positioning parameter, which is used as the automatic positioning result of magnetic resonance imaging for the current magnetic resonance scan. The beneficial effect is to effectively improve the efficiency of the magnetic resonance imaging workflow and reduce the uncertainty brought by the operator's human factors.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic resonance imaging, and particularly to a method and system for automatic positioning of magnetic resonance imaging based on image registration. Background Art

[0002] Magnetic resonance imaging technology (abbreviated as MRI) refers to using corresponding radio frequency pulses to excite the hydrogen nuclei (protons) of the object to be examined placed in a static magnetic field, collecting the generated magnetic resonance signals, and reconstructing them into images through computer processing. Since magnetic resonance imaging technology has the advantages of no ionizing radiation damage, high soft tissue contrast, high image resolution, flexible selection of imaging parameters and scanning orientations, and the ability to display blood vessels without the need for a contrast agent, it is widely used in clinical medical diagnosis, and its application and development prospects are very broad.

[0003] In clinical applications and scientific research, it is usually required that the obtained magnetic resonance images have accurate and consistent spatial positioning. This requires pre-scanning positioning at the beginning of the formal magnetic resonance imaging sequence scan. However, in clinical or scientific magnetic resonance scans, it usually depends on the operator (such as radiologists, researchers, etc.) to perform manual positioning before the start of the scan process. That is, first, a set of positioning images are collected, and the operator identifies the characteristic structures therein and uses them as a basis for positioning. However, this method depends on the operator's experience, is easily interfered by the operator's subjective factors, and is not conducive to accelerating the scanning process.

[0004] In the prior art, some researchers have used the left-right symmetry of the human anatomical structure for automatic positioning of magnetic resonance imaging scans. However, when there are organic lesions or severe deformities, the symmetry of the human anatomical structure is relatively low; and the symmetry of the abdomen is not high enough. Many researchers have also developed automatic positioning algorithms, which can first automatically identify specific structures in the imaging area and then determine the positioning parameters. However, in the case of organic lesions in the tissue, the corresponding structures under normal conditions are significantly different from the specific structures. In this case, it is either difficult to accurately identify the specific structures, or the identified specific structures cannot be used to obtain accurate positioning parameters. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention provides a method for automatic positioning of magnetic resonance imaging based on image registration, including:

[0006] Step S1, obtaining a three-dimensional magnetic resonance image of a part to be imaged by magnetic resonance scanning, performing image registration on the three-dimensional magnetic resonance image with a plurality of pre-generated three-dimensional magnetic resonance template images respectively to obtain registration images and corresponding first registration parameters, and adding each of the registration images, the corresponding three-dimensional magnetic resonance template images, and the registration parameters into a data set;

[0007] In step S2, for each of the said data sets, calculate the similarity between the registration image and the magnetic resonance three-dimensional template image respectively, and process a second registration parameter based on each of the similarities and the first registration parameter in the corresponding data set;

[0008] In step S3, process an automatic positioning parameter based on the second registration parameter and a standard positioning parameter associated with the to-be-imaged part obtained in advance, and use it as the automatic positioning result of magnetic resonance imaging for the current magnetic resonance scan.

[0009] Preferably, before performing step S1, it further includes the process of obtaining the standard positioning parameter, including:

[0010] In step A1, collect at least one transverse magnetic resonance image, one sagittal magnetic resonance image and one coronal magnetic resonance image of the to-be-imaged part respectively, and perform feature structure annotation on them to obtain a transverse annotation image, a sagittal annotation image and a coronal annotation image;

[0011] In step A2, process the standard positioning parameter based on the transverse annotation image, the sagittal annotation image and the coronal annotation image.

[0012] Preferably, the magnetic resonance three-dimensional image and each of the magnetic resonance three-dimensional template images have a first resolution, the transverse magnetic resonance image, the sagittal magnetic resonance image and the coronal magnetic resonance image have a second resolution, and the first resolution is less than the second resolution.

[0013] Preferably, step S2 includes:

[0014] In step S21a, for each of the said data sets, calculate the similarity between the registration image and the magnetic resonance three-dimensional template image respectively;

[0015] In step S22a, extract the first registration parameter in the data set corresponding to the maximum value among each of the similarities, and use the first registration parameter as the second registration parameter.

[0016] Preferably, step S2 includes:

[0017] In step S21b, for each of the said data sets, calculate the similarity between the registration image and the magnetic resonance three-dimensional template image respectively, and arrange each of the similarities in descending order to obtain a similarity sequence;

[0018] Step S22b: successively extract the first registration parameters in the data set corresponding to the preset number of the similarities with higher rankings from the similarity sequence, and perform weighted average processing on each of the first registration parameters to obtain the second registration parameter.

[0019] Preferably, the second registration parameter includes a registration matrix formed by three registration translation parameters and three registration rotation parameters, and the standard positioning parameter includes a standard positioning matrix formed by three standard translation parameters and three standard rotation parameters;

[0020] Then in the step S3, multiply the registration matrix and the standard positioning matrix to obtain an automatic positioning matrix, and extract three positioning translation parameters and three positioning rotation parameters from the automatic positioning matrix as the automatic positioning parameters.

[0021] Preferably, each of the magnetic resonance three-dimensional template images is associated with corresponding first basic information of the scanned subject, and the magnetic resonance three-dimensional image obtained by magnetic resonance scanning is associated with corresponding second basic information of the scanned subject;

[0022] Then in the step S1, it further includes extracting each of the magnetic resonance three-dimensional template images corresponding to the second basic information that matches the first basic information, and then performing image registration on the magnetic resonance three-dimensional image and each of the extracted magnetic resonance three-dimensional template images to obtain the registered image and the corresponding first registration parameter.

[0023] Preferably, the first basic information and the second basic information include the imaging part to be imaged of the scanned subject, and / or the age group.

[0024] Preferably, in the step S1, the image registration method is an affine transformation method, or a rigid body transformation method, or a non-linear transformation method.

[0025] The present invention also provides a magnetic resonance imaging automatic positioning system based on image registration, which applies the above magnetic resonance imaging automatic positioning method. The magnetic resonance imaging automatic positioning system includes:

[0026] An image registration module, configured to perform image registration on a magnetic resonance three-dimensional image of an imaging part to be imaged obtained by magnetic resonance scanning respectively with a plurality of pre-generated magnetic resonance three-dimensional template images, obtain a registered image and the corresponding first registration parameter, and add each of the registered images, the corresponding magnetic resonance three-dimensional template images, and the registration parameters to a data set respectively;

[0027] An image processing module, connected to the image registration module, is configured to calculate the similarity between the registered image and the magnetic resonance three-dimensional template image for each of the data sets, and process a second registration parameter based on each of the similarities and the first registration parameter in the corresponding data set;

[0028] An automatic positioning module, connected to the image processing module, is configured to process an automatic positioning parameter based on the second registration parameter and a standard positioning parameter associated with the to-be-imaged part obtained in advance, and use it as the automatic positioning result of magnetic resonance imaging for the current magnetic resonance scan.

[0029] The above technical solution has the following advantages or beneficial effects: The automatic positioning of magnetic resonance imaging is performed by means of image registration, which considers the global information of the to-be-imaged part rather than the positioning method of a specific structure. Therefore, it has higher stability, effectively improves the efficiency of the magnetic resonance imaging workflow, and reduces the uncertainty brought by human factors of the operator. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 In a preferred embodiment of the present invention, it is a schematic flowchart of a method for automatic positioning of magnetic resonance imaging based on image registration;

[0031] Figure 2 In a preferred embodiment of the present invention, it is a schematic flowchart of the process for obtaining the standard positioning parameter;

[0032] Figure 3 In a preferred embodiment of the present invention, it is a schematic flowchart of the process for obtaining the second registration parameter;

[0033] Figure 4 In a preferred embodiment of the present invention, it is a schematic flowchart of the process for obtaining the second registration parameter;

[0034] Figure 5 In a preferred embodiment of the present invention, it is a schematic structural diagram of a system for automatic positioning of magnetic resonance imaging based on image registration. DETAILED DESCRIPTION OF THE INVENTION

[0035] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment, and other embodiments may also fall within the scope of the present invention as long as they conform to the gist of the present invention.

[0036] In a preferred embodiment of the present invention, in view of the above problems existing in the prior art, a method for automatic positioning of magnetic resonance imaging based on image registration is provided, as shown in Figure 1 and includes:

[0037] Step S1: Obtain a three-dimensional magnetic resonance image of a part to be imaged by magnetic resonance scanning. Perform image registration on the three-dimensional magnetic resonance image with multiple pre-generated three-dimensional magnetic resonance template images respectively to obtain registration images and corresponding first registration parameters, and add each registration image, the corresponding three-dimensional magnetic resonance template image, and the registration parameters into a data set respectively;

[0038] Step S2: For each data set, calculate the similarity between the registration image and the three-dimensional magnetic resonance template image respectively, and process according to each similarity and the first registration parameter in the corresponding data set to obtain a second registration parameter;

[0039] Step S3: Process according to the second registration parameter and a standard positioning parameter associated with the part to be imaged obtained in advance to obtain an automatic positioning parameter, which is used as the automatic positioning result of magnetic resonance imaging for the current magnetic resonance scanning.

[0040] Specifically, in this embodiment, before performing automatic positioning of magnetic resonance imaging, a template library can be constructed in advance to store multiple pre-obtained three-dimensional magnetic resonance template images and standard positioning parameters. Among them, the above three-dimensional magnetic resonance template images can be obtained by image reconstruction of low-resolution three-dimensional data with the same resolution in each direction, such as 8×8×8mm 3 spatial resolution, and preferably T1-weighted images obtained by three-dimensional gradient echo (GRE) imaging are used. Using T1-weighted or proton density weighted images can minimize the influence of tissue lesions on positioning as much as possible. Further preferably, the scanning parameters of each three-dimensional magnetic resonance template image are kept as consistent as possible. The scanning parameters include but are not limited to repetition time (TR), echo time (TE), flip angle (FA), field of view (FOV), acquisition matrix size, acquisition bandwidth, etc.

[0041] The process of obtaining the standard positioning parameter is as Figure 2 shown, and includes:

[0042] Step A1: Collect at least one transverse magnetic resonance image, one sagittal magnetic resonance image, and one coronal magnetic resonance image of the part to be imaged respectively, and perform feature structure annotation to obtain a transverse annotation image, a sagittal annotation image, and a coronal annotation image respectively;

[0043] Step A2: Process the cross-sectional labeled image, sagittal labeled image, and coronal labeled image to obtain standard positioning parameters.

[0044] Among them, the three-dimensional magnetic resonance image and each three-dimensional magnetic resonance template image have a first resolution, and the cross-sectional magnetic resonance image, sagittal magnetic resonance image, and coronal magnetic resonance image have a second resolution, and the first resolution is less than the second resolution. In other words, the cross-sectional magnetic resonance image, sagittal magnetic resonance image, and coronal magnetic resonance image need to have a sufficiently high resolution, such as a spatial resolution of 2×2×10mm 3 so that an experienced clinical radiologist can manually label the characteristic structures using the high-resolution cross-sectional magnetic resonance image, sagittal magnetic resonance image, and coronal magnetic resonance image. Preferably, the cross-sectional magnetic resonance image, sagittal magnetic resonance image, and coronal magnetic resonance image can be T2-weighted (T2weighted) images, which are beneficial for clinical radiologists to identify specific anatomical structures. More preferably, after obtaining multiple three-dimensional magnetic resonance template images with the first resolution, each three-dimensional magnetic resonance template image can be preprocessed to improve its resolution. The preprocessing methods include, but are not limited to: zero-padding in the k-space and then performing image reconstruction; using interpolation in the image domain, including linear interpolation, square interpolation, cubic interpolation, spline interpolation, etc.

[0045] Further preferably, multiple experienced clinical radiologists can use high-resolution images to manually position the characteristic structures, and then respectively obtain the standard positioning parameters processed after each doctor's manual positioning. Subsequently, the standard positioning parameters processed by each clinical radiologist's manual positioning are averaged, and the result of the averaging process is used as the final standard positioning parameter to further eliminate the influence of personal subjective experience on the subsequent automatic positioning result and further improve the accuracy of automatic positioning.

[0046] The above-mentioned characteristic structure has different positioning criteria for different parts to be imaged, and the corresponding medical positioning criteria can be specifically referred to. Taking the brain as an example of the part to be imaged, a commonly used medical positioning criterion is to adopt a cross-sectional view and make the imaging plane parallel to the line connecting the anterior commissure (AC) and the posterior commissure (PC), that is, the anterior-posterior commissure line (AC-PC line), to obtain multi-layer images covering the entire brain area from the skull vertex to the skull base. In addition to using the anterior-posterior commissure line as the reference line, there are also many other options, including the line connecting the external auditory meatus and the lower edge of the ipsilateral orbit, that is, the orbito-meatal line (OM line); the line connecting the genu of the corpus callosum and the splenium of the corpus callosum, that is, the subcallosal line (SC line); the brainstem vertical line, etc. Preferably, for brain magnetic resonance images, the corresponding characteristic structure includes marking the anterior-posterior commissure line in the sagittal plane annotation image, and marking the left-right symmetry line in the coronal plane annotation image and the cross-sectional plane annotation image. Furthermore, the translation amounts of the intersection point of the above three lines in three directions relative to the image center are the three standard translation parameters, and the angles of each line in the corresponding image, including the pitch angle, yaw angle, and roll angle, are respectively the three standard rotation parameters. The three standard translation parameters and the three standard rotation parameters form the standard positioning parameters.

[0047] For a superconducting magnetic resonance imaging system, the three standard translation parameters tx, ty, and tz can respectively correspond to the left-right direction x, the anterior-posterior direction y, and the head-foot direction z; the three standard rotation parameters are respectively defined as rotations around the x, y, and z coordinate axes of the three directions. Borrowing the convention in the aerospace field, they can be respectively called the pitch angle, yaw angle, and roll angle, and are denoted as θx, θy, and θz.

[0048] After obtaining each magnetic resonance three-dimensional template image and standard positioning parameters in advance, before formally performing magnetic resonance imaging, firstly perform magnetic resonance scanning on the imaging part to obtain a corresponding magnetic resonance three-dimensional image, and then perform image registration on the magnetic resonance three-dimensional image with each magnetic resonance three-dimensional template image to obtain a registered image and a corresponding first registration parameter, and each registered image, the corresponding magnetic resonance three-dimensional template image and the registration parameter are respectively added to the same data set. Taking the magnetic resonance image as image A and the magnetic resonance three-dimensional template images as images B1, B2 and B3, image A is configured with images B1, B2 and B3 respectively, and the corresponding registration image C1 and the first registration parameter D1, the registration image C2 and the first registration parameter D2, the registration image C3 and the first registration parameter D3 are obtained. The corresponding data sets include the following three: {C1, B1, D1}, {C2, B2, D2}, {C3, B3, D3}, and then for each data set, the similarity between images C1 and B1, the similarity between images C2 and B2, and the similarity between images C3 and B3 are calculated respectively, and then the second registration parameters are obtained according to the similarities and the first registration parameters in the corresponding data set.

[0049] There are two preferred ways to obtain the second registration parameter processed above. The first way is to directly select the first registration parameter in the data set with the largest similarity as the second registration parameter. Specifically, Figure 3 As shown, step S2 includes:

[0050] Step S21a, for each data set, respectively calculating the similarity between the registration image and the magnetic resonance three-dimensional template image;

[0051] Step S22a, extracting the first registration parameter in the data set corresponding to the maximum value of each similarity, and using the first registration parameter as the second registration parameter.

[0052] The second method is to select multiple first registration parameters with high similarity and perform weighted averaging to obtain the result. Specifically, Figure 4 As shown, step S2 includes:

[0053] Step S21b, for each data set, respectively calculating the similarity between the registration image and the magnetic resonance three-dimensional template image, and arranging the similarities in descending order to obtain a similarity sequence;

[0054] Step S22b, sequentially extracting first registration parameters in a data set corresponding to a preset number of similarities ranked top from the similarity sequence, and performing weighted averaging processing on each first registration parameter to obtain a second registration parameter.

[0055] In this embodiment, the above preset quantity can be any number greater than 1, which can be specifically configured according to requirements. Moreover, for the similarity with a higher ranking, the corresponding weight configured during the weighted average operation is greater.

[0056] The above similarity can be measured by including, but not limited to, the correlation coefficient, structural similarity (SSIM), mutual information, etc.

[0057] Further specifically, the second registration parameter includes a registration matrix formed by three registration translation parameters and three registration rotation parameters, and the standard positioning parameter includes a standard positioning matrix formed by three standard translation parameters and three standard rotation parameters.

[0058] Then, in step S3, the registration matrix and the standard positioning matrix are multiplied to obtain an automatic positioning matrix, and three positioning translation parameters and three positioning rotation parameters are extracted from the automatic positioning matrix as the automatic positioning parameters.

[0059] In this embodiment, the automatic positioning matrix is calculated using the following formula:

[0060] M = M1 × M2

[0061] Wherein, M is used to represent the automatic positioning matrix, M1 is used to represent the registration matrix, and M2 is used to represent the standard positioning matrix.

[0062] In a preferred embodiment of the present invention, each magnetic resonance three-dimensional template image is associated with the first basic information of the corresponding subject being scanned, and the magnetic resonance three-dimensional image obtained by magnetic resonance scanning is associated with the second basic information of the corresponding subject being scanned.

[0063] Then, in step S1, it further includes extracting each magnetic resonance three-dimensional template image corresponding to the second basic information that matches the first basic information, and then respectively performing image registration on the magnetic resonance three-dimensional image and each extracted magnetic resonance three-dimensional template image to obtain the registered image and the corresponding first registration parameter.

[0064] In a preferred embodiment of the present invention, the first basic information and the second basic information include the imaging part to be imaged of the subject being scanned, and / or the age group.

[0065] Specifically, in this embodiment, considering that when there are significant differences in the ages of different patients, the sizes and shapes of the parts to be imaged also vary; and when there are organic lesions, the anatomical structures of the parts to be imaged also differ. In these cases, it is difficult to meet the requirements of various situations that may be encountered during childbirth using a single template, while using multiple templates can effectively solve this problem. Further, during image registration, corresponding magnetic resonance three-dimensional template images of the parts to be imaged and the corresponding age groups can be selected from each magnetic resonance three-dimensional template image for image registration, so as to effectively reduce the amount of computation.

[0066] In a preferred embodiment of the present invention, in step S1, the image registration method is an affine transformation method, or a rigid body transformation method, or a non-linear transformation method.

[0067] Specifically, in this embodiment, preferably, different image registration methods can be configured according to different parts to be imaged. Among them, the affine transformation method fully considers the rotation, translation, scale, and shear of the image. After obtaining the affine transformation matrix, the parameters of rotation and translation can be extracted from it, and specifically, it can be achieved by decomposing the affine transformation matrix into translation, rotation, scale, and shear matrices. The affine transformation matrix can be decomposed as follows:

[0068]

[0069] Among them,

[0070]

[0071] Among them, T(t x ,t y ,t z ) is the translation matrix, R(θ x ), R(θ y ), R(θ z ) are the rotation matrices along the x, y, and z directions respectively, S(s x ,s y ,s z ) is the scale matrix, and H(h x ,h y ,h z ) is the shear matrix. Solving the above equation can extract the translation parameters tx, ty, tz and the rotation parameters θx, θy, and θz.

[0072] Among them, the rigid body transformation method can be used for image registration. Specifically, the rotation parameters can be estimated by Fourier-Mellin Transformation, and the translation parameters can be estimated by phase matching analysis in the frequency domain. The specific calculation process is not elaborated here.

[0073] For abdominal imaging, a non-linear transformation can be considered to estimate the warp field, also known as the deformation field, and then the translation and rotation parameters can be extracted from it. In specific implementation, the objective function of the non-linear transformation can be the mean square error (MSE), that is, the average of the squared image differences; it can also be the maximization of mutual information, etc.

[0074] The present invention also provides a magnetic resonance imaging automatic positioning system based on image registration, which applies the above-mentioned magnetic resonance imaging automatic positioning method. As Figure 5 shown, the magnetic resonance imaging automatic positioning system includes:

[0075] An image registration module 1, configured to perform image registration on the magnetic resonance three-dimensional image of a to-be-imaged part obtained by magnetic resonance scanning with a plurality of pre-generated magnetic resonance three-dimensional template images respectively, obtain the registered images and the corresponding first registration parameters, and add each registered image, the corresponding magnetic resonance three-dimensional template image, and the registration parameters into a data set respectively;

[0076] An image processing module 2, connected to the image registration module 1, configured to calculate the similarity between the registered image and the magnetic resonance three-dimensional template image for each data set respectively, and process a second registration parameter based on each similarity and the first registration parameter in the corresponding data set;

[0077] An automatic positioning module 3, connected to the image processing module 2, configured to process an automatic positioning parameter based on the second registration parameter and a standard positioning parameter associated with the to-be-imaged part obtained in advance, and use it as the magnetic resonance imaging automatic positioning result of the current magnetic resonance scanning.

[0078] The above are only the preferred embodiments of the present invention, and do not limit the implementation manners and protection scope of the present invention. For those skilled in the art, it should be realized that all the equivalent replacements and obvious changes made by using the content of this specification and the drawings should be included in the protection scope of the present invention.

Claims

1. An automatic positioning method for magnetic resonance imaging based on image registration, characterized in that, include: Step S1, obtaining a magnetic resonance three-dimensional image of a part to be imaged by magnetic resonance scanning, performing image registration on the magnetic resonance three-dimensional image and a plurality of pre-generated magnetic resonance three-dimensional template images to obtain registered images and corresponding first registration parameters, and adding each of the registered images, the corresponding magnetic resonance three-dimensional template image and the first registration parameters to a data set; Step S2, for each of the data sets, respectively calculating the similarity between the registered image and the magnetic resonance three-dimensional template image, and obtaining a second registration parameter according to each of the similarities and the first registration parameter in the corresponding data set; Step S3, obtaining automatic positioning parameters according to the second registration parameters and a pre-acquired standard positioning parameter associated with the part to be imaged as the magnetic resonance imaging automatic positioning result of the current magnetic resonance scan; Before executing step S1, a process of obtaining the standard positioning parameters is also included, including: Step A1, respectively collecting at least one cross-sectional magnetic resonance image, one sagittal magnetic resonance image and one coronal magnetic resonance image of the part to be imaged, and respectively performing feature structure annotation to obtain a cross-sectional annotated image, a sagittal annotated image and a coronal annotated image; Step A2, obtaining the standard positioning parameters according to the cross-sectional labeled image, the sagittal labeled image and the coronal labeled image; The three-dimensional magnetic resonance image and each of the three-dimensional magnetic resonance template images have a first resolution, the transverse magnetic resonance image, the sagittal magnetic resonance image and the coronal magnetic resonance image have a second resolution, and the first resolution is smaller than the second resolution; The second registration parameters include a registration matrix formed by three registration translation parameters and three registration rotation parameters, and the standard positioning parameters include a standard positioning matrix formed by three standard translation parameters and three standard rotation parameters; Then in step S3, the registration matrix and the standard positioning matrix are multiplied to obtain an automatic positioning matrix, and three positioning translation parameters and three positioning rotation parameters are extracted from the automatic positioning matrix as the automatic positioning parameters.

2. The automatic positioning method for magnetic resonance imaging according to claim 1, wherein The step S2 comprises: Step S21a, for each of the data sets, respectively calculating the similarity between the registration image and the magnetic resonance three-dimensional template image; Step S22a: extracting the first registration parameter in the data set corresponding to the maximum value of each similarity, and using the first registration parameter as the second registration parameter.

3. The automatic positioning method for magnetic resonance imaging according to claim 1, characterized in that, The step S2 comprises: Step S21b, for each of the data sets, respectively calculating the similarity between the registered image and the magnetic resonance three-dimensional template image, and arranging the similarities in descending order to obtain a similarity sequence; Step S22b, sequentially extracting the first registration parameters in the data set corresponding to a preset number of the similarities ranked top from the similarity sequence, and performing weighted averaging processing on each of the first registration parameters to obtain the second registration parameters.

4. The automatic positioning method for magnetic resonance imaging according to claim 1, wherein Each of the magnetic resonance three-dimensional template images is associated with corresponding first basic information of the subject to be scanned, and the magnetic resonance three-dimensional images obtained by magnetic resonance scanning are associated with corresponding second basic information of the subject to be scanned; Then, in the step S1, it further includes extracting each of the magnetic resonance three-dimensional template images corresponding to the second basic information that matches the first basic information, and then performing image registration on the magnetic resonance three-dimensional images and the extracted magnetic resonance three-dimensional template images respectively to obtain the registered images and the corresponding first registration parameters.

5. The automatic positioning method for magnetic resonance imaging according to claim 4, wherein, The first basic information and the second basic information include the part to be imaged of the subject to be scanned, and / or the age group.

6. The automatic positioning method for magnetic resonance imaging according to claim 1, characterized in that, In the step S1, the image registration method is an affine transformation method, or a rigid body transformation method, or a non-linear transformation method.

7. An automatic positioning system for magnetic resonance imaging based on image registration, characterized in that, Applying the magnetic resonance imaging automatic positioning method according to any one of claims 1-6, the magnetic resonance imaging automatic positioning system includes: An image registration module, configured to perform image registration on the magnetic resonance three-dimensional images of a part to be imaged obtained by magnetic resonance scanning and a plurality of pre-generated magnetic resonance three-dimensional template images respectively, obtain the registered images and the corresponding first registration parameters, and add each of the registered images, the corresponding magnetic resonance three-dimensional template images, and the first registration parameters into a data set respectively; An image processing module, connected to the image registration module, configured to calculate the similarity between the registered image and the magnetic resonance three-dimensional template image for each data set respectively, and process the obtained second registration parameter according to each similarity and the first registration parameter in the corresponding data set; An automatic positioning module, connected to the image processing module, configured to process the second registration parameter and a standard positioning parameter associated with the part to be imaged obtained in advance to obtain an automatic positioning parameter, which is used as the magnetic resonance imaging automatic positioning result of the current magnetic resonance scanning.

Images