Medical image registration method and device, equipment and storage medium

Abstract

The invention relates to a medical image registration method and device, equipment and a storage medium. The method comprises the following steps: acquiring a single reference image and a single floating image; the single reference image is one image in three-dimensional reference images, and the single floating image is one image in three-dimensional floating images; registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image; and performing registration processing on the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter, and determining a second registration parameter between the three-dimensional reference image and the three-dimensional floating image. By adopting the method, the registration between images with incomplete visual fields can be stably realized.

Description

Technical Field

[0001] The present application relates to the technical field of medical image processing, and in particular to a medical image registration method, apparatus, device and storage medium. Background Art

[0002] Surgical robotics refers to the technology that allows doctors to operate machines away from the operating table to perform surgery. At present, the workflow of surgical robots can be roughly divided into three stages: preoperative planning, intraoperative navigation, and intraoperative execution. In the intraoperative navigation and intraoperative execution stages, it is generally necessary to map the surgical parameters planned preoperatively to the intraoperative space. Here, the mapping is generally to first align the image of the preoperative planning surgical parameters with the intraoperative image to form a registration relationship, and then map the planned surgical parameters to the intraoperative space through the registration relationship.

[0003] In the related art, in the process of aligning the image used for preoperative planning of surgical parameters with the intraoperative image, the gradient intensity of the tissue boundary in each preoperative and intraoperative image is generally identified, and then the reference image and the moving image are aligned through the gradient intensity to obtain the alignment relationship between the two.

[0004] However, when applied to a scenario where the field of view of images captured at either the preoperative or intraoperative stage is incomplete, it is difficult to achieve registration between the images at these two stages using the above technology. Summary of the invention

[0005] Based on this, it is necessary to provide a medical image registration method, device, equipment and storage medium that can stably achieve registration between images with incomplete fields of view in order to address the above technical problems.

[0006] In a first aspect, the present application provides a medical image registration method, the method comprising:

[0007] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0008] Registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image;

[0009] The three-dimensional reference image and the three-dimensional floating image are registered according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0010] In one embodiment, registering both the single reference image and the single floating image to a preset three-dimensional standard space and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image includes:

[0011] Registering the single reference image to the three-dimensional standard space, and determining a first initial registration parameter of the single reference image in the three-dimensional standard space;

[0012] Registering the single floating image to the three-dimensional standard space, and determining a second initial registration parameter of the single floating image in the three-dimensional standard space;

[0013] A first registration parameter is determined according to the first initial registration parameter and the second initial registration parameter.

[0014] In one embodiment, the above-mentioned registering the single reference image to the three-dimensional standard space and determining the first initial registration parameter of the single reference image in the three-dimensional standard space includes:

[0015] Inputting a single reference image into a preset neural network, predicting the registration parameters of the single reference image in a three-dimensional standard space, and determining a first initial registration parameter;

[0016] The neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training images. The standard registration parameters are registration parameters of the single training images in a three-dimensional standard space.

[0017] In one embodiment, the method of obtaining the single training image and the standard registration parameters corresponding to the single training image includes:

[0018] Acquire a three-dimensional training image, and register the three-dimensional training image into a three-dimensional standard space;

[0019] Acquire a single training image from the three-dimensional training images registered to the three-dimensional standard space, and obtain the translation amount and angle corresponding to the single training image;

[0020] The translation amount and angle corresponding to a single training image are converted into parameters to obtain the standard registration parameters corresponding to the single training image.

[0021] In one embodiment, the registering process of the 3D reference image and the 3D floating image according to the first registration parameter to determine the second registration parameter between the 3D reference image and the 3D floating image includes:

[0022] Performing registration processing on the three-dimensional floating image using the first registration parameter to determine a transformed three-dimensional floating image;

[0023] According to the visual fields corresponding to the three-dimensional reference image and the transformed three-dimensional floating image, one of the three-dimensional reference image and the transformed three-dimensional floating image is determined as a target three-dimensional image and the other is determined as a non-target three-dimensional image; the visual field of the target three-dimensional image is smaller than that of the non-target three-dimensional image;

[0024] Perform redundancy removal on non-target three-dimensional images;

[0025] The non-target three-dimensional image after the redundancy removal processing and the target three-dimensional image are registered to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0026] In one embodiment, the above-mentioned redundancy removal process for the non-target three-dimensional image includes:

[0027] According to the field of view of the target three-dimensional image, redundancy removal is performed in the non-target three-dimensional image.

[0028] In one embodiment, the above-mentioned registration processing of the non-target three-dimensional image after the redundancy removal processing and the target three-dimensional image to determine the second registration parameter between the three-dimensional reference image and the three-dimensional floating image includes:

[0029] Calculate the similarity measure between the non-target three-dimensional image after redundancy processing and the target three-dimensional image;

[0030] The non-target three-dimensional image after the redundancy removal process and the target three-dimensional image are registered according to the similarity measure to determine the second registration parameter.

[0031] In a second aspect, the present application further provides a medical image registration device, the device comprising:

[0032] An image acquisition module, used to acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0033] A first registration module, used to register the single reference image and the single floating image to a preset three-dimensional standard space, and determine a first registration parameter between the three-dimensional reference image and the three-dimensional floating image;

[0034] The second registration module is used to perform registration processing on the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0035] In a third aspect, the present application further provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0036] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0037] Registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image;

[0038] The three-dimensional reference image and the three-dimensional floating image are registered according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0040] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0041] Registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image;

[0042] The three-dimensional reference image and the three-dimensional floating image are registered according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0043] In a fifth aspect, the present application further provides a computer program product, the computer program product comprising a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0044] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0045] Registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image;

[0046] The three-dimensional reference image and the three-dimensional floating image are registered according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0047] The above-mentioned medical image registration method, device, equipment and storage medium, by acquiring a single reference image in a three-dimensional reference image and acquiring a single floating image in a three-dimensional floating image, register the single reference image and the single floating image to a preset three-dimensional standard space, determine the first registration parameter between the three-dimensional reference image and the three-dimensional floating image, and register the two three-dimensional images according to the first registration parameter, and determine the second registration parameter between the two three-dimensional images. In this method, since two single images can be obtained from two three-dimensional images respectively, and then the two single images are registered to the three-dimensional standard space to obtain the registration parameters for registering the two three-dimensional images, the registration between the two three-dimensional images can be achieved regardless of whether the field of view of the two three-dimensional images is complete, and the image registration has high robustness; then for two three-dimensional images with incomplete fields of view, the registration between the two three-dimensional images with incomplete fields of view can naturally be stably achieved, ensuring accurate registration between images in scenes with incomplete fields of view. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is an internal structure diagram of a computer device in one embodiment;

[0049] Figure 2 is a schematic diagram of a process of a medical image registration method in one embodiment;

[0050] Figure 3 is a schematic flow chart of a medical image registration method in another embodiment;

[0051] Figure 4 is a schematic flow chart of a medical image registration method in another embodiment;

[0052] Figure 5 is a schematic flow chart of a medical image registration method in another embodiment;

[0053] Figure 6 is a schematic flow chart of a medical image registration method in another embodiment;

[0054] Figure 7 is a schematic flow chart of a medical image registration method in another embodiment;

[0055] Figure 8 FIG. 4 is a structural block diagram of a medical image registration device in one embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] At present, for preoperative scan data, such as MRA (magnetic resonance angiography) images, due to the distribution of diseased blood vessels, there are often incomplete FOV (Field of View). This results in the traditional registration method based on gradient descent being extremely prone to falling into the trap of local optimality without initial relatively accurate alignment / registration. Especially in extreme scenarios, for ultra-narrow FOV, it may even cause the registration algorithm to fail. The requirements for registration accuracy of neurosurgery robots are usually at the sub-millimeter level. Failure or failure of registration will seriously affect the subsequent surgical processing procedures. Therefore, the registration of images with incomplete field of view is particularly important. Based on this, the embodiments of the present application provide a medical image registration method, device, equipment and storage medium that can solve the above technical problems.

[0058] The medical image registration method provided in the embodiment of the present application can be applied to a computer device, which can be a server or a terminal. Taking the terminal as an example, its internal structure diagram can be as follows: Figure 1 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a medical image registration method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0059] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may 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, Figure 2 As shown, a medical image registration method is provided, which is applied to Figure 1 Taking the computer device in the example as an example, the method may include the following steps:

[0061] S202, acquiring a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image.

[0062] In this step, a three-dimensional reference image and a three-dimensional floating image may be acquired in advance, and the three-dimensional reference image and the three-dimensional floating image may be referred to as the three-dimensional images to be registered. The three-dimensional reference image and the three-dimensional floating image may be images of any modality, such as CT (Computed Tomography) images, MR (Magnetic Resonance) images, PET (Positron Emission Tomography) images, and other modality images; for the three-dimensional reference image and the three-dimensional floating image, the modalities of the two may be the same or different, such as the three-dimensional reference image may be a CT image, and the three-dimensional floating image may be a PET image.

[0063] Here, the 3D reference image and the 3D floating image can be 3D images with an incomplete field of view or 3D images with a complete field of view. In other words, the 3D reference image and the 3D floating image can both be 3D images with an incomplete field of view, or either one of them can be a 3D image with an incomplete field of view, or both can be 3D images with a complete field of view.

[0064] For three-dimensional images with incomplete field of view, it means that the field of view is incomplete when the image is taken. For example, the part to be measured may not be fully photographed. For example, for a complete brain, the entire area from the top of the skull to the base of the skull should be scanned completely, but sometimes only the top of the skull or the base of the skull is scanned. This can be called incomplete field of view.

[0065] It should be noted that both the three-dimensional reference image and the three-dimensional floating image here include at least one two-dimensional image, that is, at least one two-dimensional slice image.

[0066] The three-dimensional reference image and the three-dimensional floating image can be three-dimensional images of the same object to be measured or the same part of the same object to be measured, obtained under different conditions, where the different conditions may include different time, different scanning protocols, different scanning periods, etc. Generally, the three-dimensional floating image is registered to the coordinate system of the three-dimensional reference image, so as to achieve registration between the two.

[0067] The method for acquiring the three-dimensional reference image and the three-dimensional floating image may be a three-dimensional image obtained by scanning and reconstructing the same object to be measured or the same part of the same object to be measured under different conditions; or it may be a pre-stored three-dimensional reference image and a three-dimensional floating image obtained from the cloud or a server; or it may be other acquisition methods, which are not specifically limited here.

[0068] After obtaining the three-dimensional reference image and the three-dimensional floating image, a two-dimensional image can be selected from the three-dimensional reference image as a single reference image, and a two-dimensional image can also be selected from the three-dimensional floating image as a single floating image; that is, the single reference image and the single floating image here are generally two-dimensional images, and the two-dimensional image generally only corresponds to a part of the field of view during shooting / scanning, that is, the single reference image and the single floating image can be considered as images with incomplete field of view.

[0069] For selecting a single reference image from the three-dimensional reference image and selecting a single floating image from the three-dimensional floating image, both can be a two-dimensional image selected arbitrarily from the corresponding three-dimensional image; in addition, for the selected single reference image and single floating image, the positions of the two in their respective three-dimensional images can be corresponding, that is, the two can be two-dimensional images at the same position in the two three-dimensional images. Of course, the positions of the single reference image and the single floating image in their respective three-dimensional images can also be non-corresponding, that is, the two can be two-dimensional images at different positions in the two three-dimensional images.

[0070] S204: registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0071] In this step, the preset three-dimensional standard space refers to a standard three-dimensional space pre-set for a certain part or a certain organ tissue, etc., which can be a three-dimensional standard image space. The standard three-dimensional space is generally a digital three-dimensional standard space with specific group characteristics, complete annotations, accuracy and group representativeness, which can provide a spatial basis or navigation guidance for the subsequent analysis or processing of the part or organ tissue. For example, the part or the organ tissue of different objects can refer to the corresponding three-dimensional standard space for subsequent processing or analysis, etc. For example, a standard three-dimensional brain space can be pre-set for the brain, and the subsequent brain image processing of different objects can be processed based on the three-dimensional brain space as the data basis.

[0072] In addition, for a pre-set three-dimensional standard space, the orientation information (such as position, angle, etc.) of any point therein can be obtained when the three-dimensional standard space is set.

[0073] Furthermore, after obtaining the single reference image as described above, the single reference image can be compared with the three-dimensional standard space. The specific comparison can be, for example, comparing the orientation of each point on the single reference image with the orientation of each point on the three-dimensional standard space, determining the orientation of each point of the single reference image in the three-dimensional standard space, and then comprehensively obtaining the orientation of the single reference image in the three-dimensional standard space. Similarly, after obtaining a single floating image, the single floating image can also be compared with the three-dimensional standard space. The specific comparison process can refer to the comparison process of the single reference image. In short, the orientation of the single floating image in the three-dimensional standard space can be finally obtained. Afterwards, the orientation of the single reference image in the three-dimensional standard space and the orientation of the single floating image in the three-dimensional standard space are aligned to obtain the alignment parameter, which is the first alignment parameter between the three-dimensional reference image and the three-dimensional floating image.

[0074] Regarding the above-mentioned comparison of the orientation of each point on the single reference image with the orientation of each point in the three-dimensional standard space to determine the orientation of each point in the single reference image in the three-dimensional standard space, specifically, a feature point recognition method can be adopted, that is, feature points are extracted in the single reference image, and corresponding feature points are also extracted in the three-dimensional standard space. Usually, three non-coplanar points can determine a three-dimensional transformation matrix, and then the orientation of each point in the single reference image in the three-dimensional standard space can be calculated according to the feature point matching algorithm (such as ICP (Iterative Closest Point) algorithm).

[0075] Alternatively, after obtaining the single reference image and the single floating image as described above, the orientation of each point on the single reference image and the orientation of each point on the single floating image can also be compared with the orientation of each point in the three-dimensional standard space at the same time to obtain the orientation alignment parameters of the single reference image and the single floating image in the three-dimensional standard space, which are recorded as the first alignment parameters between the three-dimensional reference image and the three-dimensional floating image.

[0076] It should be noted that the first registration parameter here is a coarse registration parameter between the 3D reference image and the 3D floating image, that is, the 3D reference image and the 3D floating image can be pre-registered and pre-positioned first to facilitate subsequent rapid and accurate registration.

[0077] S206: Perform registration processing on the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0078] In this step, after obtaining the first registration parameter, the first registration parameter can be used to perform registration processing on the three-dimensional floating image, and the three-dimensional floating image is registered to the coordinate system where the three-dimensional reference image is located, and the three-dimensional floating image after rough registration is determined. At this time, the rough registration of the three-dimensional floating image is completed. Generally, the registration accuracy of the three-dimensional floating image after rough registration is not enough, and further fine registration is required, that is, it is necessary to obtain fine registration parameters for further fine registration, which can be recorded as second registration parameters. It should be noted that the registration processing of a certain image mentioned in this article refers to transform sampling or conversion processing of the image.

[0079] Regarding the method for determining the second registration parameters, a first method can be to perform fine registration based on the three-dimensional reference image and the three-dimensional floating image that has been coarsely registered using the first registration parameters to obtain fine registration parameters, and then multiply the first registration parameters and the fine registration parameters to obtain final registration parameters, i.e., obtain the second registration parameters.

[0080] The second method is to continuously compare and iterate the first registration parameter (i.e., the rough registration parameter) through the similarity between the three-dimensional floating image and the three-dimensional reference image after rough registration, so that the registration accuracy between the three-dimensional floating image and the three-dimensional reference image in the iterative process reaches a certain requirement. The registration parameter at this time is the registration parameter that meets the registration accuracy requirement, which can be recorded as the second registration parameter between the three-dimensional reference image and the three-dimensional floating image. In other words, the second registration parameter is obtained by continuously iterating the first registration parameter, and the registration accuracy of the second registration parameter for the image is higher than the registration accuracy of the first registration parameter for the image. The three-dimensional floating image is registered again by the second registration parameter, so that the spatial position of the registered three-dimensional floating image is closer to the spatial position of the three-dimensional reference image, which can provide a more accurate data basis for the subsequent preoperative or intraoperative scenes that require registration of images, so that the subsequent processing process can be accurately carried out.

[0081] It should be noted that both the first registration parameter and the second registration parameter can be represented by a registration matrix or a transformation matrix.

[0082] In the above-mentioned medical image registration method, by acquiring a single reference image in a three-dimensional reference image and acquiring a single floating image in a three-dimensional floating image, the single reference image and the single floating image are registered to a preset three-dimensional standard space, a first registration parameter between the three-dimensional reference image and the three-dimensional floating image is determined, and the two three-dimensional images are registered according to the first registration parameter to determine a second registration parameter between the two three-dimensional images. In this method, since two single images can be first acquired from two three-dimensional images respectively, and then the two single images are registered to the three-dimensional standard space to obtain the registration parameters for registering the two three-dimensional images, the registration between the two three-dimensional images can be achieved regardless of whether the fields of view of the two three-dimensional images are complete, and the image registration has high robustness; then for two three-dimensional images with incomplete fields of view, the registration between the two three-dimensional images with incomplete fields of view can naturally be stably achieved, ensuring accurate registration between images in scenes with incomplete fields of view.

[0083] The above embodiments mentioned that a single reference image and a single floating image can be registered to a three-dimensional standard space respectively. The following embodiments illustrate a possible implementation of the process.

[0084] In another embodiment, another medical image registration method is provided. Based on the above embodiment, Figure 3 As shown, the above S204 may include the following steps:

[0085] S302: register the single reference image to the three-dimensional standard space, and determine a first initial registration parameter of the single reference image in the three-dimensional standard space.

[0086] In this step, when registering a single reference image to a three-dimensional standard space, a neural network may be used for registration. As an optional embodiment, a single reference image may be input into a preset neural network, and the registration parameters of the single reference image in the three-dimensional standard space are predicted and processed to determine the first initial registration parameters. The neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training images, and the standard registration parameters are the registration parameters of the single training image in the three-dimensional standard space.

[0087] That is to say, a neural network can be pre-trained, and the neural network can be trained by multiple single training images (i.e., two-dimensional slice images) and the registration parameters of each single training image in a three-dimensional standard space. Here, the registration parameters of each single training image in the three-dimensional standard space can include orientation parameters or posture parameters in the three-dimensional standard space, etc., wherein the orientation parameters or posture parameters can include, for example, rotation angles, translation amounts, etc. in the three-dimensional standard space. Of course, the orientation parameters or posture parameters can also be represented by Euler angles, quaternions, rotation matrices, anchor points AnchorPoints, etc. Here, the parameters such as Euler angles, quaternions, rotation matrices, anchor points AnchorPoints, etc. can also be converted to rotation angles and translation amounts. Here, each single training image can be pre-registered to the three-dimensional standard space to obtain the corresponding standard registration parameters, or other methods can be used to obtain them. For example, the atlas space can be used as the three-dimensional standard space, and the image of the complete FOV can be used as the three-dimensional floating image. The position of each slice in the three-dimensional floating image in the standard three-dimensional space is obtained directly through the traditional gradient descent rigid registration method. In addition, this embodiment does not specifically limit the specific structure of the neural network.

[0088] After the neural network is pre-trained as mentioned above, the obtained single reference image can be input into the neural network for prediction of the registration parameters. By continuously performing convolution, pooling, activation, full connection and other processes in the neural network, the predicted registration parameters corresponding to the single reference image can be output eventually, that is, the orientation or posture of the single reference image in the three-dimensional standard space, which is recorded as the first initial registration parameter.

[0089] S304: register the single floating image to the three-dimensional standard space, and determine a second initial registration parameter of the single floating image in the three-dimensional standard space.

[0090] In this step, after obtaining a single floating image, the prediction method of the above-mentioned single reference image can be referred to, that is, the single floating image can be input into a preset neural network, and the registration parameters of the single floating image in the three-dimensional standard space can be predicted and processed to obtain the predicted registration parameters corresponding to the single floating image, which are recorded as the second initial registration parameters.

[0091] S306: Determine a first registration parameter according to the first initial registration parameter and the second initial registration parameter.

[0092] The first initial registration parameter corresponding to the single reference image may be the registration matrix of the single reference image in the three-dimensional standard space. Similarly, the second initial registration parameter corresponding to the single floating image may be the registration matrix of the single floating image in the three-dimensional standard space.

[0093] In this step, after obtaining the first initial registration parameters corresponding to the single reference image and the second initial registration parameters corresponding to the single floating image, since both the single reference image and the single floating image are registered to the three-dimensional standard space, the first initial registration parameters and the second initial registration parameters can be converted through the three-dimensional standard space to obtain the registration relationship between the first initial registration parameters and the second initial registration parameters, which can correspond to the registration relationship between the three-dimensional reference image and the three-dimensional floating image, that is, the first registration parameters between the three-dimensional reference image and the three-dimensional floating image are obtained.

[0094] In this embodiment, a single reference image and a single floating image are registered to the three-dimensional standard space to obtain their respective initial registration parameters, and then the coarse registration parameters between the two three-dimensional images are obtained through the two initial registration parameters. This ensures that the coarse registration parameters between the two three-dimensional images can still be accurately obtained for images with incomplete fields of view, thereby facilitating the subsequent successful implementation of the fine registration process. In addition, the registration parameters of a single image are predicted and processed through a neural network, so that the registration parameters can be obtained quickly and accurately, thereby improving the efficiency and accuracy of image registration.

[0095] The above embodiments mentioned that the registration parameters of a single image can be predicted by a trained neural network. The following embodiments illustrate the method of obtaining the training images used in the neural network training.

[0096] In another embodiment, another medical image registration method is provided. Based on the above embodiment, Figure 4 As shown, the method for obtaining the above-mentioned single training image and the standard registration parameters corresponding to the single training image includes the following steps:

[0097] S402, acquiring a three-dimensional training image, and registering the three-dimensional training image into a three-dimensional standard space.

[0098] In this step, three-dimensional training images of multiple modalities, such as CT three-dimensional training images, MR three-dimensional training images, PET three-dimensional training images, etc., can be acquired in advance, and then these three-dimensional training data can be preprocessed, such as data scalping, grayscale normalization, etc. After that, the above-mentioned preprocessed three-dimensional training images can be registered to the three-dimensional standard space, specifically, each three-dimensional training image can be aligned or registered to the three-dimensional standard space according to the orientation of the three-dimensional standard space. For example, the three-dimensional standard space here can be an atlas space, and a three-dimensional rigid registration method can be used to directly register the three-dimensional training image and the three-dimensional standard space.

[0099] S404, obtaining a single training image from the three-dimensional training images registered to the three-dimensional standard space, and obtaining a translation amount and an angle corresponding to the single training image.

[0100] In this step, the three-dimensional training images of each of the above modalities may include multiple single two-dimensional training images. After each three-dimensional training image is aligned to the three-dimensional standard space, a series of two-dimensional slice images can be generated at the position and angle of any point in the three-dimensional standard space. Here, the position of any point can represent any translation amount in the three-dimensional standard space.

[0101] According to different combinations of translation amounts and angles, the three-dimensional training images registered to the three-dimensional standard space are randomly sampled, and training samples of single two-dimensional slice images are taken out and recorded as single training images. At the same time, the corresponding translation amounts and angles of each single training image taken out are recorded.

[0102] S406, performing parameter conversion on the translation amount and angle corresponding to the single training image to obtain standard registration parameters corresponding to the single training image.

[0103] In this step, the standard registration parameters of a single training image generally used by a neural network during training may include parameters such as Euler angles, quaternions, rotation matrices, anchor points AnchorPoints, etc., all of which can be converted from the translation amounts and angles corresponding to the single training image. Therefore, after obtaining the translation amounts and angles corresponding to each single training image, the translation amounts and angles can be converted to Euler angles, quaternions, rotation matrices, anchor points AnchorPoints, etc. parameters to obtain the standard registration parameters corresponding to the single training image.

[0104] In this embodiment, by aligning the three-dimensional training images of each modality to the three-dimensional standard space, and after the alignment, sampling multiple single training images according to different translation amounts and angles, and obtaining the standard alignment parameters corresponding to each single training image, and then training the neural network, the trained neural network can be applied to image alignment of various modalities, thereby improving the generalization performance of the neural network; at the same time, by randomly sampling a single training image from the aligned three-dimensional training images, the richness of the obtained single training images and the efficiency of obtaining the single training images can also be improved.

[0105] The above embodiment mentioned that two three-dimensional images can be registered by using coarse registration parameters to determine further fine registration parameters. The following embodiment specifically describes the process.

[0106] In another embodiment, another medical image registration method is provided. Based on the above embodiment, Figure 5 As shown, the above S206 may include the following steps:

[0107] S502: Performing registration processing on the three-dimensional floating image using the first registration parameter to determine a transformed three-dimensional floating image.

[0108] In this step, after obtaining the first registration parameter between the two three-dimensional images, the floating image is generally registered to the coordinate system where the reference image is located. Therefore, the first registration parameter can be used here to register the three-dimensional floating image, so as to roughly register the three-dimensional floating image to the coordinate system where the three-dimensional reference image is located, and obtain the three-dimensional floating image after rough registration, which is recorded as the transformed three-dimensional floating image.

[0109] S504, determining one of the three-dimensional reference image and the transformed three-dimensional floating image as a target three-dimensional image and the other as a non-target three-dimensional image according to the respective corresponding field of view ranges of the three-dimensional reference image and the transformed three-dimensional floating image; the field of view range of the target three-dimensional image is smaller than the field of view range of the non-target three-dimensional image.

[0110] S506: performing redundancy removal processing on the non-target three-dimensional image.

[0111] In S504 and S506, the field of view (e.g., field of view angle) of the three-dimensional reference image and the three-dimensional floating image when they are captured may be different. Therefore, when the two three-dimensional images with different fields of view are registered, there may be some asymmetric redundant information, thereby affecting the accuracy of the registration. Based on this, this step proposes to remove the redundancy of the two three-dimensional images with different fields of view, so as to eventually make the field of view of the two registered three-dimensional images close, thereby improving the registration accuracy.

[0112] For the above-mentioned redundancy removal processing, as an optional embodiment, it can be to obtain a first field of view corresponding to the three-dimensional reference image, and to obtain a second field of view corresponding to the transformed three-dimensional floating image; the above-mentioned first field of view and second field of view are different; based on the first field of view and the second field of view, redundancy removal processing is performed on the three-dimensional reference image and the transformed three-dimensional floating image.

[0113] That is to say, the field of view of the two three-dimensional images can be obtained first. The field of view of the image can be obtained according to the setting parameters of the scanning device during the image shooting process; the field of view of the three-dimensional reference image obtained here can be recorded as the first field of view. The rough alignment of the three-dimensional floating image here usually does not change its field of view, so the field of view of the three-dimensional floating image can be obtained here, or the field of view of the transformed three-dimensional floating image (that is, the three-dimensional floating image after rough alignment) can be obtained. The two are the same and are recorded as the second field of view.

[0114] Afterwards, based on the two fields of view, the target field of view and the non-target field of view in the first field of view and the second field of view can be determined; wherein the target field of view is smaller than the non-target field of view; that is, the first field of view and the second field of view can be compared to obtain a smaller field of view, which is recorded as the target field of view, and the larger field of view can be recorded as the non-target field of view. Then, the image of the smaller field of view can be recorded as the target three-dimensional image, and the image of the larger field of view can be recorded as the non-target three-dimensional image, that is, one of the three-dimensional reference image and the three-dimensional floating image is the target three-dimensional image, and the other is the non-target three-dimensional image.

[0115] Then, redundancy removal can be performed in the non-target three-dimensional image according to the target field of view; that is, the image with a large field of view can be intercepted according to the image with a small field of view, and the field of view can be expanded to a certain extent, and then the redundant image corresponding to the image with a large field of view can be obtained, that is, the redundancy removal of the image with a large field of view is realized.

[0116] For example, a single slice image in which both the reference image and the floating image are two-dimensional is taken as an example, and the three-dimensional reference image and the three-dimensional floating image are similar to this, see Figure 6 As shown in the example diagram of redundancy removal processing (the reference image is a single slice image of the three-dimensional reference image in this embodiment, and the moving image is a single slice image of the three-dimensional floating image in this embodiment), there may be two situations. One is that the field of view of the reference image is larger than the field of view of the floating image. In this case, the reference image can be subjected to redundancy removal processing. Specifically, an image with the same field of view as the floating image is intercepted on the reference image (for example, the portion between the solid lines in the left figure of Figure a), and then the field of view of the intercepted image is expanded to a certain extent at both ends (for example, the portion between the dotted lines in the left figure of Figure a). After that, the grayscale of the image outside the expanded field of view on the reference image can be set as the background grayscale, and a redundant image after the redundancy removal processing of the reference image can be obtained (for example, the portion between the dotted lines in the left figure of Figure a).

[0117] The other is that the field of view of the reference image is smaller than that of the floating image. In this case, the floating image can be processed for redundancy removal. Specifically, an image with the same field of view as the reference image is cut out from the floating image (not shown in Figure b), and then the field of view of the cut image is expanded to both ends by a certain amount (such as the part between the dotted lines in the right picture of Figure b). After that, the grayscale of the image outside the expanded field of view on the floating image can be set to the background grayscale, and a redundant image after the floating image is processed for redundancy removal can be obtained (such as the part between the dotted lines in the right picture of Figure b).

[0118] Of course, after removing redundancy from the large field of view three-dimensional image (i.e., the non-target three-dimensional image) of the two three-dimensional images, the non-target three-dimensional image and the target three-dimensional image after removing redundancy can also be rigidly registered based on the ROI region of interest to remove mismatched information outside the field of view and improve the accuracy of subsequent fine registration.

[0119] S508: performing registration processing on the non-target three-dimensional image after the redundancy removal processing and the target three-dimensional image to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0120] In this step, after the non-target three-dimensional image is subjected to redundancy removal processing, a non-target three-dimensional image after redundancy removal processing can be obtained. Then, the non-target three-dimensional image after redundancy removal and the target three-dimensional image can be finely registered.

[0121] During fine registration, as an optional embodiment, a similarity measure between the non-target three-dimensional image after redundancy removal and the target three-dimensional image may be calculated; the non-target three-dimensional image after redundancy removal and the target three-dimensional image are registered according to the similarity measure to determine a second registration parameter.

[0122] That is to say, the similarity measure between the two three-dimensional images after removing the redundancy can be calculated, specifically: the similarity measure of the ROI area of ​​the two three-dimensional images after removing the redundancy is calculated, and then the fine registration is performed according to the similarity measure to obtain the registration parameters of this time, and then the registration parameters here can be derived, and the new registration parameters are used to update the registration parameters of this time, and then the updated registration parameters are used to transform or align the two three-dimensional images, and then the similarity of the two three-dimensional images is evaluated again by the similarity measure, that is, whether the similarity measure meets the threshold condition is determined, if it is satisfied, the new registration parameters are used as the fine registration parameters; if not, the process is iterated repeatedly until the set convergence threshold or number of iterations is reached, and the registration parameters at the end of the iteration are obtained, which are the fine registration parameters.

[0123] Then, the fine registration parameter and the first registration parameter may be multiplied to obtain a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0124] Another optional implementation of fine registration is given below. The process may include: 1. Based on the coarse registration parameters, deform the three-dimensional floating image to obtain the deformed three-dimensional floating image; 2. Based on the deformed three-dimensional floating image and the three-dimensional reference image, determine the range of redundant selection; 3. Invert the coarse registration parameters, transform the redundant range of the deformed three-dimensional floating image to the redundant range before deformation, and obtain the redundant range of the undeformed three-dimensional floating image; 4. Cut the three-dimensional reference image according to the redundant range; 5. Based on the coarse registration parameters, iterate the fine registration of the three-dimensional reference image cut based on the redundant range and the three-dimensional floating image. The specific iterative process may also be the iterative method in the above-mentioned optional embodiment, which will not be repeated here.

[0125] The following mathematically describes the above image fine registration process, that is, for the three-dimensional floating image I m and the 3D reference image I f , we hope to find a set of transformation parameters T (i.e., registration parameters) so that the transformed three-dimensional floating image and the three-dimensional reference image can achieve a matching relationship when the similarity S is maximum. Specifically, it can be expressed by the following formula (2):

[0126]

[0127] in, represents a composite mapping, R is a regularization term, and S is a similarity measure, which can also be called similarity.

[0128] In this embodiment, by performing redundancy processing on the two three-dimensional images based on the field of view of the two three-dimensional images, it is possible to avoid the difference in the field of view of the two three-dimensional images being too large to affect the accuracy of subsequent fine registration, thereby improving the accuracy of subsequent fine registration. In addition, by performing redundancy processing on the large field of view image through the small field of view image in the two three-dimensional images, a relatively accurate redundancy result can be obtained, thereby ensuring the accuracy of subsequent image registration through the redundant result. Furthermore, in the fine registration process, fine registration can be performed based on the similarity between the two images, so that the fine registration process can be achieved relatively quickly and accurately.

[0129] In order to better illustrate the method of the embodiment of the present application, a detailed embodiment is given below to illustrate the technical solution of the present application. Figure 7 As shown in the registration framework diagram, based on the above embodiment, the method may include the following steps:

[0130] S1, acquiring a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0131] S2, inputting the single reference image into a preset neural network, predicting the registration parameters of the single reference image in the three-dimensional standard space, and determining the first initial registration parameters;

[0132] S3, inputting the single floating image into a preset neural network, predicting the registration parameters of the single floating image in the three-dimensional standard space, and determining the second initial registration parameters;

[0133] S4, determining a first registration parameter according to the first initial registration parameter and the second initial registration parameter;

[0134] S5, performing registration processing on the three-dimensional floating image using the first registration parameter to determine a transformed three-dimensional floating image;

[0135] S6, determining one of the three-dimensional reference image and the transformed three-dimensional floating image as a target three-dimensional image and the other as a non-target three-dimensional image according to the respective corresponding fields of view of the three-dimensional reference image and the transformed three-dimensional floating image; the field of view of the target three-dimensional image is smaller than the field of view of the non-target three-dimensional image;

[0136] S7, performing redundancy removal processing in the non-target three-dimensional image according to the field of view of the target three-dimensional image;

[0137] S8, calculating a similarity measure between the non-target three-dimensional image after redundancy removal and the target three-dimensional image;

[0138] S9, performing registration processing on the non-target three-dimensional image after the redundancy removal processing and the target three-dimensional image according to the similarity measurement, and determining a second registration parameter.

[0139] It should be noted that Figure 7 The standard spatial orientation in is the three-dimensional standard space in the embodiment of the present application, the coarse registration matrix is ​​the first registration parameter in the embodiment of the present application, the fine registration matrix is ​​the second registration parameter in the embodiment of the present application, the reference image (1Slice) refers to a single reference image, the floating image (1Slice) refers to a single floating image, the reference image refers to a three-dimensional reference image, and the floating image refers to a three-dimensional floating image.

[0140] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages 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 to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0141] Based on the same inventive concept, the embodiment of the present application also provides a medical image registration device for implementing the medical image registration method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more medical image registration device embodiments provided below can refer to the limitations of the medical image registration method above, and will not be repeated here.

[0142] In one embodiment, Figure 8 As shown, a medical image registration device is provided, comprising: an image acquisition module, a first registration module and a second registration module, wherein:

[0143] An image acquisition module, used to acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image;

[0144] A first registration module, used to register the single reference image and the single floating image to a preset three-dimensional standard space, and determine a first registration parameter between the three-dimensional reference image and the three-dimensional floating image;

[0145] The second registration module is used to perform registration processing on the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0146] In another embodiment, another medical image registration device is provided. Based on the above embodiment, the above first registration module may include:

[0147] A first initial registration unit, used to register the single reference image to the three-dimensional standard space, and determine a first initial registration parameter of the single reference image in the three-dimensional standard space;

[0148] A second initial registration unit, used to register the single floating image to the three-dimensional standard space, and determine a second initial registration parameter of the single floating image in the three-dimensional standard space;

[0149] The first registration unit is used to determine a first registration parameter according to the first initial registration parameter and the second initial registration parameter.

[0150] Optionally, the above-mentioned first initial registration unit is specifically used to input a single reference image into a preset neural network, predict the registration parameters of the single reference image in the three-dimensional standard space, and determine the first initial registration parameters; wherein the above-mentioned neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training image, and the above-mentioned standard registration parameters are the registration parameters of the single training image in the three-dimensional standard space.

[0151] In another embodiment, another medical image registration device is provided. Based on the above embodiment, the above device may further include: a training image acquisition module, wherein:

[0152] The training image acquisition module is used to acquire three-dimensional training images and align the three-dimensional training images to the three-dimensional standard space; acquire a single training image from the three-dimensional training images aligned to the three-dimensional standard space, and obtain the translation amount and angle corresponding to the single training image; perform parameter conversion on the translation amount and angle corresponding to the single training image, and obtain the standard alignment parameters corresponding to the single training image.

[0153] In another embodiment, another medical image registration device is provided. Based on the above embodiment, the first registration unit may include:

[0154] A standard space registration subunit, configured to register a single reference image to a three-dimensional standard space according to a first initial registration parameter, and to register a single floating image to a three-dimensional standard space according to a second initial registration parameter;

[0155] The first registration subunit is used to establish a registration relationship between a single reference image and a single floating image in a three-dimensional standard space, and determine a first registration parameter.

[0156] In another embodiment, another medical image registration device is provided. Based on the above embodiment, the second registration module may include:

[0157] A transformation unit, used to perform a registration process on the three-dimensional floating image using a first registration parameter to determine a transformed three-dimensional floating image;

[0158] an image determination unit, for determining one of the three-dimensional reference image and the transformed three-dimensional floating image as a target three-dimensional image and the other as a non-target three-dimensional image according to the respective corresponding fields of view of the three-dimensional reference image and the transformed three-dimensional floating image; the field of view of the target three-dimensional image is smaller than the field of view of the non-target three-dimensional image;

[0159] A redundancy processing unit, used for performing redundancy processing on non-target three-dimensional images;

[0160] The second registration unit is used to perform registration processing on the non-target three-dimensional image after redundancy removal processing and the target three-dimensional image, and determine the second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0161] Optionally, the redundancy processing unit is specifically configured to perform redundancy removal processing in the non-target three-dimensional image according to the field of view of the target three-dimensional image.

[0162] Optionally, the second registration unit is specifically used to calculate a similarity measure between the non-target three-dimensional image after redundancy removal and the target three-dimensional image; perform registration processing on the non-target three-dimensional image after redundancy removal and the target three-dimensional image according to the similarity measure to determine a second registration parameter.

[0163] Each module in the above-mentioned medical image registration device can be implemented in whole or in part by software, hardware or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0164] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0165] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image; register the single reference image and the single floating image to a preset three-dimensional standard space, and determine a first registration parameter between the three-dimensional reference image and the three-dimensional floating image; register the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter, and determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0166] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0167] A single reference image is registered to a three-dimensional standard space to determine a first initial registration parameter of the single reference image in the three-dimensional standard space; a single floating image is registered to a three-dimensional standard space to determine a second initial registration parameter of the single floating image in the three-dimensional standard space; and a first registration parameter is determined based on the first initial registration parameter and the second initial registration parameter.

[0168] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0169] A single reference image is input into a preset neural network, and the registration parameters of the single reference image in the three-dimensional standard space are predicted and processed to determine the first initial registration parameters; wherein the above-mentioned neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training image, and the above-mentioned standard registration parameters are the registration parameters of the single training image in the three-dimensional standard space.

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

[0171] Acquire three-dimensional training images, and align each three-dimensional training image to the three-dimensional standard space; acquire a single training image from the three-dimensional training images aligned to the three-dimensional standard space, and obtain the translation amount and angle corresponding to the single training image; perform parameter conversion on the translation amount and angle corresponding to the single training image, and obtain the standard alignment parameters corresponding to the single training image.

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

[0173] A single reference image is registered to the three-dimensional standard space according to the first initial registration parameter, and a single floating image is registered to the three-dimensional standard space according to the second initial registration parameter; a registration relationship between the single reference image and the single floating image is established in the three-dimensional standard space, and the first registration parameter is determined.

[0174] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0175] The three-dimensional floating image is registered by using the first registration parameter to determine the transformed three-dimensional floating image; one of the three-dimensional reference image and the transformed three-dimensional floating image is determined as the target three-dimensional image and the other is determined as the non-target three-dimensional image according to the visual field ranges corresponding to the three-dimensional reference image and the transformed three-dimensional floating image respectively; the visual field range of the target three-dimensional image is smaller than that of the non-target three-dimensional image; the non-target three-dimensional image is subjected to redundancy removal processing; the non-target three-dimensional image subjected to the redundancy removal processing and the target three-dimensional image are registered to determine the second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

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

[0177] According to the field of view of the target three-dimensional image, redundancy removal is performed in the non-target three-dimensional image.

[0178] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0179] The similarity measure between the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image is calculated; the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image are registered according to the similarity measure to determine the second registration parameter.

[0180] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0181] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image; register the single reference image and the single floating image to a preset three-dimensional standard space, and determine a first registration parameter between the three-dimensional reference image and the three-dimensional floating image; register the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter, and determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0182] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0183] A single reference image is registered to a three-dimensional standard space to determine a first initial registration parameter of the single reference image in the three-dimensional standard space; a single floating image is registered to a three-dimensional standard space to determine a second initial registration parameter of the single floating image in the three-dimensional standard space; and a first registration parameter is determined based on the first initial registration parameter and the second initial registration parameter.

[0184] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0185] A single reference image is input into a preset neural network, and the registration parameters of the single reference image in the three-dimensional standard space are predicted and processed to determine the first initial registration parameters; wherein the above-mentioned neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training image, and the above-mentioned standard registration parameters are the registration parameters of the single training image in the three-dimensional standard space.

[0186] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0187] Acquire three-dimensional training images, and align each three-dimensional training image to the three-dimensional standard space; acquire a single training image from the three-dimensional training images aligned to the three-dimensional standard space, and obtain the translation amount and angle corresponding to the single training image; perform parameter conversion on the translation amount and angle corresponding to the single training image, and obtain the standard alignment parameters corresponding to the single training image.

[0188] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0189] A single reference image is registered to the three-dimensional standard space according to the first initial registration parameter, and a single floating image is registered to the three-dimensional standard space according to the second initial registration parameter; a registration relationship between the single reference image and the single floating image is established in the three-dimensional standard space, and the first registration parameter is determined.

[0190] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0191] The three-dimensional floating image is registered by using the first registration parameter to determine the transformed three-dimensional floating image; one of the three-dimensional reference image and the transformed three-dimensional floating image is determined as the target three-dimensional image and the other is determined as the non-target three-dimensional image according to the visual field ranges corresponding to the three-dimensional reference image and the transformed three-dimensional floating image respectively; the visual field range of the target three-dimensional image is smaller than that of the non-target three-dimensional image; the non-target three-dimensional image is subjected to redundancy removal processing; the non-target three-dimensional image subjected to the redundancy removal processing and the target three-dimensional image are registered to determine the second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0192] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0193] According to the field of view of the target three-dimensional image, redundancy removal is performed in the non-target three-dimensional image.

[0194] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0195] The similarity measure between the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image is calculated; the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image are registered according to the similarity measure to determine the second registration parameter.

[0196] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0197] Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image; register the single reference image and the single floating image to a preset three-dimensional standard space, and determine a first registration parameter between the three-dimensional reference image and the three-dimensional floating image; register the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter, and determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0198] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0199] A single reference image is registered to a three-dimensional standard space to determine a first initial registration parameter of the single reference image in the three-dimensional standard space; a single floating image is registered to a three-dimensional standard space to determine a second initial registration parameter of the single floating image in the three-dimensional standard space; and a first registration parameter is determined based on the first initial registration parameter and the second initial registration parameter.

[0200] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0201] A single reference image is input into a preset neural network, and the registration parameters of the single reference image in the three-dimensional standard space are predicted and processed to determine the first initial registration parameters; wherein the above-mentioned neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training image, and the above-mentioned standard registration parameters are the registration parameters of the single training image in the three-dimensional standard space.

[0202] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0203] Acquire three-dimensional training images, and align each three-dimensional training image to the three-dimensional standard space; acquire a single training image from the three-dimensional training images aligned to the three-dimensional standard space, and obtain the translation amount and angle corresponding to the single training image; perform parameter conversion on the translation amount and angle corresponding to the single training image, and obtain the standard alignment parameters corresponding to the single training image.

[0204] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0205] A single reference image is registered to the three-dimensional standard space according to the first initial registration parameter, and a single floating image is registered to the three-dimensional standard space according to the second initial registration parameter; a registration relationship between the single reference image and the single floating image is established in the three-dimensional standard space, and the first registration parameter is determined.

[0206] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0207] The three-dimensional floating image is registered by using the first registration parameter to determine the transformed three-dimensional floating image; one of the three-dimensional reference image and the transformed three-dimensional floating image is determined as the target three-dimensional image and the other is determined as the non-target three-dimensional image according to the visual field ranges corresponding to the three-dimensional reference image and the transformed three-dimensional floating image respectively; the visual field range of the target three-dimensional image is smaller than that of the non-target three-dimensional image; the non-target three-dimensional image is subjected to redundancy removal processing; the non-target three-dimensional image subjected to the redundancy removal processing and the target three-dimensional image are registered to determine the second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

[0208] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0209] According to the field of view of the target three-dimensional image, redundancy removal is performed in the non-target three-dimensional image.

[0210] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0211] The similarity measure between the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image is calculated; the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image are registered according to the similarity measure to determine the second registration parameter.

[0212] It should be noted that the data involved in this application (including but not limited to data used for analysis, data stored, data displayed, etc.) are all information fully authorized by all parties.

[0213] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in 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 memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access 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 and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0214] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0215] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A medical image registration method, characterized in that: The method comprises: Acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image; Registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image; The three-dimensional reference image and the three-dimensional floating image are registered according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

2. The method according to claim 1, characterized in that: The step of registering the single reference image and the single floating image to a preset three-dimensional standard space and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image includes: Registering the single reference image to the three-dimensional standard space, and determining a first initial registration parameter of the single reference image in the three-dimensional standard space; Registering the single floating image to the three-dimensional standard space, and determining a second initial registration parameter of the single floating image in the three-dimensional standard space; The first registration parameter is determined according to the first initial registration parameter and the second initial registration parameter.

3. The method according to claim 2, characterized in that The step of registering the single reference image to the three-dimensional standard space and determining a first initial registration parameter of the single reference image in the three-dimensional standard space includes: Inputting the single reference image into a preset neural network, performing prediction processing on the registration parameters of the single reference image in the three-dimensional standard space, and determining the first initial registration parameters; The neural network is trained based on multiple single training images and standard registration parameters corresponding to the single training images, and the standard registration parameters are registration parameters of the single training image in the three-dimensional standard space.

4. The method according to claim 3, characterized in that: The method of acquiring the single training image and the standard registration parameters corresponding to the single training image includes: Acquire a three-dimensional training image, and register the three-dimensional training image into the three-dimensional standard space; Acquire a single training image from the three-dimensional training images registered to the three-dimensional standard space, and obtain a translation amount and an angle corresponding to the single training image; Parameter conversion is performed on the translation amount and angle corresponding to the single training image to obtain the standard registration parameters corresponding to the single training image.

5. The method according to any one of claims 1 to 4, characterized in that: The performing registration processing on the three-dimensional reference image and the three-dimensional floating image according to the first registration parameter to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image comprises: Performing registration processing on the three-dimensional floating image using the first registration parameter to determine a transformed three-dimensional floating image; According to the respective corresponding fields of view of the three-dimensional reference image and the transformed three-dimensional floating image, one of the three-dimensional reference image and the transformed three-dimensional floating image is determined as a target three-dimensional image and the other is determined as a non-target three-dimensional image; the field of view of the target three-dimensional image is smaller than the field of view of the non-target three-dimensional image; performing redundancy removal processing on the non-target three-dimensional image; The non-target three-dimensional image after the redundancy removal processing and the target three-dimensional image are registered to determine a second registration parameter between the three-dimensional reference image and the three-dimensional floating image.

6. The method according to claim 5, characterized in that The performing redundancy removal processing on the non-target three-dimensional image includes: According to the field of view of the target three-dimensional image, redundancy removal processing is performed in the non-target three-dimensional image.

7. The method according to claim 5, characterized in that The performing registration processing on the non-target three-dimensional image after the redundancy removal processing and the target three-dimensional image to determine the second registration parameter between the three-dimensional reference image and the three-dimensional floating image includes: Calculating a similarity measure between the non-target three-dimensional image after the redundancy removal process and the target three-dimensional image; The non-target three-dimensional image after redundancy removal and the target three-dimensional image are registered according to the similarity measure to determine the second registration parameter.

8. A medical image registration device, characterized in that: The device comprises: An image acquisition module, used to acquire a single reference image and a single floating image; the single reference image is an image in the three-dimensional reference image, and the single floating image is an image in the three-dimensional floating image; A first registration module, used for registering the single reference image and the single floating image to a preset three-dimensional standard space, and determining a first registration parameter between the three-dimensional reference image and the three-dimensional floating image; The second registration module is used to perform registration processing on the 3D reference image and the 3D floating image according to the first registration parameter to determine a second registration parameter between the 3D reference image and the 3D floating image.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images