Medical image registration method and device, computer equipment and storage medium

Abstract

The invention relates to a medical image registration method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a tumor image and a functional map image; the tumor image is a part image with a tumor area; performing velocity field calculation on the tumor image and the functional map image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the interior of a tumor area in the tumor image; the tumor external velocity field is a velocity field for the exterior of a tumor area in the tumor image; and according to the tumor external velocity field and the tumor internal velocity field, registering the functional atlas image to the tumor image. By adopting the method, the standard functional map partition space can be accurately registered into the placeholder tumor image data.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a medical image registration method, apparatus, computer equipment, storage medium and computer program product. Background Art

[0002] Tumors are neoplasms formed by the proliferation of local tissue cells under the action of various tumorigenic factors. In clinical practice, multimodal magnetic resonance imaging is one of the main methods for tumor screening and diagnosis. By spatially registering the standard functional atlas partitions to the tumor imaging data, the matching of the anatomical structure and function of the site is completed, which is of great significance for the diagnosis and treatment of the patient's condition.

[0003] However, since the space-occupying effect of the tumor will cause non-conservative deformation of the surrounding tissue, it is difficult to accurately register the standard functional atlas partition space to the space-occupying tumor image data in the related art. Therefore, a solution is needed to accurately register the standard functional atlas partition space to the space-occupying tumor image data, so as to reduce the impact of the tumor space-occupying lesions on the operation of completing the matching of anatomical structure and function. Summary of the invention

[0004] Based on this, it is necessary to provide a medical image registration method, apparatus, computer equipment, computer-readable storage medium and computer program product that can more accurately spatially register standard functional map partitions to space-occupying tumor image data in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a medical image registration method. The method comprises:

[0006] Obtaining a tumor image and a functional atlas image; the tumor image is an image of a region where a tumor exists;

[0007] The velocity field is calculated for the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field; the tumor internal velocity field is the velocity field for the inside of the tumor area in the tumor image; the tumor external velocity field is the velocity field for the outside of the tumor area in the tumor image;

[0008] The functional atlas image is registered to the tumor image based on the tumor external velocity field and the tumor internal velocity field.

[0009] In one embodiment, the velocity field outside the tumor includes a forward velocity field outside the tumor; the velocity field inside the tumor includes a forward velocity field inside the tumor;

[0010] According to the velocity field outside the tumor and the velocity field inside the tumor, the functional atlas image is registered to the tumor image, including:

[0011] The forward velocity field outside the tumor and the forward velocity field inside the tumor are fused to obtain the forward velocity field;

[0012] According to the forward velocity field, the functional atlas image and the tumor image are iteratively registered until the preset iteration end condition is met.

[0013] In one embodiment, the iterative registration of the functional atlas image and the tumor image according to the forward velocity field until a preset iteration termination condition is met includes:

[0014] Integrate the forward velocity field used in the current iteration in the time domain to obtain the current forward deformation field;

[0015] Driving the functional atlas image to deform according to the current positive deformation field to obtain a new functional atlas image;

[0016] According to the first similarity between the new functional atlas image and the tumor image, it is determined whether the iteration end condition is satisfied.

[0017] In one embodiment, when it is determined that the iteration end condition is not satisfied, the iterative registration of the functional atlas image and the tumor image until a preset iteration end condition is satisfied includes:

[0018] The new functional atlas image is used as the functional atlas image, and the step of performing velocity field calculation on the tumor image and the functional atlas image to obtain a forward velocity field outside the tumor and a forward velocity field inside the tumor is returned until the iteration end condition is met.

[0019] In one embodiment, the tumor external velocity field also includes a tumor external reverse velocity field; the tumor internal velocity field also includes a tumor internal reverse velocity field;

[0020] The registering the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field comprises:

[0021] fusing the reverse velocity field outside the tumor and the reverse velocity field inside the tumor to obtain a reverse velocity field;

[0022] The functional atlas image and the tumor image are iteratively registered according to the forward velocity field and the reverse velocity field until a preset iteration end condition is met.

[0023] In one embodiment, the iterative registration of the functional atlas image and the tumor image according to the forward velocity field and the reverse velocity field until a preset iteration end condition is met includes:

[0024] According to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration, the current averaged forward velocity field and the current averaged reverse velocity field are obtained;

[0025] The functional atlas image and the tumor image are iteratively registered according to the current averaged forward velocity field and the current averaged reverse velocity field until a preset iteration end condition is met.

[0026] In one embodiment, obtaining the current averaged forward velocity field and the current averaged reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration comprises:

[0027] Obtaining the current averaged forward velocity field according to a difference obtained by subtracting the reverse velocity field used in the current iteration from the forward velocity field used in the current iteration;

[0028] The current averaged reverse velocity field is obtained according to a difference obtained by subtracting the forward velocity field used in the current iteration from the reverse velocity field used in the current iteration.

[0029] In one embodiment, the iterative registration of the functional atlas image and the tumor image according to the current averaged forward velocity field and the current averaged reverse velocity field until a preset iteration termination condition is met includes:

[0030] Integrating the current averaged forward velocity field and the current averaged reverse velocity field in the time domain to obtain a current forward deformation field and a current reverse deformation field;

[0031] Driving the functional atlas image to deform according to the current positive deformation field to obtain a new functional atlas image;

[0032] driving the tumor image to deform according to the current inverse deformation field to obtain a new tumor image;

[0033] Whether the iteration end condition is satisfied is determined according to a first similarity between the new functional atlas image and the tumor image and a second similarity between the new tumor image and the functional atlas image.

[0034] In one embodiment, when it is determined that the iteration end condition is not satisfied, the iterative registration of the functional atlas image and the tumor image until a preset iteration end condition is satisfied includes:

[0035] The new functional atlas image is used as the functional atlas image, the new tumor image is used as the tumor image, and the step of performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field is returned until the iteration end condition is met.

[0036] In one embodiment, the performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field includes:

[0037] Performing velocity field calculation on the tumor image and the functional atlas image according to the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image, to obtain the external forward velocity field of the tumor and the external reverse velocity field of the tumor;

[0038] The image difference is the difference between the tumor image and the functional atlas image.

[0039] In one embodiment, the performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field includes:

[0040] Determine a centroid point of an internal tumor mask corresponding to the tumor image; the internal tumor mask corresponding to the tumor image represents the interior of the tumor region in the tumor image;

[0041] The reverse velocity field inside the tumor and the forward velocity field inside the tumor are determined according to the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the tumor image.

[0042] In one embodiment, the functional atlas image includes a deformed functional atlas image; and the step of acquiring the tumor image and the functional atlas image includes:

[0043] An initial functional atlas image is obtained, and the initial functional atlas image and the tumor image are input into a trained tumor registration network to obtain a forward deformation field and a deformed functional atlas image; wherein the deformed functional atlas image is obtained by deforming the initial functional atlas image using the forward deformation field.

[0044] In one of the embodiments, the method further comprises: acquiring a tumor sample image, a functional atlas sample image, and a gold standard deformation field;

[0045] According to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, and the gold standard deformation field, the tumor registration network to be trained is trained to obtain the trained tumor registration network;

[0046] The tumor interior mask corresponding to the tumor sample image represents the interior of the tumor region in the tumor sample image;

[0047] The tumor interior mask corresponding to the functional atlas sample image represents the interior of the tumor region in the functional atlas sample image.

[0048] In one embodiment, according to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, and the gold standard deformation field, a tumor registration network to be trained is trained to obtain the trained tumor registration network, including:

[0049] Input the tumor sample image and the functional atlas sample image into the tumor registration network to be trained to obtain a predicted deformation field and a deformed functional atlas sample image; wherein the deformed functional atlas sample image is obtained by deforming the functional atlas sample image by using the predicted deformation field;

[0050] Obtaining a network loss value for the tumor registration network to be trained according to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, the deformed functional atlas sample image, the predicted deformation field, and the gold standard deformation field; the network loss value carries at least one of the first supervisory signal or the second supervisory signal;

[0051] The first supervisory signal is used to characterize the similarity between the outside of the tumor region in the deformed functional atlas sample image and the outside of the tumor region in the tumor sample image;

[0052] The second supervisory signal is used to characterize the difference between the predicted deformation field and the gold standard deformation field when the deformation field is outside the tumor region in the functional atlas sample image;

[0053] The tumor registration network to be trained is trained according to the network loss value.

[0054] In a second aspect, the present application also provides a medical image registration device. The device comprises:

[0055] An acquisition module, used to acquire a tumor image and a functional atlas image; the tumor image is an image of a part where a tumor exists;

[0056] a calculation module, configured to perform velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the interior of the tumor region in the tumor image; and the tumor external velocity field is a velocity field for the exterior of the tumor region in the tumor image;

[0057] A registration module is used to register the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

[0058] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0059] Acquire a tumor image and a functional atlas image; the tumor image is an image of a region where a tumor exists;

[0060] Performing velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the inside of the tumor region in the tumor image; the tumor external velocity field is a velocity field for the outside of the tumor region in the tumor image;

[0061] The functional atlas image is registered to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

[0062] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0063] Acquire a tumor image and a functional atlas image; the tumor image is an image of a region where a tumor exists;

[0064] Performing velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the inside of the tumor region in the tumor image; the tumor external velocity field is a velocity field for the outside of the tumor region in the tumor image;

[0065] The functional atlas image is registered to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

[0066] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0067] Acquire a tumor image and a functional atlas image; the tumor image is an image of a region where a tumor exists;

[0068] Performing velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the inside of the tumor region in the tumor image; the tumor external velocity field is a velocity field for the outside of the tumor region in the tumor image;

[0069] The functional atlas image is registered to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

[0070] The above-mentioned medical image registration method, device, computer equipment, storage medium and computer program product obtain a tumor image and a functional atlas image; the tumor image is a site image where a tumor area exists; velocity field calculation is performed on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is the velocity field for the inside of the tumor area in the tumor image; the tumor external velocity field is the velocity field for the outside of the tumor area in the tumor image; and the functional atlas image is registered to the tumor image based on the tumor external velocity field and the tumor internal velocity field.

[0071] In this way, by calculating the velocity field of the tumor image and the functional atlas image, the internal velocity field of the tumor region in the tumor image and the external velocity field of the tumor region outside the tumor image are obtained, so as to align the functional atlas image to the tumor image where the tumor region exists. Since the space-occupying effect of the tumor will cause non-conservative deformation, the deformation of the regions inside and outside the tumor will be different, that is, the shape and structure of the tumor will change inside the tumor region, and the morphology and structure of the surrounding tissues outside the tumor region will change. Therefore, by using different registration modes inside and outside the tumor region, it is achieved that the deformation outside the tumor region and the deformation inside the tumor region are taken into account in the registration process, which effectively solves the problem of the space-occupying effect of the tumor, and then the standard functional atlas partition space can be accurately registered to the space-occupying tumor image data, completing the matching of the anatomical structure and function of the site. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of a process of a medical image registration method in an embodiment;

[0073] Figure 2A schematic diagram of a process for registering a brain function atlas image and a brain tumor image in one embodiment;

[0074] Figure 3 A schematic diagram of a process of a brain tumor image data registration method based on deep learning in one embodiment;

[0075] Figure 4 A schematic flow chart of a medical image registration method in another embodiment;

[0076] Figure 5 is a structural block diagram of a medical image registration device in one embodiment;

[0077] Figure 6 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0078] 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.

[0079] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0080] In one embodiment, Figure 1 As shown, a medical image registration method is provided. This embodiment uses the method applied to a computer device as an example. It can be understood that in actual applications, the computer device is a user terminal, and can also be implemented by an independent server or a server cluster composed of multiple servers. It can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0081] Step S110, acquiring a tumor image and a functional atlas image.

[0082] The tumor image may be an image of a region where the tumor exists.

[0083] The part image may be an image obtained by scanning the target part of the inspected object.

[0084] Among them, the mass effect of the tumor can cause non-conservative deformation of the surrounding tissues.

[0085] Among them, conservation and non-conservation are usually used to describe whether there is a change in the structure itself. For example, conservation means that the tissue has only deformed, but no new structure has been generated (that is, the topological structure remains unchanged). Non-conservation means that the tissue has a new structure, such as the growth of a tumor, the tissue has a tear, etc. (that is, the topological structure has changed).

[0086] Among them, the tumor image can be obtained by performing multimodal magnetic resonance imaging on the target part of the object being examined.

[0087] The tumor image corresponding to the object under examination can be pre-stored in the memory of the computer device, and the computer device can directly obtain it from the memory when needed. This embodiment does not limit the specific method of obtaining the tumor image, as long as its function can be achieved.

[0088] Among them, the functional atlas image is the functional atlas image corresponding to the target part.

[0089] For example, if the target part is the brain, then the corresponding tumor image can be a brain tumor image, which is a brain image of the tumor area; the functional atlas image corresponding to the brain is a brain functional atlas image. In the following embodiments, the target part is the brain, the tumor image is the brain tumor image, and the functional atlas image is the brain functional atlas image. It can be understood that the target part can also be other parts besides the brain.

[0090] The brain function atlas image is an image obtained based on a standard brain function atlas partition space. The standard brain function atlas partition space may be a spatial template that divides the brain into different regions and defines functional characteristics for each region.

[0091] Among them, for brain tumors, the mass effect of the tumor will cause non-conservative deformation of the surrounding brain tissue.

[0092] In a specific implementation, the computer device can obtain a brain function atlas image and a brain tumor image corresponding to the object being examined.

[0093] Step S120, performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field.

[0094] The tumor internal velocity field is a velocity field inside the tumor region in the tumor image.

[0095] The tumor external velocity field is a velocity field outside the tumor area in the tumor image.

[0096] The velocity field represents the displacement velocity per unit time of each pixel point in the image to the corresponding pixel point in another image during the registration process.

[0097] In a specific implementation, a computer device can perform velocity field calculation on a brain tumor image and a brain function map image to obtain a velocity field inside the tumor region in the brain tumor image as the tumor internal velocity field; and obtain a velocity field outside the tumor region in the brain tumor image as the tumor external velocity field.

[0098] Step S130, registering the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

[0099] In a specific implementation, a computer device can use a brain tumor image as a reference image and a brain function map image as a moving image; the brain function map image and the brain tumor image are registered according to the external velocity field and the internal velocity field of the tumor, and the brain function map image is registered to the brain tumor image to achieve spatial registration of the standard brain function map partitions to the space-occupying tumor imaging data, thereby completing the matching of brain anatomical structure and brain function.

[0100] In the above-mentioned medical image registration method, a tumor image and a functional atlas image are obtained; the tumor image is a partial image of a tumor area; velocity fields are calculated for the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the inside of the tumor area in the tumor image; the tumor external velocity field is a velocity field for the outside of the tumor area in the tumor image; and the functional atlas image is registered to the tumor image based on the tumor external velocity field and the tumor internal velocity field.

[0101] In this way, by calculating the velocity field of the tumor image and the functional atlas image, the internal velocity field of the tumor region in the tumor image and the external velocity field of the tumor region outside the tumor image are obtained, so as to align the functional atlas image to the tumor image where the tumor region exists. Since the space-occupying effect of the tumor will cause non-conservative deformation, the deformation of the regions inside and outside the tumor will be different, that is, the shape and structure of the tumor will change inside the tumor region, and the morphology and structure of the surrounding tissues outside the tumor region will change. Therefore, by using different registration modes inside and outside the tumor region, it is achieved that the deformation outside the tumor region and the deformation inside the tumor region are taken into account in the registration process, which effectively solves the problem of the space-occupying effect of the tumor, and then the standard functional atlas partition space can be accurately registered to the space-occupying tumor image data, completing the matching of the anatomical structure and function of the site.

[0102] When the local image is an image corresponding to the brain, the standard brain function atlas partition space can be accurately registered to the space-occupying brain tumor image data, thereby completing the matching of brain anatomical structure and brain function.

[0103] In one embodiment, the tumor external velocity field includes the tumor external forward velocity field; the tumor internal velocity field includes the tumor internal forward velocity field; according to the tumor external velocity field and the tumor internal velocity field, the functional atlas image is aligned to the tumor image, including: fusing the tumor external forward velocity field and the tumor internal forward velocity field to obtain the forward velocity field; according to the forward velocity field, iteratively aligning the functional atlas image and the tumor image until a preset iteration end condition is met.

[0104] The tumor external positive velocity field is the positive velocity field corresponding to the outside of the tumor area.

[0105] The forward velocity field inside the tumor is the forward velocity field corresponding to the interior of the tumor region.

[0106] The forward velocity field is the velocity field used to map the functional atlas image onto the tumor image.

[0107] In a specific implementation, the external velocity field of the tumor includes the external forward velocity field of the tumor; the internal velocity field of the tumor includes the internal forward velocity field of the tumor; in the process of registering the brain function map image to the brain tumor image based on the external velocity field of the tumor and the internal velocity field of the tumor, the computer device can fuse the external forward velocity field of the tumor and the internal forward velocity field of the tumor to obtain the forward velocity field, and iteratively register the brain function map image and the brain tumor image through the forward velocity field until a preset iteration end condition is met, so as to register the brain function map image to the brain tumor image.

[0108] In one of the embodiments, the forward velocity field outside the tumor and the forward velocity field inside the tumor are fused to obtain the forward velocity field, including: determining the product of the forward velocity field inside the tumor and the tumor internal mask corresponding to the tumor image as the first product; the tumor internal mask corresponding to the tumor image represents the inside of the tumor area in the tumor image; determining the product of the forward velocity field outside the tumor and the tumor external mask corresponding to the tumor image as the second product; the tumor external mask corresponding to the tumor image represents the outside of the tumor area in the tumor image; and determining the forward velocity field based on the sum of the first product and the second product.

[0109] In a specific implementation, when the computer device fuses the external forward velocity field of the tumor and the internal forward velocity field of the tumor to obtain the forward velocity field, the internal tumor mask corresponding to the brain tumor image can characterize the inside of the tumor region in the brain tumor image, and the computer device can determine the product of the internal forward velocity field of the tumor and the internal tumor mask corresponding to the brain tumor image, and use the product as the first product. At the same time, the external tumor mask corresponding to the brain tumor image can characterize the outside of the tumor region in the brain tumor image. The computer device can determine the product of the external forward velocity field of the tumor and the external tumor mask corresponding to the brain tumor image, and use the product as the second product.

[0110] In this way, the computer device can accurately determine the forward velocity field that fuses the forward velocity field outside the tumor and the forward velocity field inside the tumor according to the sum of the first product and the second product, and obtain a complete forward velocity field. Thus, the computer device can iteratively align the brain function atlas image and the brain tumor image according to the complete forward velocity field until the preset iteration end condition is met.

[0111] In practical applications, the tumor internal mask may be a mask image for the inside of the tumor region in the brain tumor image, and the mask image has the same size as the brain tumor image. Specifically, the tumor internal mask may be a binary image with the same size as the brain tumor image, and pixels with a pixel value of 1 in the tumor internal mask constitute the inside of the tumor region. The tumor external mask may be determined by the difference between the all-one matrix and the mask image. The size of the all-one matrix may be equal to the size of the mask image.

[0112] Thus, taking Mask representation of the mask image inside the tumor area in the brain tumor image as an example, the formula for determining the forward velocity field can be as follows:

[0113] Mask×Forward velocity field inside the tumor+(1-Mask)×Forward velocity field outside the tumor

[0114] Among them, "Mask×tumor internal forward velocity field" can represent the first product; "1" can represent the all-one matrix; "(1-Mask)" can represent the tumor external mask; "(1-Mask)×tumor external forward velocity field" can represent the second product.

[0115] It should be noted that in order to make the fusion more natural, the mask image can be smoothed during the fusion process.

[0116] In the technical solution of this embodiment, the tumor external velocity field includes the tumor external forward velocity field; the tumor internal velocity field includes the tumor internal forward velocity field; the forward velocity field is obtained by fusing the tumor external forward velocity field and the tumor internal forward velocity field; according to the forward velocity field, the functional atlas image and the tumor image are iteratively registered until the preset iteration end condition is met. In this way, since the forward velocity field fuses the tumor internal velocity field for the tumor region in the tumor image and the tumor external velocity field for the tumor region outside the tumor image, the forward velocity field is a complete velocity field for the functional atlas image. In the process of using the forward velocity field for registration, different registration modes can be used inside and outside the tumor region, respectively, which not only considers the deformation outside the tumor region, but also considers the deformation inside the tumor region, which can effectively improve the registration accuracy of the process of registering the functional atlas image to the tumor image.

[0117] In one embodiment, the functional atlas image and the tumor image are iteratively aligned according to the forward velocity field until a preset iteration end condition is met, including: integrating the forward velocity field used for the current iteration in the time domain to obtain the current forward deformation field; driving the functional atlas image to deform according to the current forward deformation field to obtain a new functional atlas image; and judging whether the iteration end condition is met according to a first similarity between the new functional atlas image and the tumor image.

[0118] Among them, the current forward deformation field is the forward deformation field used in the current iteration.

[0119] The forward deformation field is a deformation field used to deform the moving image so as to map the moving image onto the reference image.

[0120] Among them, the deformation field can be used to represent the displacement or deformation of each pixel in the image in space.

[0121] In a specific implementation, when a computer device iteratively aligns a brain function map image and a brain tumor image according to a forward velocity field until a preset iteration end condition is met, the computer device can integrate the forward velocity field used for the current iteration in the time domain to obtain a current forward deformation field; and drive the brain function map image to deform according to the current forward deformation field to obtain a new brain function map image; in this way, the computer device can determine the similarity between the new brain function map image and the brain tumor image as the first similarity, and judge whether the iteration end condition is met according to the first similarity between the new brain function map image and the brain tumor image.

[0122] Among them, methods such as NCC (Normalized cross-correlation, a matching algorithm based on the similarity of pixel gray values), MI, MSE (mean square error) and the like may be used for similarity testing, and the specific method is not limited here.

[0123] The computer device may determine that the iteration end condition is satisfied when the first similarity reaches a preset convergence threshold; and may determine that the iteration end condition is not satisfied when the first similarity does not reach the preset convergence threshold.

[0124] In addition, the computer device may also determine that an iteration end condition is satisfied when the number of iterations reaches a preset number threshold.

[0125] When it is determined that the iteration end condition is not met, the computer device can use the new brain function atlas image as the brain function atlas image, return to the step of performing velocity field calculation on the brain tumor image and the brain function atlas image, and obtain the forward velocity field outside the tumor and the forward velocity field inside the tumor, until the iteration end condition is met.

[0126] The technical solution of this embodiment is to obtain the current forward deformation field by integrating the forward velocity field used in the current iteration in the time domain; to drive the functional atlas image to deform according to the current forward deformation field to obtain a new functional atlas image; and to determine whether the iteration end condition is met according to the first similarity between the new functional atlas image and the tumor image. In this way, in the process of iteratively registering the functional atlas image and the tumor image using the forward velocity field, the functional atlas image is driven to deform according to the current forward deformation field obtained based on the forward velocity field used in the current iteration, and the similarity between the new functional atlas image and the tumor image is used as the iteration end judgment condition, so as to accurately determine whether the tumor image and the functional atlas image have completed the registration.

[0127] In one embodiment, the tumor external velocity field also includes the tumor external inverse velocity field; the tumor internal velocity field also includes the tumor internal inverse velocity field; according to the tumor external velocity field and the tumor internal velocity field, the functional atlas image is aligned to the tumor image, including: fusing the tumor external inverse velocity field and the tumor internal inverse velocity field to obtain the inverse velocity field; according to the forward velocity field and the inverse velocity field, iteratively aligning the functional atlas image and the tumor image until a preset iteration end condition is met.

[0128] The reverse velocity field outside the tumor is the reverse velocity field corresponding to the outside of the tumor area.

[0129] The reverse velocity field inside the tumor is the reverse velocity field corresponding to the inside of the tumor area.

[0130] The inverse velocity field is a velocity field used to map the tumor image onto the functional atlas image.

[0131] In a specific implementation, when a computer device aligns a brain function atlas image to a brain tumor image based on an external velocity field and an internal velocity field of the tumor, the computer device can fuse an external inverse velocity field of the tumor and an internal inverse velocity field of the tumor to obtain an inverse velocity field; in this way, the computer device can iteratively align the brain function atlas image and the brain tumor image based on the forward velocity field and the inverse velocity field until a preset iteration end condition is met.

[0132] In one of the embodiments, the inverse velocity field outside the tumor and the inverse velocity field inside the tumor are fused to obtain the inverse velocity field, including: determining the product of the inverse velocity field inside the tumor and the tumor internal mask corresponding to the tumor image as the third product; the tumor internal mask corresponding to the tumor image represents the inside of the tumor area in the tumor image; determining the product of the inverse velocity field outside the tumor and the tumor external mask corresponding to the tumor image as the fourth product; the tumor external mask corresponding to the tumor image represents the outside of the tumor area in the tumor image; and determining the inverse velocity field according to the sum of the third product and the fourth product.

[0133] In a specific implementation, when the computer device fuses the tumor external inverse velocity field and the tumor internal inverse velocity field to obtain the inverse velocity field, the tumor internal mask corresponding to the brain tumor image can characterize the inside of the tumor region in the brain tumor image, and the computer device can determine the product of the tumor internal inverse velocity field and the tumor internal mask corresponding to the brain tumor image, and use the product as the third product. At the same time, the tumor external mask corresponding to the brain tumor image can characterize the outside of the tumor region in the brain tumor image. The computer device can determine the product of the tumor external inverse velocity field and the tumor external mask corresponding to the brain tumor image, and use the product as the fourth product.

[0134] In this way, the computer device can accurately determine the reverse velocity field that fuses the reverse velocity field outside the tumor and the reverse velocity field inside the tumor according to the sum of the third product and the fourth product, and obtain a complete reverse velocity field. Thus, the computer device can iteratively align the brain function atlas image and the brain tumor image according to the complete reverse velocity field and the complete forward velocity field until the preset iteration end condition is met.

[0135] In practical applications, the tumor internal mask may be a mask image for the tumor region in the brain tumor image, and the mask image has the same size as the brain tumor image. The tumor external mask may be determined by the difference between the all-one matrix and the mask image. The size of the all-one matrix may be equal to the size of the mask image.

[0136] Thus, taking the mask image inside the tumor area in the brain tumor image as an example, the formula for determining the inverse velocity field can be as follows:

[0137] Mask×inverse velocity field inside the tumor+(1-Mask)×inverse velocity field outside the tumor

[0138] Among them, "Mask×interior inverse velocity field of tumor" can represent the third product; "1" can represent the all-one matrix; "(1-Mask)" can represent the external mask of tumor; "(1-Mask)×exterior inverse velocity field of tumor" can represent the fourth product.

[0139] It should be noted that in order to make the fusion more natural, the mask image can be smoothed during the fusion process.

[0140] In the technical solution of this embodiment, the tumor external velocity field also includes the tumor external reverse velocity field; the tumor internal velocity field also includes the tumor internal reverse velocity field; the reverse velocity field is obtained by fusing the tumor external reverse velocity field and the tumor internal reverse velocity field; the functional atlas image and the tumor image are iteratively registered according to the forward velocity field and the reverse velocity field until the preset iteration end condition is met. In this way, considering the contribution of the forward velocity field and the reverse velocity field together, the functional atlas image and the tumor image can be iteratively registered more accurately.

[0141] In one embodiment, the functional atlas image and the tumor image are iteratively registered according to the forward velocity field and the reverse velocity field until a preset iteration end condition is met, including: obtaining the current averaged forward velocity field and the current averaged reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration; and iteratively registering the functional atlas image and the tumor image according to the current averaged forward velocity field and the current averaged reverse velocity field until a preset iteration end condition is met.

[0142] In a specific implementation, since the functional atlas image and the tumor image need to be iteratively registered, the computer device iteratively registers the brain functional atlas image and the brain tumor image according to the forward velocity field and the reverse velocity field until the preset iteration end condition is met. The computer device can obtain the current average forward velocity field and the current average reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration; then, iteratively register the brain functional atlas image and the brain tumor image according to the current average forward velocity field and the current average reverse velocity field until the preset iteration end condition is met.

[0143] The current averaged forward velocity field and the current averaged reverse velocity field are obtained according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration, including: obtaining the current averaged forward velocity field according to the difference obtained by subtracting the reverse velocity field used in the current iteration from the forward velocity field used in the current iteration; and obtaining the current averaged reverse velocity field according to the difference obtained by subtracting the current forward velocity field from the current reverse velocity field.

[0144] In a specific implementation, in the process of obtaining the current averaged forward velocity field and the current averaged reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration, the computer device may subtract the reverse velocity field used in the current iteration from the forward velocity field used in the current iteration, and obtain the averaged forward velocity field used in the current iteration (i.e., the current averaged forward velocity field) according to the difference obtained by subtracting the reverse velocity field used in the current iteration from the forward velocity field used in the current iteration. Furthermore, the computer device may subtract the forward velocity field used in the current iteration from the reverse velocity field used in the current iteration, and obtain the averaged reverse velocity field used in the current iteration (i.e., the current averaged reverse velocity field) according to the difference obtained by subtracting the forward velocity field used in the current iteration from the reverse velocity field used in the current iteration.

[0145] Specifically, v direct Characterizes the forward velocity field used in the current iteration, expressed as v inverse Characterizes the inverse velocity field used in the current iteration, expressed as v average_direct Characterizes the current average forward velocity field, with v average_inverse Characterize the current average rear reverse velocity field, then the current average rear forward velocity field and the current average rear reverse velocity field can be obtained by the following formula:

[0146] v average_direct =(v direct -v inverse ) / 2

[0147] vaverage_invrtse =(v inverse -v direct ) / 2

[0148] In this way, after obtaining the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration, by obtaining the current averaged forward velocity field and the current averaged reverse velocity field as above, the current averaged forward velocity field and the current averaged reverse velocity field can be used as symmetrical forward velocity fields and reverse velocity fields, and the average contribution of the two velocity fields is considered together, so that in the process of iteratively aligning the brain function atlas image and the brain tumor image according to the current averaged forward velocity field and the current averaged reverse velocity field, the symmetrical force method is used for bilateral driving, and the forward and reverse aligned velocity fields can be obtained at the same time.

[0149] The technical solution of this embodiment is to obtain the current averaged forward velocity field and the current averaged reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration; and to iteratively align the functional atlas image and the tumor image according to the current averaged forward velocity field and the current averaged reverse velocity field until the preset iteration end condition is met. In this way, based on the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration, the current averaged forward velocity field and the current averaged reverse velocity field are obtained. The current averaged forward velocity field and the current averaged reverse velocity field can be used as symmetrical forward velocity fields and reverse velocity fields. Considering the average contribution of the two velocity fields together, the deformation relationship between the functional atlas image and the tumor image can be more comprehensively captured, so that the current averaged forward velocity field and the current averaged reverse velocity field can be used to iteratively align the functional atlas image and the tumor image more accurately.

[0150] In one embodiment, the functional atlas image and the tumor image are iteratively registered according to the current averaged forward velocity field and the current averaged reverse velocity field until a preset iteration end condition is met, including: integrating the current averaged forward velocity field and the current averaged reverse velocity field in the time domain to obtain the current forward deformation field and the current reverse deformation field; driving the functional atlas image to deform according to the current forward deformation field to obtain a new functional atlas image; driving the tumor image to deform according to the current reverse deformation field to obtain a new tumor image; judging whether the iteration end condition is met according to a first similarity between the new functional atlas image and the tumor image, and a second similarity between the new tumor image and the functional atlas image.

[0151] Among them, the current inverse deformation field is the inverse deformation field used in the current iteration.

[0152] The inverse deformation field is a deformation field used to deform the reference image so as to map the reference image onto the moving image.

[0153] In a specific implementation, the computer device iteratively aligns the brain function atlas image and the brain tumor image according to the current average rear forward velocity field and the current average rear reverse velocity field until a preset iteration end condition is met. The computer device can integrate the current average rear forward velocity field and the current average rear reverse velocity field in the time domain to obtain the current forward deformation field corresponding to the current average rear forward velocity field and the current reverse deformation field corresponding to the current average rear reverse velocity field.

[0154] In the process of integrating the current averaged forward velocity field and the current averaged reverse velocity field in the time domain to obtain the current forward deformation field and the current reverse deformation field, the computer device can ensure the regularity of the deformation field through differential isomorphism operation or deformation field smoothing step to obtain the current forward deformation field (which can be expressed as Disp direct characterization) and the current inverse deformation field (which can be represented by Disp inverse characterization).

[0155] Then, the computer device can drive the brain function atlas image to deform according to the current forward deformation field to obtain a new brain function atlas image; and drive the brain tumor image to deform according to the current reverse deformation field to obtain a new brain tumor image. In this way, the computer device can determine whether the iteration end condition is met according to the new brain function atlas image, the new brain tumor image, the brain tumor image and the brain function atlas image.

[0156] Specifically, in the process of judging whether the iteration end condition is satisfied according to the new brain function atlas image, the new brain tumor image, the brain tumor image and the brain function atlas image, the computer device can map the brain function atlas image onto the brain tumor image due to the forward velocity field, thereby obtaining a new brain function atlas image;

[0157] The inverse velocity field can be used to map the brain tumor image onto the brain function atlas image to obtain a new brain tumor image; the computer device can determine the similarity between the new brain function atlas image and the brain tumor image as the first similarity; and determine the similarity between the new brain tumor image and the brain function atlas image as the second similarity. Thus, the computer device can accurately determine whether the iteration end condition is met based on the first similarity and the second similarity to complete the registration.

[0158] The computer device may perform a similarity test on the new brain function atlas image and the brain tumor image to obtain a first similarity. Similarly, the computer device may perform a similarity test on the new brain tumor image and the brain function atlas image to obtain a second similarity.

[0159] Among them, methods such as NCC (Normalized cross-correlation, a matching algorithm based on the similarity of pixel gray values), MI, MSE (mean square error) and the like may be used for similarity testing, and the specific method is not limited here.

[0160] In a specific implementation, when a computer device determines whether an iteration end condition is met based on a first similarity between a new brain function map image and a brain tumor image, and based on a second similarity between the new brain tumor image and the brain function map image, the computer device may determine that the iteration end condition is met if the sum of the first similarity and the second similarity reaches a preset convergence threshold; and determine that the iteration end condition is not met if the sum of the first similarity and the second similarity does not reach the preset convergence threshold.

[0161] In this way, by judging whether the sum of the first similarity and the second similarity reaches the set convergence threshold, it can be accurately determined whether the brain tumor image and the brain function atlas image are fully registered.

[0162] It should be noted that, when the number of iterations reaches a preset iteration number threshold, it can be determined that the iteration end condition is met.

[0163] In one of the embodiments, when the iteration end condition is not met, the functional atlas image and the tumor image are iteratively aligned until the preset iteration end condition is met, including: taking the new functional atlas image as the functional atlas image, taking the new tumor image as the tumor image, returning to the step of calculating the velocity field of the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field, until the iteration end condition is met.

[0164] In a specific implementation, when it is determined that the iteration end condition is not met according to the first similarity between the new brain function atlas image and the brain tumor image, and the second similarity between the new brain tumor image and the brain function atlas image, the computer device can use the new brain function atlas image as the brain function atlas image and the new brain tumor image as the brain tumor image, and return to the step of calculating the velocity field of the brain tumor image and the brain function atlas image to obtain the tumor external velocity field and the tumor internal velocity field, until the iteration end condition is met. In this way, when the sum of the first similarity and the second similarity does not reach the set convergence threshold, iterative registration is continued, and the forward velocity field and the reverse velocity field are optimized again, so that the deformation relationship between the brain tumor image and the brain function atlas image can be more comprehensively captured to improve the registration accuracy between the brain tumor image and the brain function atlas image.

[0165] The technical solution of this embodiment is to obtain the current forward deformation field and the current reverse deformation field by integrating the current averaged forward velocity field and the current averaged reverse velocity field in the time domain; drive the functional atlas image to deform according to the current forward deformation field to obtain a new functional atlas image; drive the tumor image to deform according to the current reverse deformation field to obtain a new tumor image; and judge whether the iteration end condition is met according to the first similarity between the new functional atlas image and the tumor image, and the second similarity between the new tumor image and the functional atlas image. In this way, in the process of iterative registration, bilateral driving is realized by means of symmetrical force, and the velocity fields of forward and reverse registration can be continuously and simultaneously obtained to deform the functional atlas image and the tumor image, so as to realize the iterative registration of the functional atlas image and the tumor image efficiently and accurately.

[0166] In one embodiment, velocity field calculation is performed on the tumor image and the functional atlas image to obtain the velocity field outside the tumor and the velocity field inside the tumor, including: performing velocity field calculation on the tumor image and the functional atlas image according to the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image to obtain the forward velocity field outside the tumor and the reverse velocity field outside the tumor;

[0167] The image difference is the difference between the tumor image and the functional atlas image.

[0168] In a specific implementation, when the computer device calculates the velocity field of the tumor image and the functional atlas image to obtain the velocity field outside the tumor and the velocity field inside the tumor, the computer device can determine the difference between the brain tumor image and the brain functional atlas image as the image difference. In addition, the computer device can determine the gradient corresponding to the brain functional atlas image as the first gradient, and determine the gradient corresponding to the brain tumor image as the second gradient. In this way, the computer device can determine the external forward velocity field of the tumor and the external reverse velocity field of the tumor based on the image difference, the first gradient corresponding to the brain functional atlas image, and the second gradient corresponding to the brain tumor image.

[0169] In one embodiment, velocity fields of the tumor image and the functional atlas image are calculated based on the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image to obtain a forward velocity field outside the tumor and a reverse velocity field outside the tumor, including: determining a grayscale difference obtained by subtracting the tumor image from the functional atlas image as a first image difference; and determining a forward velocity field outside the tumor based on the first image difference and the second gradient.

[0170] In a specific implementation, the computer device calculates the velocity field of the brain tumor image and the brain function atlas image according to the image difference, the first gradient corresponding to the brain function atlas image, and the second gradient corresponding to the brain tumor image to obtain the external forward velocity field of the tumor and the external reverse velocity field of the tumor. For the external forward velocity field of the tumor, the computer device can determine the grayscale difference obtained by subtracting the brain tumor image from the brain function atlas image as the first image difference; the first image difference can represent the difference between the brain function atlas image and the brain tumor image; then, the computer device can determine the external forward velocity field of the tumor according to the first image difference and the second gradient.

[0171] Specifically, the following formula can be used to calculate the external forward velocity field of the tumor:

[0172]

[0173] Among them, g is the brain function atlas image as a moving image, f is the brain tumor image as a reference image, is the gradient of the reference image.

[0174] In one embodiment, velocity fields of the brain tumor image and the brain function map image are calculated based on image differences, a first gradient corresponding to the brain function map image, and a second gradient corresponding to the brain tumor image to obtain a forward velocity field outside the tumor and a reverse velocity field outside the tumor, including: determining a grayscale difference obtained by subtracting the brain function map image from the brain tumor image as a second image difference; and determining a reverse velocity field outside the tumor based on the second image difference and the first gradient.

[0175] In a specific implementation, the computer device calculates the velocity field of the brain tumor image and the brain function atlas image according to the image difference, the first gradient corresponding to the brain function atlas image, and the second gradient corresponding to the brain tumor image to obtain the forward velocity field outside the tumor and the reverse velocity field outside the tumor. For the reverse velocity field outside the tumor, the computer device can determine the difference obtained by subtracting the brain function atlas image from the brain tumor image as the second image difference; the second image difference can represent the difference between the brain tumor image and the brain function atlas image; and the reverse velocity field outside the tumor is determined according to the second image difference and the first gradient.

[0176] Specifically, for the formula for calculating the reverse velocity field outside the tumor, it is only necessary to replace f with g and g with f in the above formula for calculating the forward velocity field outside the tumor. Replace with (the gradient corresponding to the moving image) is enough.

[0177] The technical solution of this embodiment calculates the velocity field of the tumor image and the functional atlas image according to the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image, so as to obtain the external forward velocity field of the tumor and the external reverse velocity field of the tumor; wherein the image difference is the difference between the tumor image and the functional atlas image. In this way, the image details and edge information can be provided through the gradient information of the image, and combined with the image difference between the tumor image and the functional atlas image, the movement information outside the tumor area can be more accurately determined when the tumor image and the functional atlas image are registered, so as to more accurately obtain the external forward velocity field of the tumor and the external reverse velocity field of the tumor.

[0178] In one embodiment, velocity field calculation is performed on a tumor image and a functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field, including: determining a centroid point of a tumor internal mask corresponding to the tumor image; the tumor internal mask corresponding to the tumor image represents the interior of the tumor region in the tumor image; and determining a tumor internal reverse velocity field and a tumor internal forward velocity field according to the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the tumor image.

[0179] In a specific implementation, when the computer device calculates the velocity field of the brain tumor image and the brain function atlas image to obtain the external velocity field and the internal velocity field of the tumor, the computer device can determine the centroid point of the internal tumor mask corresponding to the brain tumor image. Specifically, the internal tumor mask corresponding to the brain tumor image represents the interior of the tumor area in the brain tumor image, and the size of the internal tumor mask can be the same as the size of the brain tumor image. Among them, the internal tumor mask can be a binary image, and the pixels with a pixel value of 1 in the internal tumor mask constitute the interior of the tumor area. Exemplarily, the coordinates of the centroid point of the internal tumor mask can be determined based on the coordinates of the pixel points with a pixel value of 1 in the internal tumor mask. After determining the centroid point of the internal tumor mask, the computer device can determine the position coordinates of the pixel points corresponding to the centroid point of the internal tumor mask in the brain function atlas image as the coordinates of the seed points to be placed.

[0180] In this way, the computer device can determine the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the brain tumor image, and determine the tumor internal velocity field based on the distance as an adjustment parameter of the velocity field, which includes the tumor internal reverse velocity field and the tumor internal forward velocity field.

[0181] It should be noted that since the tumor internal mask is a binary image, the pixels with a pixel value of 1 in the tumor internal mask constitute the interior of the tumor area. When the computer device determines the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the brain tumor image, it actually determines the distance between the centroid point and each pixel point with a pixel value of 1 in the tumor internal mask.

[0182] In a specific implementation, in the process of determining the reverse velocity field inside the tumor and the forward velocity field inside the tumor according to the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the brain tumor image, the computer device can determine the reverse velocity field inside the tumor used for a single iteration and the forward velocity field inside the tumor used for a single iteration according to the distance between the centroid point of the tumor internal mask corresponding to the brain tumor image and each pixel point in the tumor internal mask, and a preset number of iterations.

[0183] Specifically, the computer device can determine the total velocity field inside the tumor based on the distance between the centroid point of the tumor internal mask corresponding to the brain tumor image and each pixel point in the tumor internal mask, and then divide the total velocity field inside the tumor by a preset number of iterations to obtain the tumor internal reverse velocity field used for a single iteration and the tumor internal forward velocity field used for a single iteration.

[0184] For example, the formula for calculating the forward velocity field inside the tumor for a single iteration can be as follows:

[0185]

[0186] in, It represents the distance between the centroid point and each pixel in the tumor mask corresponding to the brain tumor image, N it is the preset number of iterations, so the forward velocity of the tumor growth in each iteration process can be calculated to obtain the forward velocity field inside the tumor.

[0187] Specifically, during the iteration process, for the velocity field inside the tumor, it is necessary to divide the velocity field composed of the distance between the centroid point and each pixel point in the mask inside the tumor by the number of iterations, because each iteration is done slowly, rather than all at once. Therefore, the above formula for calculating the iterative velocity field distributes the distance in each iteration process as the true velocity field inside the tumor, and can accurately obtain the velocity field inside the tumor that meets the iterative registration scenario. For the calculation of the reverse velocity field inside the tumor in a single iteration, it is only necessary to invert the above formula for calculating the forward velocity field inside the tumor in a single iteration, that is, directly add a negative sign to the above formula for calculating the forward velocity field inside the tumor in a single iteration.

[0188] In this way, in view of the space-occupying effect of the tumor, in the process of calculating the external forward velocity field of the tumor, the external reverse velocity field of the tumor, the internal forward velocity field of the tumor, and the internal reverse velocity field of the tumor, the seed point growth method corresponding to the internal velocity field of the tumor and the classical registration method corresponding to the external velocity field of the tumor are coupled with each other, and different registration modes are considered to be used inside and outside the tumor, respectively. In the process of registration using the internal forward velocity field of the tumor and the internal reverse velocity field of the tumor, based on the seed point coordinates in the brain function atlas image, the tumor can be grown in the brain function atlas image while being registered, so as to obtain the tumor area in the brain function atlas image and solve the changes in the non-differential homology of the topological structure caused by the space-occupying effect of the tumor.

[0189] The above-mentioned lightweight seed growth and velocity field coupling registration process, this lightweight registration model, can ensure engineering and time performance.

[0190] In order to facilitate the understanding of those skilled in the art, Figure 2 A schematic diagram of registering a brain function atlas image and a brain tumor image is provided. Figure 2 As shown, the brain tumor image is input as the reference image, and the tumor internal mask representing the tumor area in the brain tumor image is determined as the tumor internal mask corresponding to the brain tumor image; the brain function atlas image is input as the moving image, and the seed point coordinates of the brain function atlas image are determined according to the centroid point of the tumor internal mask corresponding to the brain tumor image. According to the brain tumor image, the tumor internal mask corresponding to the brain tumor image, the brain function atlas image and the centroid point position, the tumor external forward velocity field, the tumor external reverse velocity field, the tumor internal forward velocity field and the tumor internal reverse velocity field are calculated, and the deformation field is obtained by integrating the velocity field. The final registration result is obtained according to the deformation field and output.

[0191] The forward velocity field can be obtained by fusing the forward velocity field outside the tumor and the forward velocity field inside the tumor; the reverse velocity field can be obtained by fusing the reverse velocity field outside the tumor and the reverse velocity field inside the tumor. The forward deformation field is obtained by integrating the forward velocity field, and the reverse deformation field is obtained by integrating the reverse velocity field. The brain function atlas image and the brain tumor image are iteratively registered according to the forward deformation field to register the brain function atlas image to the brain tumor image. Alternatively, the brain function atlas image and the brain tumor image are iteratively registered according to the forward deformation field and the reverse deformation field to register the brain function atlas image to the brain tumor image.

[0192] It should be noted that the specific definition of the above steps can refer to the specific definition of a medical image registration method above.

[0193] In one embodiment, the functional atlas image includes a deformed functional atlas image; obtaining a tumor image and a functional atlas image includes: obtaining an initial functional atlas image, inputting the initial functional atlas image and the tumor image into a trained tumor registration network to obtain a forward deformation field and a deformed functional atlas image; the deformed functional atlas image is obtained by deforming the initial functional atlas image according to the forward deformation field.

[0194] Among them, the forward deformation field is determined by the trained tumor registration network based on the transformation required to map each pixel point of the initial functional atlas image to the corresponding position on the tumor image.

[0195] The trained tumor registration network is a pre-trained tumor-specific customized registration network. The tumor-specific customized registration network is a registration network based on deep learning and can be used for tumor registration tasks in medical images.

[0196] The initial functional atlas image may be pre-stored in a memory of a computer device, and the computer device can directly obtain it from the memory when needed. This embodiment does not limit the specific method of obtaining the initial functional atlas image, as long as its function can be achieved.

[0197] In a specific implementation, the functional atlas image includes a deformed functional atlas image. In the process of acquiring a brain tumor image and a brain function atlas image, the computer device can acquire an initial brain function atlas image, use the initial brain function atlas image as a moving image, use the brain tumor image as a reference image, and input the initial brain function atlas image and the brain tumor image into a trained tumor registration network. The trained tumor registration network deforms the initial brain function atlas image according to a forward deformation field calculated using the initial brain function atlas image as a moving image and the brain tumor image as a reference image, and obtains and outputs a deformed brain function atlas image and a forward deformation field.

[0198] Specifically, the forward deformation field is determined by the trained tumor registration network based on the transformation required to map each pixel point of the initial brain function atlas image to the corresponding position on the brain tumor image.

[0199] It is understandable that the trained tumor registration network can also output an inverse deformation field, which is used to deform the brain tumor image to deform the edge data in the brain tumor image to the brain function atlas image, such as mapping the blood vessel data and neuron data to the same atlas space for data analysis. Therefore, in addition to outputting the deformed brain function atlas image and the forward deformation field, the trained tumor registration network can also output at least one of the inverse deformation field and the deformed brain tumor image.

[0200] In this way, the deformed brain function atlas image output by the trained tumor registration network can be used as the first stage of registering the brain function atlas image to the brain tumor image. After the trained tumor registration network outputs the deformed brain function atlas image, the computer device can enter the second stage of registering the brain function atlas image to the brain tumor image, that is, the computer device can execute the steps of acquiring the brain tumor image and the brain function atlas image, performing velocity field calculation on the brain tumor image and the brain function atlas image to obtain the external velocity field of the tumor and the internal velocity field of the tumor; and registering the brain function atlas image to the brain tumor image according to the external velocity field of the tumor and the internal velocity field of the tumor.

[0201] In one embodiment, if Figure 3 As shown in FIG, a flowchart of a tumor image data registration method based on deep learning is provided. Figure 3 As shown, the method also includes: obtaining a tumor sample image, a functional atlas sample image and a gold standard deformation field; training a tumor registration network to be trained according to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, and the gold standard deformation field to obtain a trained tumor registration network; the tumor internal mask corresponding to the tumor sample image represents the interior of the tumor region in the tumor sample image; the tumor internal mask corresponding to the functional atlas sample image represents the interior of the tumor region in the functional atlas sample image.

[0202] The tumor registration network to be trained may adopt FCN (Fully Convolutional Network) or other deep learning networks, which are not specifically limited here.

[0203] The tumor sample image is a sample image of a part where a tumor exists.

[0204] The tumor sample images and the functional map sample images can be pre-stored in the memory of the computer device, and the computer device can directly obtain them from the memory when needed. This embodiment does not limit the specific method of obtaining the tumor sample images and the functional map sample images, as long as the functions can be achieved.

[0205] Correspondingly, when the target site is the brain, the tumor sample image may be a brain tumor sample image, and the function map sample image may be a brain function map sample image.

[0206] Taking the optimization of the forward deformation field predicted and output by the tumor registration network to be trained as an example, the computer device trains the tumor registration network to be trained according to the brain tumor sample image, the brain function atlas sample image, the tumor internal mask corresponding to the brain tumor sample image, the tumor internal mask corresponding to the brain function atlas sample image, and the gold standard deformation field, and in the process of obtaining the trained tumor registration network, the computer device can use the brain tumor sample image as a reference image and the brain function atlas sample image as a moving image, input the tumor registration network to be trained, and obtain the predicted deformation field and the deformed brain function atlas sample image. Among them, the predicted deformation field can be the forward deformation field predicted by the tumor registration network to be trained for registering the brain function atlas sample image to the brain tumor sample image.

[0207] Then, the computer device can obtain the network loss value for the tumor registration network to be trained based on the brain tumor sample image, the brain function map sample image, the tumor internal mask corresponding to the brain tumor sample image, the tumor internal mask corresponding to the brain function map sample image, the deformed brain function map sample image, the predicted deformation field and the gold standard deformation field.

[0208] The deformed functional map sample image is obtained by deforming the functional map sample image through predicting the deformation field.

[0209] The network loss value carries at least one of the first supervisory signal or the second supervisory signal.

[0210] The first supervisory signal is used to characterize the similarity between the outside of the tumor region in the deformed functional map sample image and the outside of the tumor region in the tumor sample image.

[0211] The second supervisory signal is used to characterize the difference between the predicted deformation field and the positive gold standard deformation field when the deformation field is outside the tumor region in the functional atlas sample image.

[0212] The computer device can train the tumor registration network to be trained according to the network loss value until the network loss value meets the preset training end condition, thereby obtaining the trained tumor registration network.

[0213] The computer device may use a preset loss function to determine the first supervisory signal and the second supervisory signal. For example, the computer device may use the Loss function of the image similarity, taking LNCC (Local Normalized Cross-Correlation) Loss as an example, that is, the local correlation coefficient, to calculate the similarity between the outside of the tumor region in the image after registration deformation, and obtain the first supervisory signal. When calculating the loss function, the similarity may be calculated based on the internal mask of the tumor, and only the outside of the tumor region is calculated.

[0214] For example, taking the optimization of the forward deformation field predicted by the tumor registration network to be trained as an example, based on the internal tumor mask corresponding to the brain tumor sample image and the internal tumor mask corresponding to the deformed brain function map sample image, the similarity between the outside of the tumor area in the deformed brain function map sample image and the outside of the tumor area in the brain tumor sample image can be calculated.

[0215] Taking the optimization of the inverse deformation field predicted by the trained tumor registration network as an example, based on the internal tumor mask corresponding to the deformed brain tumor sample image and the internal tumor mask corresponding to the brain function map sample image, the similarity between the outside of the tumor area in the brain function map sample image and the outside of the tumor area in the deformed brain tumor sample image can be calculated.

[0216] In addition, for the second supervisory signal, the computer device can use the gold standard L1Loss function of the deformation field to only calculate the loss function outside the tumor area. That is, taking the optimization of the forward deformation field predicted by the tumor registration network to be trained as an example, the computer device can use the L1Loss function to calculate the difference between the forward deformation field predicted and the forward gold standard deformation field for the outside of the tumor area in the brain function atlas sample image. Taking the optimization of the inverse deformation field predicted by the tumor registration network to be trained as an example, the computer device can use the L1Loss function to calculate the difference between the inverse deformation field predicted and the inverse gold standard deformation field for the outside of the tumor area in the brain tumor sample image.

[0217] Among them, the principle of optimizing the inverse deformation field predicted by the tumor registration network to be trained is similar to the principle of optimizing the forward deformation field, which will not be repeated here.

[0218] It should be noted that when the network loss value carries the first supervisory signal and the second supervisory signal, the network loss value can be determined by the average of the loss value corresponding to the first supervisory signal and the loss value corresponding to the second supervisory signal. It can also be determined by the weighted average of the loss value corresponding to the first supervisory signal and the second supervisory signal.

[0219] In addition, for the final loss function, in addition to the above-mentioned image similarity Loss function and the gold standard L1Loss function of the deformation field, you can also choose the Smooth Loss function, Inverse Loss function, and Jac Loss function.

[0220] The Smooth Loss function is a loss constrained by the smoothness of the deformation field. When calculating the loss function, different smoothing coefficients can be set inside and outside the tumor area according to the internal mask of the tumor. The Inverse Loss function is a loss function with inverse consistency, which ensures that the forward and reverse deformation fields are reversible. For Jac, that is, Jacobi, the normal Jacobi constraint is used.

[0221] In the actual training process, for these loss functions, one of them can be selected, or a weighted average of multiple loss functions can be selected in a weighted manner. In an optional embodiment of the present application, the loss function can select at least one of the Loss function of image similarity and the gold standard L1Loss function of the deformation field. Alternatively, at least one of the Loss function of image similarity and the gold standard L1Loss function of the deformation field, and at least one of the Smooth Loss function, Inverse Loss function, and Jac Loss function can be selected.

[0222] in, Figure 3 The first loss function in may be a Smooth Loss function, the second loss function may be an Inverse Loss function, the third loss function may be an LNCC Loss function, the fourth loss function may be a Jac Loss function, and the fifth loss function may be an L1Loss function.

[0223] The design of the weights of each loss function needs to be based on the specific problem and training set. During the actual training process, specific adjustments should be made based on the loss function curve and the registration results of the validation set.

[0224] It should be noted that the data preprocessing (scalp removal, cropping, sampling, etc.) and matrix and deformation field post-processing are also included in the solution and will not be repeated here.

[0225] Thus, in order to solve the large deformation problem caused by the mass effect of the tumor, the technical solution of this embodiment trains the tumor image data by deep learning alone. In order to solve the large deformation problem, the gold standard deformation field of large deformation is used as supervision, and the tumor internal mask is used to solve the problem of inconsistent deformation degree between the inside and outside of the tumor.

[0226] In another embodiment, Figure 4 As shown, a medical image registration method is provided, which is described by taking the method applied to a computer device as an example, and includes the following steps:

[0227] Step S402, obtaining a tumor sample image and a functional atlas sample image.

[0228] Step S404: input the tumor sample image and the functional atlas sample image into the tumor registration network to be trained to obtain the predicted deformation field and the deformed functional atlas sample image.

[0229] Step S406, obtaining a network loss value for the tumor registration network to be trained according to the tumor sample image, the functional map sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional map sample image, the deformed functional map sample image, the predicted deformation field and the gold standard deformation field.

[0230] Step S408: training the tumor registration network to be trained according to the network loss value to obtain a trained tumor registration network.

[0231] Step S410, obtaining an initial functional atlas image, inputting the initial functional atlas image and the tumor image into the trained tumor registration network, and obtaining a deformed functional atlas image.

[0232] Step S412, performing velocity field calculation on the tumor image and the functional atlas image according to the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image, to obtain a forward velocity field outside the tumor and a reverse velocity field outside the tumor.

[0233] Step S414, determining the centroid point of the tumor internal mask corresponding to the tumor image.

[0234] Step S416, determining the reverse velocity field inside the tumor and the forward velocity field inside the tumor according to the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the tumor image.

[0235] Step S418, fusing the forward velocity field outside the tumor and the forward velocity field inside the tumor to obtain the forward velocity field.

[0236] Step S420: fusing the reverse velocity field outside the tumor and the reverse velocity field inside the tumor to obtain a reverse velocity field.

[0237] Step S422, iteratively aligning the functional atlas image and the tumor image according to the forward velocity field and the reverse velocity field until a preset iteration end condition is met.

[0238] In this way, through the first stage of deep learning-based brain tumor image data registration and the second stage of lightweight seed growth and velocity field coupling registration process, not only the large deformation problem caused by the tumor's space effect can be solved, but also the problem of non-differential isomorphic changes in the topological structure caused by the tumor's space occupancy can be solved.

[0239] It should be noted that the specific definition of the above steps can refer to the specific definition of a medical image registration method above.

[0240] 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.

[0241] 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 mentioned 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 a medical image registration method above, and will not be repeated here.

[0242] In one embodiment, Figure 5 As shown, a medical image registration device is provided, comprising: an acquisition module 510, a calculation module 520 and a registration module 530, wherein:

[0243] The input module 510 is used to obtain a tumor image and a functional atlas image; the tumor image is an image of a part where a tumor exists.

[0244] The calculation module 520 is used to perform velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the inside of the tumor area in the tumor image; and the tumor external velocity field is a velocity field for the outside of the tumor area in the tumor image.

[0245] The registration module 530 is used to register the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

[0246] In one of the embodiments, the tumor external velocity field includes a tumor external forward velocity field; the tumor internal velocity field includes a tumor internal forward velocity field; the registration module 530 is specifically used to fuse the tumor external forward velocity field and the tumor internal forward velocity field to obtain a forward velocity field; according to the forward velocity field, the functional atlas image and the tumor image are iteratively registered until a preset iteration end condition is met.

[0247] In one embodiment, the registration module 530 is specifically used to integrate the forward velocity field used in the current iteration in the time domain to obtain a current forward deformation field; drive the functional atlas image to deform according to the current forward deformation field to obtain a new functional atlas image; and determine whether the iteration end condition is met based on a first similarity between the new functional atlas image and the tumor image.

[0248] In one of the embodiments, when it is determined that the iteration end condition is not met, the registration module 530 is specifically used to use the new functional atlas image as the functional atlas image, return to the step of performing velocity field calculation on the tumor image and the functional atlas image to obtain the forward velocity field outside the tumor and the forward velocity field inside the tumor, until the iteration end condition is met.

[0249] In one of the embodiments, the tumor external velocity field also includes a tumor external reverse velocity field; the tumor internal velocity field also includes a tumor internal reverse velocity field; the registration module 530 is specifically used to fuse the tumor external reverse velocity field and the tumor internal reverse velocity field to obtain a reverse velocity field; according to the forward velocity field and the reverse velocity field, the functional atlas image and the tumor image are iteratively registered until a preset iteration end condition is met.

[0250] In one embodiment, the registration module 530 is specifically used to obtain the current average forward velocity field and the current average reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration; and iteratively align the functional atlas image and the tumor image according to the current average forward velocity field and the current average reverse velocity field until a preset iteration end condition is met.

[0251] In one embodiment, the alignment module 530 is specifically used to obtain the current average forward velocity field according to the difference obtained by subtracting the reverse velocity field used in the current iteration from the forward velocity field used in the current iteration; and to obtain the current average reverse velocity field according to the difference obtained by subtracting the forward velocity field used in the current iteration from the reverse velocity field used in the current iteration.

[0252] In one embodiment, the registration module 530 is specifically used to integrate the current averaged forward velocity field and the current averaged reverse velocity field in the time domain to obtain a current forward deformation field and a current reverse deformation field; drive the functional atlas image to deform according to the current forward deformation field to obtain a new functional atlas image; drive the tumor image to deform according to the current reverse deformation field to obtain a new tumor image; and determine whether the iteration end condition is met based on a first similarity between the new functional atlas image and the tumor image, and a second similarity between the new tumor image and the functional atlas image.

[0253] In one of the embodiments, when it is determined that the iteration end condition is not met, the registration module 530 is specifically used to use the new functional atlas image as the functional atlas image, use the new tumor image as the tumor image, and return to the step of calculating the velocity field of the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field until the iteration end condition is met.

[0254] In one embodiment, the calculation module 520 is specifically used to perform velocity field calculation on the tumor image and the functional atlas image according to the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image, to obtain the forward velocity field outside the tumor and the reverse velocity field outside the tumor; the image difference is the difference between the tumor image and the functional atlas image.

[0255] In one embodiment, the calculation module 520 is specifically used to determine the centroid point of the tumor internal mask corresponding to the tumor image; the tumor internal mask corresponding to the tumor image represents the interior of the tumor area in the tumor image; based on the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the tumor image, the reverse velocity field inside the tumor and the forward velocity field inside the tumor are determined.

[0256] In one embodiment, the calculation module 520 is specifically used to determine the tumor internal reverse velocity field used for a single iteration and the tumor internal forward velocity field used for a single iteration based on the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the tumor image, and a preset number of iterations.

[0257] In one embodiment, the functional atlas image includes a deformed functional atlas image; the acquisition module 510 is specifically used to acquire an initial functional atlas image, input the initial functional atlas image and the tumor image into a trained tumor registration network, and obtain a forward deformation field and a deformed functional atlas image; wherein the deformed functional atlas image is obtained by deforming the initial functional atlas image using the forward deformation field.

[0258] In one embodiment, the device also includes: a training module, used to obtain a tumor sample image, a functional map sample image and a gold standard deformation field; according to the tumor sample image, the functional map sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional map sample image, and the gold standard deformation field, the tumor registration network to be trained is trained to obtain the trained tumor registration network; the tumor internal mask corresponding to the tumor sample image represents the interior of the tumor area in the tumor sample image; the tumor internal mask corresponding to the functional map sample image represents the interior of the tumor area in the functional map sample image.

[0259] In one embodiment, the training module is specifically used to input the tumor sample image and the functional map sample image into the tumor registration network to be trained to obtain a predicted deformation field and a deformed functional map sample image; wherein the deformed functional map sample image is obtained by deforming the functional map sample image by the predicted deformation field; according to the tumor sample image, the functional map sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional map sample image, the deformed functional map sample image, the predicted deformation field, the functional map sample image, the deformed The method comprises the following steps: comparing the deformation field and the gold standard deformation field to obtain a network loss value for the tumor registration network to be trained; the network loss value carries at least one of a first supervisory signal or a second supervisory signal; the first supervisory signal is used to characterize the similarity between the outside of the tumor region in the deformed functional atlas sample image and the outside of the tumor region in the tumor sample image; the second supervisory signal is used to characterize the difference between the predicted deformation field and the gold standard deformation field when targeting the outside of the tumor region in the functional atlas sample image; and according to the network loss value, the tumor registration network to be trained is trained.

[0260] 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.

[0261] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. 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 input / output interface of the computer device is used to exchange information between the processor and the external device. 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 implemented 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 unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen 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 covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0262] Those skilled in the art will understand that Figure 6 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.

[0263] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0264] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0265] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0266] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0267] 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.

[0268] 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.

[0269] 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 construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications 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 tumor image and a functional atlas image; the tumor image is an image of a region where a tumor exists; Performing velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the inside of the tumor region in the tumor image; the tumor external velocity field is a velocity field for the outside of the tumor region in the tumor image; The functional atlas image is registered to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

2. The method according to claim 1, characterized in that The tumor external velocity field includes the tumor external forward velocity field; the tumor internal velocity field includes the tumor internal forward velocity field; The registering the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field comprises: fusing the external forward velocity field of the tumor and the internal forward velocity field of the tumor to obtain a forward velocity field; The functional atlas image and the tumor image are iteratively registered according to the forward velocity field until a preset iteration end condition is met.

3. The method according to claim 2, characterized in that The iterative registration of the functional atlas image and the tumor image according to the forward velocity field until a preset iteration end condition is met includes: Integrate the forward velocity field used in the current iteration in the time domain to obtain the current forward deformation field; Driving the functional atlas image to deform according to the current positive deformation field to obtain a new functional atlas image; Whether the iteration end condition is satisfied is determined according to a first similarity between the new functional atlas image and the tumor image.

4. The method according to claim 3, characterized in that In the case where it is determined that the iteration end condition is not satisfied, the iterative registration of the functional atlas image and the tumor image until a preset iteration end condition is satisfied includes: The new functional atlas image is used as the functional atlas image, and the step of performing velocity field calculation on the tumor image and the functional atlas image to obtain a forward velocity field outside the tumor and a forward velocity field inside the tumor is returned until the iteration end condition is met.

5. The method according to claim 2, characterized in that: The tumor external velocity field also includes the tumor external reverse velocity field; the tumor internal velocity field also includes the tumor internal reverse velocity field; The registering the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field comprises: fusing the reverse velocity field outside the tumor and the reverse velocity field inside the tumor to obtain a reverse velocity field; The functional atlas image and the tumor image are iteratively registered according to the forward velocity field and the reverse velocity field until a preset iteration end condition is met.

6. The method according to claim 5, characterized in that The iterative registration of the functional atlas image and the tumor image according to the forward velocity field and the reverse velocity field until a preset iteration end condition is met includes: According to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration, the current averaged forward velocity field and the current averaged reverse velocity field are obtained; The functional atlas image and the tumor image are iteratively registered according to the current averaged forward velocity field and the current averaged reverse velocity field until a preset iteration end condition is met.

7. The method according to claim 6, characterized in that The step of obtaining the current averaged forward velocity field and the current averaged reverse velocity field according to the difference between the forward velocity field used in the current iteration and the reverse velocity field used in the current iteration comprises: Obtaining the current averaged forward velocity field according to a difference obtained by subtracting the reverse velocity field used in the current iteration from the forward velocity field used in the current iteration; The current averaged reverse velocity field is obtained according to a difference obtained by subtracting the forward velocity field used in the current iteration from the reverse velocity field used in the current iteration.

8. The method according to claim 7, characterized in that The iterative registration of the functional atlas image and the tumor image according to the current averaged forward velocity field and the current averaged reverse velocity field until a preset iteration end condition is met includes: Integrating the current averaged forward velocity field and the current averaged reverse velocity field in the time domain to obtain a current forward deformation field and a current reverse deformation field; Driving the functional atlas image to deform according to the current positive deformation field to obtain a new functional atlas image; driving the tumor image to deform according to the current inverse deformation field to obtain a new tumor image; Whether the iteration end condition is satisfied is determined according to a first similarity between the new functional atlas image and the tumor image and a second similarity between the new tumor image and the functional atlas image.

9. The method according to claim 8, characterized in that In the case where it is determined that the iteration end condition is not satisfied, the iterative registration of the functional atlas image and the tumor image until a preset iteration end condition is satisfied includes: The new functional atlas image is used as the functional atlas image, the new tumor image is used as the tumor image, and the step of performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field is returned until the iteration end condition is met.

10. The method according to claim 5, characterized in that The performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field comprises: Performing velocity field calculation on the tumor image and the functional atlas image according to the image difference, the first gradient corresponding to the functional atlas image, and the second gradient corresponding to the tumor image, to obtain the external forward velocity field of the tumor and the external reverse velocity field of the tumor; The image difference is the difference between the tumor image and the functional atlas image.

11. The method according to claim 5, characterized in that The performing velocity field calculation on the tumor image and the functional atlas image to obtain the tumor external velocity field and the tumor internal velocity field comprises: Determine a centroid point of an internal tumor mask corresponding to the tumor image; the internal tumor mask corresponding to the tumor image represents the interior of the tumor region in the tumor image; The reverse velocity field inside the tumor and the forward velocity field inside the tumor are determined according to the distance between the centroid point and each pixel point in the tumor internal mask corresponding to the tumor image.

12. The method according to claim 1, characterized in that The functional atlas image includes a deformed functional atlas image; and the step of acquiring the tumor image and the functional atlas image includes: An initial functional atlas image is obtained, and the initial functional atlas image and the tumor image are input into a trained tumor registration network to obtain a forward deformation field and a deformed functional atlas image; wherein the deformed functional atlas image is obtained by deforming the initial functional atlas image using the forward deformation field.

13. The method according to claim 12, characterized in that The method further comprises: Obtain tumor sample images, functional atlas sample images, and gold standard deformation fields; According to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, and the gold standard deformation field, the tumor registration network to be trained is trained to obtain the trained tumor registration network; The tumor interior mask corresponding to the tumor sample image represents the interior of the tumor region in the tumor sample image; The tumor interior mask corresponding to the functional atlas sample image represents the interior of the tumor region in the functional atlas sample image.

14. The method according to claim 13, characterized in that The step of training the tumor registration network to be trained according to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, and the gold standard deformation field to obtain the trained tumor registration network includes: Input the tumor sample image and the functional atlas sample image into the tumor registration network to be trained to obtain a predicted deformation field and a deformed functional atlas sample image; wherein the deformed functional atlas sample image is obtained by deforming the functional atlas sample image by using the predicted deformation field; Obtaining a network loss value for the tumor registration network to be trained according to the tumor sample image, the functional atlas sample image, the tumor internal mask corresponding to the tumor sample image, the tumor internal mask corresponding to the functional atlas sample image, the deformed functional atlas sample image, the predicted deformation field, and the gold standard deformation field; the network loss value carries at least one of the first supervisory signal or the second supervisory signal; The first supervisory signal is used to characterize the similarity between the outside of the tumor region in the deformed functional atlas sample image and the outside of the tumor region in the tumor sample image; The second supervisory signal is used to characterize the difference between the predicted deformation field and the gold standard deformation field when the deformation field is outside the tumor region in the functional atlas sample image; The tumor registration network to be trained is trained according to the network loss value.

15. A medical image registration device, characterized in that: The device comprises: An acquisition module, used to acquire a tumor image and a functional atlas image; the tumor image is an image of a part where a tumor exists; a calculation module, configured to perform velocity field calculation on the tumor image and the functional atlas image to obtain a tumor external velocity field and a tumor internal velocity field; the tumor internal velocity field is a velocity field for the interior of the tumor region in the tumor image; and the tumor external velocity field is a velocity field for the exterior of the tumor region in the tumor image; A registration module is used to register the functional atlas image to the tumor image according to the tumor external velocity field and the tumor internal velocity field.

16. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 14 are implemented.

17. 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 14 are implemented.

Images