Image registration method, computer device and storage medium

Abstract

The application relates to an image registration method, a computer device and a storage medium. The method comprises the following steps: determining a registration framework which is pre-constructed and comprises at least a combined registration mode; the combined registration mode comprises a combination of a linear registration mode and a nonlinear registration mode, and the linear registration mode and the nonlinear registration mode are both suitable for image registration in different scenes; obtaining a reference image to be registered and a floating image to be registered; and calling a target registration mode in the registration framework to register the reference image and the floating image. The method can improve the application range of the registration method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an image registration method, a computer device and a storage medium. BACKGROUND

[0002] Image registration refers to the process of aligning, matching and superimposing two or more images acquired under different scenes. Through image registration, multiple images under different scenes can be better observed and analyzed to obtain more accurate analysis results.

[0003] At present, when images under different scenes are registered, an image registration method belonging to the scene is usually constructed in real time according to the actual scene each time, for example, a transformation matrix or a deformation field under the scene is constructed, so that the images under the scene are registered through the image registration method constructed in real time.

[0004] However, the above-mentioned technology needs to re-construct the image registration method each time the scene changes, and its scope of application is limited. SUMMARY

[0005] Therefore, it is necessary to provide an image registration method, a computer device and a storage medium capable of improving the scope of application of the registration method to solve the above technical problems.

[0006] In a first aspect, the present application provides an image registration method, which comprises:

[0007] determining a registration framework pre-constructed at least including a combined registration method; the combined registration method includes a combination of a linear registration method and a nonlinear registration method, and the linear registration method and the nonlinear registration method are both applicable to registering images under different scenes;

[0008] acquiring a reference image to be registered and a floating image to be registered;

[0009] calling a target registration method in the registration framework to register the reference image and the floating image.

[0010] In one of the embodiments, the determination of the registration framework pre-constructed at least including the combined registration method comprises:

[0011] constructing the linear registration method by using registration parameters under different scenes, and constructing the nonlinear registration method by using registration parameters under different scenes; the registration parameters include a degree of freedom parameter and a preset hyperparameter required in the registration process;

[0012] concatenating the linear registration method and the nonlinear registration method to obtain the combined registration method;

[0013] The registration framework is determined according to a linear registration mode, a nonlinear registration mode, and a combined registration mode.

[0014] In one embodiment, the linear registration mode includes at least one linear registration module, and each linear registration module includes at least one linear registration step.

[0015] Each linear registration module is configured to perform image-level registration on the reference image and the floating image, and each linear registration step is configured to perform linear optimization on a degree-of-freedom parameter between the reference image and the floating image.

[0016] In one embodiment, each linear registration module corresponds to image registration at a resampled image resolution, and each linear registration step corresponds to optimization of one or more target degree-of-freedom parameters.

[0017] In one embodiment, the linear registration modules are in a first cascade relationship, the linear registration steps are in a second cascade relationship, and the target degree-of-freedom parameters optimized by the linear registration steps are not all the same.

[0018] The first cascade relationship is configured to indicate that, in two adjacent linear registration modules, an output of a previous linear registration module is used as an input of a next linear registration module.

[0019] The second cascade relationship is configured to indicate that, in two adjacent linear registration steps, an output of a previous linear registration step is used as an input of a next linear registration step.

[0020] In one embodiment, the nonlinear registration mode includes at least one nonlinear registration module, and each nonlinear registration module includes at least one nonlinear registration step.

[0021] Different nonlinear registration modules correspond to image registration at different resampled image resolutions, and a resampled image resolution corresponding to a previous nonlinear registration module is smaller than a resampled image resolution corresponding to a next nonlinear registration module in two adjacent nonlinear registration modules.

[0022] In one embodiment, the target registration mode is a combined registration mode, and the target registration mode in the registration framework is called to perform registration on the reference image and the floating image, including:

[0023] An initial conversion parameter corresponding to the reference image and the floating image is determined by using a preset parameter determination mode, and the parameter determination mode includes at least one of a coordinate axis conversion mode, a centroid alignment mode, and a slice alignment mode in a body height direction.

[0024] The initial conversion parameter, the reference image and the floating image are sequentially input into the combined registration mode to perform linear registration and nonlinear registration.

[0025] In one of the embodiments, after the initial conversion parameter is determined according to the reference image and the floating image, the method further comprises:

[0026] The reference image and the floating image are subjected to different times of resampling processing to determine a reference image after each time of resampling and a floating image after each time of resampling; the reference image after each time of resampling has different image resolutions;

[0027] The reference image after each time of resampling and the floating image after each time of resampling are respectively subjected to image preprocessing;

[0028] The reference image after the image preprocessing and the floating image after the image preprocessing are sequentially input into the combined registration mode as respective reference images and floating images for linear registration and nonlinear registration.

[0029] In a second aspect, the present application further provides an image registration device, which comprises:

[0030] a framework determination module configured to determine a registration framework pre-constructed and comprising at least a combined registration mode; the combined registration mode comprises a combination of a linear registration mode and a nonlinear registration mode, and the linear registration mode and the nonlinear registration mode are both applicable to registration of images in different scenarios;

[0031] an image acquisition module configured to acquire a reference image to be registered and a floating image to be registered;

[0032] a registration module configured to call a target registration mode in the registration framework to register the reference image and the floating image.

[0033] In a third aspect, the present application further provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0034] determine a registration framework pre-constructed and comprising at least a combined registration mode; the combined registration mode comprises a combination of a linear registration mode and a nonlinear registration mode, and the linear registration mode and the nonlinear registration mode are both applicable to registration of images in different scenarios;

[0035] acquire a reference image to be registered and a floating image to be registered;

[0036] call a target registration mode in the registration framework to register the reference image and the floating image.

[0037] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0038] determining a registration framework comprising at least a combined registration method, wherein the combined registration method comprises a combination of a linear registration method and a nonlinear registration method, and the linear registration method and the nonlinear registration method are both applicable to image registration in different scenarios;

[0039] obtaining a reference image to be registered and a floating image to be registered;

[0040] calling a target registration method in the registration framework to register the reference image and the floating image.

[0041] In a fifth aspect, the present application provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the following steps:

[0042] determining a registration framework comprising at least a combined registration method, wherein the combined registration method comprises a combination of a linear registration method and a nonlinear registration method, and the linear registration method and the nonlinear registration method are both applicable to image registration in different scenarios;

[0043] obtaining a reference image to be registered and a floating image to be registered;

[0044] calling a target registration method in the registration framework to register the reference image and the floating image.

[0045] The image registration method, the computer device and the storage medium, by determining a registration framework comprising at least a combined registration method comprising a linear registration method and a nonlinear registration method, and after obtaining a reference image to be registered and a floating image to be registered, calling the registration framework to register the reference image and the floating image, wherein the linear registration method and the nonlinear registration method are both applicable to image registration in different scenarios, so that the registration framework can be used for any registration scenario, and it is not necessary to set the registration method according to different scenarios, so that the application range is wider and the universality is stronger, and the cost caused by repeatedly constructing the registration method can be saved. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 which is an internal structure diagram of the computer device in one embodiment;

[0047] Figure 2 which is a flowchart of the image registration method in one embodiment;

[0048] Figure 3This is a flowchart illustrating the image registration method in another embodiment;

[0049] Figure 4 This is a flowchart illustrating linear and nonlinear registration in another embodiment;

[0050] Figure 5 This is a flowchart illustrating the image registration method in another embodiment;

[0051] Figure 6 This is a flowchart illustrating the image registration method in another embodiment;

[0052] Figure 7 This is an example diagram of initial image registration in another embodiment;

[0053] Figure 8 This is an example diagram illustrating linear registration of images at different resolutions in another embodiment;

[0054] Figure 9 This is an example diagram illustrating nonlinear registration of images at different resolutions in another embodiment;

[0055] Figure 10 Example diagrams illustrating linear and nonlinear registration of images in another embodiment;

[0056] Figure 11 This is a structural block diagram of an image registration device in one embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0058] The image registration method provided in this application can be applied to computer devices, which can be terminals or servers. Taking a server as an example, its internal structure diagram can be as follows: Figure 1 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores pre-built registration frameworks and other data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an image registration method.

[0059] Those skilled in the art can understand that Figure 1 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0060] In one embodiment, as Figure 2 shown, an image registration method is provided, which is applied to Figure 1 a computer device as an example, and can include the following steps:

[0061] S202, determining a registration framework pre-constructed and including at least a combined registration mode; the combined registration mode includes a combination of a linear registration mode and a nonlinear registration mode, and the linear registration mode and the nonlinear registration mode are both applicable to registration of images in different scenarios.

[0062] In this step, image registration refers to the process of aligning, matching and superimposing two or more images obtained under different times, different sensors or different conditions. The different sensors can be different imaging devices, and the different conditions can include different climates, different illuminations, different camera positions and angles, etc.

[0063] Taking medical images as an example, the different scenarios herein not only include the above different times, different sensors, different conditions, etc., but also include different parts, different modalities, different resolutions, different degrees of freedom, different similarities, different pre-processing, different interpolation methods, etc.

[0064] In this step, all parameters involved in the different scenarios can be pre-collected, and then the linear registration mode and the nonlinear registration mode can be obtained by optimizing or registering all parameters involved in the different scenarios. In this way, the obtained linear registration mode and nonlinear registration mode can both be used to register images in different scenarios. In addition, the obtained linear registration mode can be a conversion parameter in linear registration, such as a conversion matrix; and the nonlinear registration mode can be a conversion parameter in nonlinear registration, such as a deformation field.

[0065] The linear registration manner and the nonlinear registration manner obtained above can be a combined registration manner, or the linear registration manner and the nonlinear registration manner can be obtained separately and then combined to obtain the combined registration manner; after the combined registration manner is obtained, the combined registration manner can be directly used as the registration framework, or the combined registration manner can be combined with other registration manners to form the registration framework; in any case, the combined registration manner including both the linear registration manner and the nonlinear registration manner can be obtained in the registration framework.

[0066] It should be noted that after the linear registration manner and the nonlinear registration manner are combined or bound, the images to be registered need to be linearly registered by the linear registration manner and nonlinearly registered by the nonlinear registration manner when the combined registration manner is used for image registration, and the registration result is obtained; the order of the linear registration manner and the nonlinear registration manner is not limited here.

[0067] S204, obtaining a reference image to be registered and a floating image to be registered.

[0068] In this step, the reference image to be registered and the floating image to be registered are generally images of the same part of the same object to be measured, but the reference image and the floating image can be obtained by scanning the same object to be measured in different scenes, for example, the reference image is obtained at a previous time and the floating image is obtained at a current time.

[0069] For the obtaining manner of the reference image and the floating image, the same object to be measured can be scanned in different scenes, and the reference image and the floating image can be obtained by image reconstruction of the scanned data; or the reference image and the floating image of the same object to be measured in different scenes can be pre-stored in the cloud or a server, and can be read from the cloud or the server when needed; or other obtaining manners can also be used, which are not limited here.

[0070] S206, calling a target registration manner in the registration framework to register the reference image and the floating image.

[0071] In this step, after the pre-constructed registration framework is determined and the reference image to be registered and the floating image to be registered are obtained, an actual registration requirement input by a user can also be obtained, and the actual registration requirement is used to represent the target registration manner selected by the user in the registration framework.

[0072] After obtaining the actual registration requirement of the user, the determined registration framework can be called first, and the target registration mode required by the user is selected from the registration framework, and then the target registration mode is used to register the reference image and the floating image to be registered to obtain the registration result. In this way, the registration mode of multiple images under the same registration framework can be realized, and it is not necessary to set different registration modes according to different scenes, and the actual registration requirement of the user can be met, so that the application range of the registration method can be improved and the user experience can be improved.

[0073] In the above image registration method, the registration framework of the combination registration mode composed of the linear registration mode and the nonlinear registration mode is determined in advance, and after the reference image and the floating image to be registered are obtained, the registration framework is called to register the reference image and the floating image. The linear registration mode and the nonlinear registration mode are both applicable to image registration in different scenes, so that the pre-constructed registration framework can be used for any registration scene, and it is not necessary to set the registration mode according to different scenes, so that the application range is wider and the universality is stronger. At the same time, the cost caused by repeatedly constructing the registration mode can be saved.

[0074] In the above embodiment, the determination of the pre-constructed registration framework is mentioned, and the following embodiment describes a possible implementation of the determination process.

[0075] In another embodiment, another image registration method is provided, which is based on the above embodiment, as shown in Figure 3 The S202 can include the following steps:

[0076] S302, a linear registration mode is constructed by using registration parameters in different scenes, and a nonlinear registration mode is constructed by using registration parameters in different scenes; the registration parameters include a degree of freedom parameter and a preset hyperparameter required in the registration process.

[0077] Among them, all the registration parameters involved in different scenes can be obtained in advance, and the all registration parameters include the degree of freedom parameter and the preset hyperparameter required in the registration process; the degree of freedom parameter mainly includes rotation, translation, scaling and other parameters; the preset hyperparameter required in the registration process mainly includes translation range and step, rotation angle range and step, scaling ratio range and step, similarity scale and other parameters.

[0078] In this step, after obtaining all the degrees of freedom parameters involved in different scenes and the preset hyperparameters required in the registration process, the linear registration method can be obtained by optimizing or registering all the degrees of freedom parameters involved in different scenes and the preset hyperparameters required in the registration process through linear registration; at the same time, the nonlinear registration method can be obtained by optimizing or registering all the degrees of freedom parameters involved in different scenes and the preset hyperparameters required in the registration process through nonlinear registration.

[0079] S304, the linear registration method and the nonlinear registration method are cascaded to obtain a combined registration method.

[0080] In this step, after obtaining the linear registration method and the nonlinear registration method, the reference image and the floating image to be registered can be taken as the input of the linear registration method to perform linear registration, and the output of the linear registration method can be taken as the input of the nonlinear registration method to perform nonlinear registration.

[0081] By the above-mentioned manner, the cascade of the linear registration method and the nonlinear registration method can be realized to obtain a combined registration method.

[0082] S306, according to the linear registration method, the nonlinear registration method and the combined registration method, a registration framework is determined.

[0083] In this step, after obtaining the linear registration method, the nonlinear registration method and the combined registration method, the linear registration method, the nonlinear registration method and the combined registration method can be taken as a kind of registration method respectively to obtain three different registration methods, and the three different registration methods can be taken as the registration framework.

[0084] In this embodiment, the linear registration method and the nonlinear registration method are constructed by using the degrees of freedom parameters and the required preset hyperparameters in different scenes, and the registration framework is obtained by cascading the two methods and the respective registration methods of the two methods. By setting different registration methods in the registration framework, subsequent images can be registered by different registration methods, and more registration method options can be provided for users, further improving the application range of the registration method and the user experience.

[0085] In the above-mentioned embodiments, it is mentioned that the linear registration method and the nonlinear registration method can register images in different scenes, and can optimize and register different degrees of freedom parameters and required preset hyperparameters. The following embodiments will respectively explain the registration details of the linear registration method and the nonlinear registration method.

[0086] Firstly, the registration details in the linear registration mode are described. In another embodiment, the linear registration mode includes at least one linear registration module, each linear registration module includes at least one linear registration step; each linear registration module is used for image-level registration of the reference image and the floating image, and each linear registration step is used for linear optimization of the degree of freedom parameters between the reference image and the floating image.

[0087] In the linear registration mode, one or more linear registration modules can be included, denoted as Block, and generally a plurality of linear registration modules Block can be included, each linear registration module Block can include one or more linear registration steps, denoted as Step; the preset parameters used by the linear registration steps Step included in each linear registration module Block can all be different, all be the same, or part be the same and part be different.

[0088] Each linear registration module Block is used for registration of the reference image and the floating image at the image level. As an optional embodiment, each linear registration module corresponds to image registration at a resampled image resolution. The linear registration step Step is used for linear optimization of the degree of freedom parameters between the reference image and the floating image. As an optional embodiment, each linear registration step corresponds to optimization of one or more target degree of freedom parameters.

[0089] For the plurality of linear registration modules Block and the plurality of linear registration steps Step, as an optional embodiment, the linear registration modules are in a first cascading relationship, the linear registration steps are in a second cascading relationship, and the target degree of freedom parameters optimized by the linear registration steps are not all the same; the first cascading relationship is used to represent that in the adjacent two linear registration modules, the output of the previous linear registration module is used as the input of the next linear registration module; and the second cascading relationship is used to represent that in the adjacent two linear registration steps, the output of the previous linear registration step is used as the input of the next linear registration step.

[0090] It should be noted that each linear registration module can preset the hyperparameters of the module, and each linear registration step can also preset the hyperparameters of the step. The hyperparameters preset by each step cover the same hyperparameters preset by the module in which the step is located. Each linear registration module optimizes and updates the degree of freedom parameters, and only when the similarity between the floating image transformed using the updated degree of freedom parameters and the reference image increases. Each linear registration step optimizes and updates the degree of freedom parameters, and only when the similarity between the floating image transformed using the updated degree of freedom parameters and the reference image increases.

[0091] For example, a plurality of linear registration modules Block are taken as an example, and the resampling image resolutions of the linear registration modules Block 1 to Block n are arranged in turn from low to high, that is, in adjacent two linear registration modules, the resampling image resolution corresponding to the previous linear registration module is less than the resampling image resolution corresponding to the next linear registration module. The highest image resolution can be the same as the resolution of the reference image or the floating image to be registered, n is a positive integer greater than or equal to 1, and the number of specific linear registration modules can be set according to actual conditions, for example, n can be 3, 4, 5, etc. That is, in the linear registration manner, different images with different resampling image resolutions can be registered respectively to speed up the registration accuracy and speed.

[0092] For example, referring to FIG. 1, Figure 4 As shown, the linear registration manner can be sequentially composed of a plurality of linear registration modules Block, such as Block 1, Block 2, …, Block n-1, Block n, which can be recorded as B1, B2, …, Bn-1, Bn for convenience of description, each linear registration module Block corresponding to a resampling image resolution; in actual registration, the reference image and the floating image to be registered and the conversion matrix T0 output by the previous registration manner can be input into B1 to obtain T 1,1 , T 1,1 is input into B2 to obtain T 1,2 , and so on, T 1,n-1 is input into Bn to obtain T 1,n .

[0093] In addition, each linear registration module Block is also composed of a plurality of linear registration steps Step in a set order, recorded as S1, S2, …, Sn, each linear registration step Step corresponding to an optimization method, such as ShiftOptim (translation parameter optimization), AngleOptim (angle parameter optimization), ScaleOptim (scale parameter optimization), SkewOptim (skew parameter optimization), GradientDscent (gradient descent parameter optimization), MassCenterOptim (mass center optimization), BodyPartOptim (body part optimization), etc.

[0094] For example, the linear registration module B1 includes two linear registration steps Step, S1 and S2, respectively, wherein S1 can optimize the translation parameter in the degree of freedom parameter, and S2 can optimize the angle parameter in the rotation parameter in the degree of freedom parameter; B2 includes one linear registration step S1, which can optimize the angle parameter in the rotation parameter in the degree of freedom parameter; the subsequent other linear registration modules Block can also include one or more linear registration steps, and sequentially optimize one or more target degree of freedom parameters in the degree of freedom parameter.

[0095] For example, T is obtained by inputting T0 into B1 1,1 For example, assume that B1 includes S 11 , S 12 ,..., and S 1n , wherein each Step can be one or more parameters in the degree of freedom parameter DOF; for example, T0 is input into S 11 to obtain T 1,1,1 , T 1,1,1 is input into S 12 to obtain T 1,1,2 ,..., T 1,1,n-1 is input into S 1,n to obtain T 1,1,n , and T 1,1,n is T 1,1 . For the subscript in T x,y,z , x, y, and z respectively represent the conversion matrix T obtained by the xth registration mode (for example, the linear registration mode), the yth linear registration module, and the zth linear registration step.

[0096] In this embodiment, by setting at least one linear registration module in the linear registration mode and at least one linear registration step in each linear registration module, linear optimization of the image level and the conversion parameter between images can be performed on the image, so that the obtained registration result is more accurate, that is, the accuracy of the registration result can be improved.

[0097] Secondly, the registration details in the nonlinear registration mode are described. In another embodiment, the above-mentioned nonlinear registration mode includes at least one nonlinear registration module, and each nonlinear registration module includes at least one nonlinear registration step;

[0098] Different nonlinear registration modules correspond to image registration under different resampled image resolutions, and the resampled image resolution corresponding to the previous nonlinear registration module in the adjacent two nonlinear registration modules is less than the resampled image resolution corresponding to the next nonlinear registration module.

[0099] Similar to linear registration, nonlinear registration can also include one or more nonlinear registration modules, denoted as Block. Generally, there can be multiple nonlinear registration modules Block. Each nonlinear registration module Block can include one or more nonlinear registration steps, denoted as Step. The nonlinear registration steps Step included in each nonlinear registration module Block can be all different, all the same, or partially the same and partially different.

[0100] For multiple nonlinear registration modules (Block) and multiple nonlinear registration steps (Step), as an optional embodiment, the nonlinear registration modules are in a first cascade relationship, and the nonlinear registration steps are in a second cascade relationship, and the target degree of freedom parameters for linear optimization of each nonlinear registration step are not completely the same; the above-mentioned first cascade relationship is used to characterize that in two adjacent nonlinear registration modules, the output of the previous nonlinear registration module is used as the input of the next nonlinear registration module.

[0101] Taking multiple nonlinear registration modules (Blocks) as an example, the resampled image resolutions of Blocks 1 to Block n are arranged from low to high. The highest image resolution can be the same as the resolution of the reference image or floating image to be registered. n is a positive integer greater than or equal to 1. The specific number of nonlinear registration modules can be set according to the actual situation. For example, n can be 3, 4, 5, etc. In other words, under this nonlinear registration method, images with different resampled image resolutions can be registered separately to speed up the registration accuracy and speed.

[0102] For example, see [link to example]. Figure 4 As shown, the nonlinear registration method can be composed of multiple nonlinear registration modules (Blocks) arranged sequentially, denoted as B1, B2, ..., B... n-1 B n Each nonlinear registration module (Block) corresponds to a resampled image resolution. During actual registration, the reference image to be registered, the floating image, and the transformation matrix T1 output from the previous registration method can be input into B1 to obtain T. 2,1 T 2,1 Input B2 to get T 2,2 And so on, T 2,n-1 Input Bn to get T 2,n , among which, T 2,n This is T2, which is the conversion parameter output by the nonlinear registration method.

[0103] For each non-linear registration step in the non-linear registration module Block, the degree of freedom of each non-linear registration step is for each pixel, so the number of degree of freedom parameters is 3 times the number of pixels, including the above-mentioned x, y, z three directions. For each non-linear registration step in the non-linear registration module Block, it can be similar to the linear registration step in the linear registration module of the above-mentioned linear registration method, which will not be repeated here.

[0104] In this embodiment, by setting at least one non-linear registration module in the non-linear registration method and at least one non-linear registration step in each non-linear registration module, the non-linear optimization of the image level and the deformation field between the images can be performed, so that the obtained registration result is more accurate, that is, the accuracy of the registration result can be improved.

[0105] The following embodiments mainly explain how to perform the registration process of the reference image and the floating image to be registered when the target registration method is a combined registration method.

[0106] In another embodiment, another image registration method is provided, which is based on the above-mentioned embodiments, as shown in Figure 5 S206 can include the following steps:

[0107] S402, determine the initial conversion parameter corresponding to the reference image and the floating image by using a preset parameter determination method; the parameter determination method includes at least one of the coordinate axis conversion method, the centroid alignment method, and the slice alignment method in the height direction of the human body.

[0108] In this step, after obtaining the reference image and the floating image to be registered, any one of the coordinate axis conversion method, the centroid alignment method, and the slice alignment method in the height direction of the human body can be used, and of course the key point registration method can also be used. The coordinate axis conversion or centroid alignment or slice alignment processing of the reference image and the floating image is performed to obtain the initial conversion parameter between the reference image and the floating image; the initial conversion parameter can be an initial conversion matrix, which can be denoted as T0.

[0109] Of course, any two or three of the above-mentioned three methods can also be performed on the reference image and the floating image, and then the conversion parameters obtained respectively can be combined to obtain the initial conversion parameter. Of course, the above-mentioned initial conversion parameter can also be pre-calculated and directly called when used.

[0110] S404, input the above-mentioned initial conversion parameter, the reference image, and the floating image into the combined registration method in sequence for linear registration and non-linear registration.

[0111] In this step, after obtaining the initial conversion parameter, assuming that the combination manner in the combination registration manner is the linear registration manner first and then the nonlinear registration manner, the initial conversion parameter and the reference image and the floating image can be input into the linear registration manner first to perform registration, to obtain the linear conversion parameter between the reference image and the floating image; then, the reference image and the floating image and the linear conversion parameter can be input into the nonlinear registration manner to perform registration, to obtain the nonlinear conversion parameter between the reference image and the floating image, and the registration between the reference image and the floating image is completed.

[0112] In this embodiment, the initial conversion parameter between the reference image and the floating image is determined by using multiple conversion or alignment manners, and the reference image and the floating image are registered by using the initial conversion parameter and the linear registration manner and the nonlinear registration manner. Thus, by using multiple registration manners, the final registration result can be more accurate.

[0113] The following embodiments mainly explain the process that after the initial conversion parameter between the reference image and the floating image is determined, the image needs to be further processed before being input into the subsequent linear or nonlinear registration manner to perform registration.

[0114] In another embodiment, another image registration method is provided, which is based on the above-mentioned embodiments, as shown in the following steps. Figure 6 The above-mentioned method can further include the following steps:

[0115] S502, the reference image and the floating image are subjected to different times of resampling processing, to determine the reference image after each time of resampling and the floating image after each time of resampling; the image resolution of the reference image after each time of resampling is different.

[0116] In this step, the number of linear registration modules set in the linear registration manner and the number of nonlinear registration modules set in the nonlinear registration manner can be used to correspondingly set the number of times of resampling required here, and the number of times of resampling can be mainly set according to the larger number. For example, assuming that the linear registration manner includes 4 linear registration modules and the nonlinear registration manner includes 3 nonlinear registration modules, the image can be resampled 4 times here.

[0117] When the reference image and the floating image are resampled, the image resolution of the reference image and the floating image is generally equal and belongs to a relatively high image resolution. Here, the resampling is mainly down-sampling processing, i.e., the reference image and the floating image are subjected to different orders of down-sampling processing. For example, assuming that the image resolution of the reference image is 256*256, and the image resolution after 1, 2, 3, and 4 times of down-sampling processing is 128*128, 64*64, 32*32, and 16*16, respectively.

[0118] By the above different order downsampling processing of the reference image and the floating image respectively, the reference image after different order downsampling and the floating image after different order downsampling can be obtained. It should be noted that the order of the reference image and the floating image is the same, for example, 4 times of downsampling processing is performed.

[0119] It should be noted that the main purpose of the downsampling in this embodiment is to optimize the parameters on the low resolution, which can accelerate the optimization process, and the optimization results of the low resolution are gradually uploaded to the high resolution as the initial value for further fine optimization, thereby improving the accuracy of the optimization results.

[0120] S504, image preprocessing is performed on each resampled reference image and each resampled floating image.

[0121] S506, the reference image after image preprocessing and the floating image after image preprocessing are input into the combined registration mode in turn as the reference image and the floating image of linear registration and nonlinear registration respectively.

[0122] The image preprocessing can include adjusting window width and window position, histogram equalization, pixel normalization, organ enhancement processing, smoothing processing, etc. The organ enhancement processing refers to segmenting the eye mask image mask of the specified organ, and then pasting the mask to the original image to realize the organ enhancement processing.

[0123] Specifically, after obtaining each order of the reference image after downsampling processing and the corresponding floating image, image preprocessing can be performed on the group of reference images and floating images, and then the group of reference images and floating images after image preprocessing are input into the corresponding linear registration module or nonlinear registration module for registration.

[0124] In addition, as mentioned in the above embodiment, the resolution of the group of reference images and floating images input by each linear registration module is the same, but different from the resolution of the group of images input by other linear registration modules; the nonlinear registration module is similar.

[0125] In this embodiment, by resampling the reference image and the floating image at different resolutions, and performing image preprocessing after resampling, the images after image preprocessing can be input into the corresponding linear registration module or nonlinear registration module for registration. By resampling images at different resolutions for registration, the speed of registration can be accelerated and the efficiency of registration can be improved. At the same time, through image processing, the result of registration can be more accurate, and the accuracy of the registration result can be improved.

[0126] A specific embodiment of linear registration and nonlinear registration of images is given below to illustrate the technical solutions of the present application. On the basis of the above embodiment, the method can comprise the following steps:

[0127] S1, constructing a linear registration mode using registration parameters in different scenarios, and constructing a nonlinear registration mode using registration parameters in different scenarios; the registration parameters comprise a degree-of-freedom parameter and a preset hyperparameter required in the registration process;

[0128] S2, cascading the linear registration mode and the nonlinear registration mode to obtain a combined registration mode;

[0129] S3, determining a registration framework according to the linear registration mode, the nonlinear registration mode, and the combined registration mode;

[0130] S4, obtaining a reference image to be registered and a floating image to be registered;

[0131] S5, determining the reference image and the floating image using a preset parameter determination mode to obtain initial conversion parameters between the reference image and the floating image; the parameter determination mode comprises at least one of a coordinate axis conversion mode, a centroid alignment mode, and a slice alignment mode in the direction of human body height;

[0132] S6, performing different times of resampling processing on the reference image and the floating image to determine a reference image after each time of resampling and a floating image after each time of resampling; the image resolutions of the reference images after each time of resampling are different;

[0133] S7, performing image preprocessing on the reference image after each time of resampling and the floating image after each time of resampling, respectively;

[0134] S8, inputting the initial conversion parameters, the reference image after image preprocessing, and the floating image into the linear registration mode to perform linear registration and obtain linear conversion parameters; the linear registration mode comprises at least one linear registration module, each linear registration module comprises at least one linear registration step; each linear registration module is used for image-level registration of the reference image and the floating image, and each linear registration step is used for linear optimization of a degree-of-freedom parameter between the reference image and the floating image; the linear registration modules have a first cascading relationship, and the linear registration steps have a second cascading relationship, and the target degree-of-freedom parameters for linear optimization of the linear registration steps are not completely the same; the first cascading relationship is used to represent that, in adjacent two linear registration modules, the output of a previous linear registration module is used as the input of a next linear registration module;

[0135] S9, inputting the linear conversion parameter, the pre-processed reference image and the floating image into a non-linear registration mode to perform linear registration, and obtaining a non-linear conversion parameter; the non-linear registration mode comprises at least one non-linear registration module, each non-linear registration module comprises at least one non-linear registration step; different non-linear registration modules correspond to image registration at different resampling image resolutions, and the resampling image resolution corresponding to a previous non-linear registration module is less than the resampling image resolution corresponding to a next non-linear registration module.

[0136] Further, for the initial conversion parameter obtained in S5 in the above embodiment, for example, the Z-axis alignment mode is adopted, then the registration result of the reference image and the floating image can be seen from Figure 7 , wherein the three images are the reference image to be registered, the floating image to be registered and the floating image after Z-axis alignment (initial registration) in sequence. As can be seen from the figure, the position of the tissue in the registered floating image is consistent with the position of the same tissue in the reference image, so that accurate image analysis can be facilitated subsequently.

[0137] For the linear registration of the image in S8 in the above embodiment, see Figure 8 , linearly registered images at different resolutions are given, the first row of images are reference images at different resolutions, the second row of images are floating images at different resolutions, and the resolutions of the reference image and the floating image in each column are the same. As can be seen from the figure, the position of the tissue in the registered floating image is consistent with the position of the same tissue in the reference image, so that accurate image analysis can be facilitated subsequently.

[0138] For the non-linear registration of the image in S9 in the above embodiment, see Figure 9 , non-linearly registered images at different resolutions are given, the first row of images are reference images at different resolutions, the second row of images are floating images at different resolutions, and the resolutions of the reference image and the floating image in each column are the same. As can be seen from the figure, the position of the tissue in the registered floating image is consistent with the position of the same tissue in the reference image, so that accurate image analysis can be facilitated subsequently.

[0139] For the result of the non-linear registration of the reference image and the floating image after linear registration, see Figure 10 , wherein the three images are the reference image to be registered, the floating image after linear registration and the floating image after non-linear registration in sequence. As can be seen from the figure, after linear registration and non-linear registration, the position of the tissue in the registered floating image is consistent with the position of the same tissue in the reference image, and the resolution of the floating image is higher than or equal to the resolution of the reference image, so that accurate image analysis can be facilitated subsequently.

[0140] It should be understood that, although each step in the flowchart involved in the embodiments described above is shown in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately executed with at least part of other steps or steps or stages in other steps.

[0141] Based on the same inventive concept, the embodiments of the present application also provide an image registration device for implementing the above-mentioned image registration method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more image registration device embodiments provided below can refer to the limitations of the image registration method in the above text, which will not be repeated here.

[0142] In one embodiment, as shown in Figure 11 An image registration device is provided, comprising: a frame determination module, an image acquisition module, and a registration module, wherein:

[0143] The frame determination module is configured to determine a registration framework pre-constructed to include at least a combined registration mode. The combined registration mode includes a combination of a linear registration mode and a nonlinear registration mode, and both the linear registration mode and the nonlinear registration mode are applicable to registering images in different scenarios.

[0144] The image acquisition module is configured to acquire a reference image to be registered and a floating image to be registered.

[0145] The registration module is configured to call a target registration mode in the registration framework to register the reference image and the floating image.

[0146] In another embodiment, another image registration device is provided, and on the basis of the above embodiment, the frame determination module can include:

[0147] The construction unit is configured to construct a linear registration mode using registration parameters in different scenarios, and construct a nonlinear registration mode using registration parameters in different scenarios. The registration parameters include a degree of freedom parameter and a preset hyperparameter required in a registration process.

[0148] a cascading unit configured to cascade the linear registration manner and the nonlinear registration manner to obtain a combined registration manner;

[0149] a frame determination unit configured to determine a registration frame according to the linear registration manner, the nonlinear registration manner, and the combined registration manner.

[0150] In another embodiment, the linear registration manner includes at least one linear registration module, each of which includes at least one linear registration step; each linear registration module is configured to perform image-level registration on the reference image and the floating image, and each linear registration step is configured to perform linear optimization on a degree-of-freedom parameter between the reference image and the floating image.

[0151] Optionally, each linear registration module corresponds to image registration at a resampled image resolution, and each linear registration step corresponds to optimization of one or more target degree-of-freedom parameters.

[0152] Optionally, the linear registration modules are in a first cascading relationship, the linear registration steps are in a second cascading relationship, and the target degree-of-freedom parameters optimized by the linear registration steps are not all the same; the first cascading relationship is configured to represent that, in two adjacent linear registration modules, the output of a previous linear registration module is input to a next linear registration module; and the second cascading relationship is configured to represent that, in two adjacent linear registration steps, the output of a previous linear registration step is input to a next linear registration step.

[0153] In another embodiment, the nonlinear registration manner includes at least one nonlinear registration module, each of which includes at least one nonlinear registration step; different nonlinear registration modules correspond to image registration at different resampled image resolutions, and the resampled image resolution corresponding to a previous nonlinear registration module is smaller than the resampled image resolution corresponding to a next nonlinear registration module in two adjacent nonlinear registration modules.

[0154] In another embodiment, another image registration apparatus is provided, which is based on the above-mentioned embodiments, and the target registration manner is a combined registration manner, and the registration module can include:

[0155] an initial registration unit configured to determine initial conversion parameters corresponding to the reference image and the floating image by using a preset parameter determination manner; the parameter determination manner includes at least one of a coordinate axis conversion manner, a centroid alignment manner, and a slice alignment manner in a body height direction;

[0156] a combined registration unit configured to input the initial conversion parameters, the reference image, and the floating image into the combined registration manner in sequence to perform linear registration and nonlinear registration.

[0157] In another embodiment, another image registration device is provided, which can further include the following based on the above-mentioned embodiment:

[0158] a resampling module configured to perform different times of resampling on the reference image and the floating image, to determine a reference image after each time of resampling and a floating image after each time of resampling, and the reference image after each time of resampling has different image resolutions;

[0159] a preprocessing module configured to perform image preprocessing on the reference image after each time of resampling and the floating image after each time of resampling, respectively;

[0160] an input module configured to input the reference image after image preprocessing and the floating image after image preprocessing into the combined registration mode in sequence, as reference images and floating images of the linear registration and the nonlinear registration, respectively.

[0161] The modules in the above-mentioned image registration device can be realized by software, hardware, or a combination thereof, in whole or in part. The modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform operations corresponding to the modules.

[0162] In one embodiment, a computer device is provided, which includes a memory and a processor, and the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0163] determining a registration framework including at least a combined registration mode, the combined registration mode including a combination of a linear registration mode and a nonlinear registration mode, the linear registration mode and the nonlinear registration mode being applicable to registration of images in different scenarios; obtaining a reference image to be registered and a floating image to be registered; and calling a target registration mode in the registration framework to register the reference image and the floating image.

[0164] In one embodiment, the processor further implements the following steps when executing the computer program:

[0165] constructing the linear registration mode using registration parameters in different scenarios, and constructing the nonlinear registration mode using registration parameters in different scenarios; the registration parameters include a degree-of-freedom parameter and a preset hyperparameter required in a registration process; cascading the linear registration mode and the nonlinear registration mode to obtain the combined registration mode; and determining the registration framework according to the linear registration mode, the nonlinear registration mode, and the combined registration mode.

[0166] In one embodiment, the linear registration manner comprises at least one linear registration module, each linear registration module comprises at least one linear registration step; each linear registration module is used for image-level registration of the reference image and the floating image, and each linear registration step is used for linear optimization of a degree of freedom parameter between the reference image and the floating image.

[0167] In one embodiment, each linear registration module corresponds to image registration at a resampled image resolution; and each linear registration step corresponds to optimization of one or more target degree of freedom parameters.

[0168] In one embodiment, the linear registration modules are in a first cascade relationship, the linear registration steps are in a second cascade relationship, and the target degree of freedom parameters optimized by the linear registration steps are not all the same; the first cascade relationship is used to represent that, in two adjacent linear registration modules, the output of a previous linear registration module is used as the input of a next linear registration module; and the second cascade relationship is used to represent that, in two adjacent linear registration steps, the output of a previous linear registration step is used as the input of a next linear registration step.

[0169] In one embodiment, the non-linear registration manner comprises at least one non-linear registration module, each non-linear registration module comprises at least one non-linear registration step; different non-linear registration modules correspond to image registration at different resampled image resolutions, and the resampled image resolution corresponding to a previous non-linear registration module is smaller than the resampled image resolution corresponding to a next non-linear registration module in two adjacent non-linear registration modules.

[0170] In one embodiment, the processor, when executing the computer program, further implements the following steps:

[0171] An initial conversion parameter corresponding to the reference image and the floating image is determined by using a preset parameter determination manner; the parameter determination manner comprises at least one of a coordinate axis conversion manner, a centroid alignment manner, and a slice alignment manner in a body height direction; and the initial conversion parameter, the reference image, and the floating image are sequentially input into the combined registration manner for linear registration and non-linear registration.

[0172] In one embodiment, the processor, when executing the computer program, further implements the following steps:

[0173] The reference image and the floating image are subjected to resampling processing of different times to determine a reference image after each resampling and a floating image after each resampling; the reference images after each resampling are different in image resolution; and the reference images after each resampling and the floating images after each resampling are respectively subjected to image preprocessing.

[0174] In one embodiment, a computer readable storage medium is provided, having stored thereon a computer program which, when executed by a processor, implements the following steps:

[0175] determining a pre-constructed registration framework comprising at least a combined registration manner; the combined registration manner comprises a combination of a linear registration manner and a nonlinear registration manner, both of which are applicable to image registration in different scenarios; obtaining a reference image to be registered and a floating image to be registered; and calling a target registration manner in the registration framework to register the reference image and the floating image.

[0176] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0177] constructing the linear registration manner using registration parameters in different scenarios, and constructing the nonlinear registration manner using registration parameters in different scenarios; the registration parameters comprise degree-of-freedom parameters and preset hyperparameters required in the registration process; concatenating the linear registration manner and the nonlinear registration manner to obtain the combined registration manner; and determining the registration framework according to the linear registration manner, the nonlinear registration manner, and the combined registration manner.

[0178] In one embodiment, the linear registration manner comprises at least one linear registration module, and each linear registration module comprises at least one linear registration step; each linear registration module is used for image-level registration of the reference image and the floating image, and each linear registration step is used for linear optimization of the degree-of-freedom parameters between the reference image and the floating image.

[0179] In one embodiment, each linear registration module corresponds to image registration at a resampling image resolution, and each linear registration step corresponds to optimization of one or more target degree-of-freedom parameters.

[0180] In one embodiment, the linear registration modules have a first concatenation relationship, the linear registration steps have a second concatenation relationship, and the target degree-of-freedom parameters for linear optimization by the linear registration steps are not all the same; the first concatenation relationship is used to represent that, in two adjacent linear registration modules, the output of a previous linear registration module is used as the input of a next linear registration module; and the second concatenation relationship is used to represent that, in two adjacent linear registration steps, the output of a previous linear registration step is used as the input of a next linear registration step.

[0181] In one embodiment, the above nonlinear registration mode comprises at least one nonlinear registration module, each nonlinear registration module comprising at least one nonlinear registration step; different nonlinear registration modules correspond to image registration at different resampling image resolutions, and in adjacent two nonlinear registration modules, the resampling image resolution corresponding to the previous nonlinear registration module is smaller than the resampling image resolution corresponding to the next nonlinear registration module.

[0182] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0183] An initial conversion parameter corresponding to the reference image and the floating image is determined by using a preset parameter determination mode; the parameter determination mode comprises at least one of a coordinate axis conversion mode, a centroid alignment mode, and a slice alignment mode in the height direction of the human body; and the initial conversion parameter, the reference image, and the floating image are sequentially input into the combined registration mode for linear registration and nonlinear registration.

[0184] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0185] The reference image and the floating image are subjected to different times of resampling processing to determine a reference image after each resampling and a floating image after each resampling; the image resolutions of the reference images after each resampling are different; and the reference image after each resampling and the floating image after each resampling are respectively subjected to image preprocessing.

[0186] In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the following steps:

[0187] A registration framework comprising at least a combined registration mode is determined in advance; the combined registration mode comprises a combination of a linear registration mode and a nonlinear registration mode, and the linear registration mode and the nonlinear registration mode are both applicable to image registration in different scenarios; a reference image to be registered and a floating image to be registered are obtained; and a target registration mode in the registration framework is called to register the reference image and the floating image.

[0188] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0189] The linear registration mode is constructed by using registration parameters in different scenarios, and the nonlinear registration mode is constructed by using registration parameters in different scenarios; the registration parameters comprise a degree of freedom parameter and a preset hyperparameter required in a registration process; the linear registration mode and the nonlinear registration mode are cascaded to obtain a combined registration mode; and a registration framework is determined according to the linear registration mode, the nonlinear registration mode, and the combined registration mode.

[0190] In one embodiment, the linear registration manner comprises at least one linear registration module, each linear registration module comprises at least one linear registration step; each linear registration module is used for image-level registration of the reference image and the floating image, and each linear registration step is used for linear optimization of the freedom degree parameters between the reference image and the floating image.

[0191] In one embodiment, each linear registration module corresponds to image registration at a resampled image resolution; and each linear registration step corresponds to optimization of one or more target freedom degree parameters.

[0192] In one embodiment, the linear registration modules are in a first cascade relationship, the linear registration steps are in a second cascade relationship, and the target freedom degree parameters optimized by the linear registration steps are not all the same; the first cascade relationship is used to represent that, in two adjacent linear registration modules, the output of a previous linear registration module is used as the input of a next linear registration module; and the second cascade relationship is used to represent that, in two adjacent linear registration steps, the output of a previous linear registration step is used as the input of a next linear registration step.

[0193] In one embodiment, the non-linear registration manner comprises at least one non-linear registration module, each non-linear registration module comprises at least one non-linear registration step; different non-linear registration modules correspond to image registration at different resampled image resolutions, and the resampled image resolution corresponding to a previous non-linear registration module is smaller than the resampled image resolution corresponding to a next non-linear registration module in two adjacent non-linear registration modules.

[0194] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0195] The initial conversion parameters corresponding to the reference image and the floating image are determined by using a preset parameter determination manner; the parameter determination manner comprises at least one of a coordinate axis conversion manner, a centroid alignment manner, and a slice alignment manner in a body height direction; and the initial conversion parameters, the reference image, and the floating image are sequentially input into the combined registration manner for linear registration and non-linear registration.

[0196] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0197] The reference image and the floating image are subjected to resampling processing of different times to determine a reference image after each resampling and a floating image after each resampling; the image resolutions of the reference images after each resampling are different; and the reference images after each resampling and the floating images after each resampling are respectively subjected to image preprocessing.

[0198] It should be noted that the data involved in the present application (including but not limited to data for analysis, stored data, displayed data, etc.) are all data authorized by the user or fully authorized by all parties.

[0199] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0200] The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0201] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. An image registration method, characterized in that, The method includes: A pre-constructed registration framework is determined, which includes at least a combination of registration methods. The combination of registration methods includes a combination of linear and nonlinear registration methods, both of which are applicable to image registration in different scenarios. These different scenarios include different imaging devices, conditions, body parts, resolutions, degrees of freedom, preprocessing, and interpolation methods. The linear and nonlinear registration methods are obtained by pre-acquiring and optimizing registration parameters for each of the different scenarios. Obtain the reference image and the floating image to be registered, and obtain the actual registration requirements input by the user; The registration framework is invoked, a target registration method matching the actual registration requirements is determined from the registration framework, and the reference image and the floating image are registered using the target registration method.

2. The method according to claim 1, characterized in that, The determination of the pre-constructed registration framework, which includes at least a combination of registration methods, includes: The linear registration method is constructed using registration parameters under different scenarios, and the nonlinear registration method is constructed using the registration parameters under the same different scenarios; the registration parameters include degree-of-freedom parameters and hyperparameters that need to be preset during the registration process; The linear registration method and the nonlinear registration method are concatenated to obtain the combined registration method; The registration framework is determined based on the linear registration method, the nonlinear registration method, and the combined registration method.

3. The method according to claim 2, characterized in that, The linear registration method includes at least one linear registration module, and each linear registration module includes at least one linear registration step; Each of the linear registration modules is used to perform image-level registration between the reference image and the floating image, and each of the linear registration steps is used to perform linear optimization of the degree-of-freedom parameters between the reference image and the floating image.

4. The method according to claim 3, characterized in that, Each of the linear registration modules corresponds to image registration at a resampled image resolution; each of the linear registration steps corresponds to optimizing one or more target degree-of-freedom parameters.

5. The method according to claim 4, characterized in that, The linear registration modules are in a first cascade relationship, and the linear registration steps are in a second cascade relationship, and the target degree of freedom parameters for linear optimization in each linear registration step are not completely the same; The first concatenation relationship is used to characterize that in two adjacent linear registration modules, the output of the previous linear registration module is used as the input of the next linear registration module; The second concatenation relationship is used to characterize that in two adjacent linear registration steps, the output of the previous linear registration step is used as the input of the next linear registration step.

6. The method according to any one of claims 2-5, characterized in that, The nonlinear registration method includes at least one nonlinear registration module, and each nonlinear registration module includes at least one nonlinear registration step; Different nonlinear registration modules correspond to image registration at different resampled image resolutions, and in two adjacent nonlinear registration modules, the resampled image resolution corresponding to the previous nonlinear registration module is smaller than the resampled image resolution corresponding to the next nonlinear registration module.

7. The method according to any one of claims 1-5, characterized in that, The target registration method is the combined registration method, and the registration of the reference image and the floating image using the target registration method includes: The initial transformation parameters corresponding to the reference image and the floating image are determined using a preset parameter determination method; the parameter determination method includes at least one of coordinate axis transformation method, centroid alignment method, and slice alignment method in the human height direction; The initial transformation parameters, the reference image, and the floating image are sequentially input into the combined registration method for linear and nonlinear registration.

8. The method according to claim 7, characterized in that, After determining the initial transformation parameters based on the reference image and the floating image, the method further includes: The reference image and the floating image are resampled a different number of times to determine the reference image and the floating image after each resampling; the image resolution of the reference image is different after each resampling. Image preprocessing is performed on the reference image after each resampling and the floating image after each resampling. The preprocessed reference image and the preprocessed floating image are sequentially input into the combined registration method as the reference image and floating image for the linear registration and nonlinear registration, respectively.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images