Image registration method, CT image registration method, computer-aided diagnosis method and CT image registration system

Abstract

The embodiment of the invention provides an image registration method, a CT image registration method, a computer-aided diagnosis method and a CT image registration system. The image registration method comprises the following steps: acquiring a fixed image and a floating image; determining registration task information according to the fixed image and the floating image; inputting the fixed image, the floating image and the registration task information into an image registration model to obtain image adjustment information output by the image registration model; the floating image is adjusted according to the image adjusting information, a registered image is obtained, and the registered image and the fixed image are aligned in space. According to the general image registration method provided by the invention, the registration task information is input into the image registration model, so that the image registration model can execute various types of image registration tasks according to a single model framework, and the necessity of training a corresponding registration model for each registration task is reduced.

Description

Technical Field

[0001] Embodiments of this specification relate to the field of computer technology, and particularly to an image registration method, a computer-aided diagnosis method, and a CT image registration system. Background Art

[0002] With the development of computer technology, in the field of image processing, especially in the field of medical computer-aided diagnosis, learning-based registration technology has made significant progress, and its accuracy and efficiency have been significantly improved. However, learning-based registration technology still faces challenges in terms of generalization ability.

[0003] In multiple registration scenarios, it is inevitable to design and train multiple single registration networks for specific tasks. That is, generally, different object registrations require training targeted registration networks. When there is a new registration task or object to be registered, a new registration network needs to be retrained to adapt to this task or object, and retraining a new registration network takes time and resources, which is time-consuming and laborious. Summary of the Invention

[0004] In view of this, embodiments of this specification provide an image registration method, a CT image registration method, a computer-aided diagnosis method, and a CT image registration system. One or more embodiments of this specification also relate to an electronic device, a computer-readable storage medium, and a computer program product to solve the technical defects existing in the prior art.

[0005] According to a first aspect of the embodiments of this specification, an image registration method is provided, including: Obtain a fixed image and a floating image; Determine registration task information according to the fixed image and the floating image; Input the fixed image, the floating image, and the registration task information into an image registration model to obtain image adjustment information output by the image registration model; Adjust the floating image according to the image adjustment information to obtain a registered image, where the registered image and the fixed image are spatially aligned.

[0006] According to a second aspect of the embodiments of this specification, a training method for an image registration model is provided, including: Obtain a sample fixed image and a sample floating image, and determine sample registration task information according to the sample fixed image and the sample floating image; Input the sample fixed image, the sample floating image, and the sample registration task information into an image registration model to obtain predicted image adjustment information output by the image registration model; Adjust the sample floating image according to the predicted image adjustment information to obtain an adjusted registration image; Calculate a model loss value according to the sample fixed image and the adjusted registration image; Adjust the model parameters of the image registration model according to the model loss value, and continue to train the image registration model until the model training stop condition is reached.

[0007] According to the third aspect of the embodiments of the present specification, a CT image registration method is provided, including: Obtain a fixed CT image and a floating CT image; Determine registration task information according to the fixed CT image and the floating CT image; Input the fixed CT image, the floating CT image, and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; Adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, where the registered CT image and the fixed CT image are spatially aligned.

[0008] According to the fourth aspect of the embodiments of the present specification, a CT image registration method applied to a cloud-side device is provided, including: Obtain a fixed CT image and a floating CT image sent by an end-side device; Determine registration task information according to the fixed CT image and the floating CT image; Input the fixed CT image, the floating CT image, and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; Adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, where the registered CT image and the fixed CT image are spatially aligned; Send the registered CT image to the end-side device.

[0009] According to the fifth aspect of the embodiments of the present specification, a computer-aided diagnosis method is provided, including: Obtain a fixed CT image and a floating CT image for a target detection area, where the floating CT image includes initial marking information for the target detection area, and determine registration task information according to the fixed CT image and the floating CT image; Input the fixed CT image, the floating CT image, and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; Adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, where the registered CT image and the fixed CT image are spatially aligned; Mark the fixed CT image according to the initial marker information in the registered CT image to obtain the target marker information corresponding to the target detection area in the fixed CT image; Detect the target detection area according to the target marker information in the fixed CT image to obtain the detection result corresponding to the target detection area.

[0010] According to the sixth aspect of the embodiments of the present specification, a CT image registration system is provided, including a client and a server; The client is configured to send a fixed CT image and a floating CT image to the server; The server is configured to determine registration task information according to the fixed CT image and the floating CT image; input the fixed CT image, the floating CT image, and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, where the registered CT image and the fixed CT image are spatially aligned; and send the registered CT image to the client.

[0011] According to the seventh aspect of the embodiments of the present specification, an electronic device is provided, including: A memory and a processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of the above method are implemented.

[0012] According to the eighth aspect of the embodiments of the present specification, a computer-readable storage medium is provided, which stores computer programs / instructions. When the computer programs / instructions are executed by a processor, the steps of the above method are implemented.

[0013] According to the ninth aspect of the embodiments of the present specification, a computer program product is provided, including computer programs / instructions. When the computer programs / instructions are executed by a processor, the steps of the above method are implemented.

[0014] One embodiment of this specification provides a general image registration method. By inputting registration task information into the image registration model, the image registration model can perform various types of image registration tasks according to a single model framework, reducing the necessity of training corresponding registration models for each registration task. It significantly reduces the dependence on model development expertise and reduces the resource consumption caused by training models. It saves time and computational overhead. Description of the Drawings

[0015] Figure 1 is a flowchart of an image registration method provided by one embodiment of this specification; Figure 2 is a schematic diagram of data flow in an image registration model provided by one embodiment of this specification; Figure 3 is a flowchart of a training method for an image registration model provided by one embodiment of this specification; Figure 4 is a flowchart of a CT image registration method provided by one embodiment of this specification; Figure 5 is a flowchart of a CT image registration method applied to a cloud-side device provided by one embodiment of this specification; Figure 6 is a flowchart of a computer-aided diagnosis method provided by one embodiment of this specification; Figure 7 is an architecture diagram of a CT image registration system provided by one embodiment of this specification; Figure 8 is a schematic structural diagram of an image registration device provided by one embodiment of this specification; Figure 9 is a block diagram of the structure of an electronic device provided by one embodiment of this specification. Detailed Embodiments

[0016] In the following description, many specific details are set forth in order to provide a thorough understanding of this specification. However, this specification can be implemented in many other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the spirit of this specification. Therefore, this specification is not limited by the specific embodiments disclosed below.

[0017] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a", "the", and "said" used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used in one or more embodiments of this specification refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0018] It should be understood that although the terms first, second, etc. may be used in one or more embodiments of this specification to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0019] 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 for analysis, stored data, displayed data, etc.) involved in this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of the relevant data need to comply with the relevant laws, regulations, and standards of the relevant region, and corresponding operation entrances are provided for the user to select authorization or rejection.

[0020] First, the noun terms involved in one or more embodiments of this specification are explained.

[0021] Image registration: Image registration is the process of aligning two or more images spatially so that the corresponding points in these images coincide precisely in the same coordinate system. Its purpose is to eliminate the differences caused by factors such as shooting angle, scale, displacement, rotation, and deformation, so that multi-source images can be effectively compared, fused, and analyzed.

[0022] Significant progress has been made in learning-based image registration techniques, and their accuracy has been comparable to that of traditional registration techniques, while there has also been a significant improvement in computational efficiency. However, learning-based image registration methods still face challenges in terms of generalization ability. In different registration scenarios, it is inevitable to design and train multiple isolated registration networks for specific tasks. That is, corresponding registration networks need to be trained for different registration scenarios. And retraining a new registration network takes time and resources and is time-consuming and laborious.

[0023] Based on this, in this specification, an image registration method, a CT image registration method, a computer-aided diagnosis method, and a CT image registration system are provided. This specification also relates to an electronic device, a computer-readable storage medium, and a computer program product, which will be described in detail one by one in the following embodiments.

[0024] See Figure 1 , Figure 1 shows a flowchart of an image registration method provided according to an embodiment of this specification, which specifically includes the following steps.

[0025] Step 102: Obtain a fixed image and a floating image.

[0026] In a specific embodiment provided in this specification, an image registration method is provided. Image registration can be understood as seeking a spatial transformation for an image so that the corresponding points on it coincide spatially with those on another image after the transformation. For example, in a medical scenario, if the first medical image is spatially transformed so that its corresponding points coincide spatially with those on the second medical image, this behavior is called registering the first medical image with the second medical image.

[0027] In the method provided in the embodiments of this specification, the image to be spatially transformed is called a floating image (moving image), and the image used to assist the floating image in spatial transformation is called a fixed image (fixed image). For example, if it is necessary to perform a spatial transformation on Image 1 so that its corresponding points coincide spatially with those on Image 2. Then Image 1 is the floating image and Image 2 is the fixed image.

[0028] In addition, the method provided in the embodiments of this specification can be used on a terminal or in a cloud server. There are also many ways for a terminal or a cloud server to obtain a fixed image and a floating image. For example, when this method is applied to a terminal, the terminal can obtain the fixed image and the floating image from a specified storage location (such as a hard disk) according to an image acquisition instruction; it can also be that the terminal obtains the fixed image and the floating image through a specified access connection; it can also be that the user uploads the fixed image and the floating image to the terminal; or it can be that after an image is captured by an intelligent imaging device and sent to the terminal. In the method provided in the embodiments of this specification, the specific manner of obtaining the fixed image and the floating image is not limited. It depends on the actual application.

[0029] Step 104: Determine registration task information according to the fixed image and the floating image.

[0030] Among them, the registration task information can be understood as the information related to the registration task to be performed when registering the current fixed image and floating image. In this embodiment, the fixed image and the floating image are registered based on the registration task, and before registration, the registration task information corresponding to the registration task is used to assist in performing the registration task.

[0031] In a specific embodiment provided in this specification, determining the registration task information according to the fixed image and the floating image includes: Determining registration object information and / or image type information according to the fixed image and the floating image.

[0032] In the method provided in the embodiments of this specification, the registration task information includes at least one of registration object information and image type information. For better subsequent processing, in a better embodiment provided in the embodiments of this specification, the registration task information includes registration object information and image type information.

[0033] The registration object information can be understood as the object corresponding to each image in the registration task. For example, taking the medical registration scenario as an example, the registration object can be determined based on the object of the CT image. For example, if it is a head and neck CT, the registration object information is the head and neck. Another example is that if it is a chest CT, the registration object information is the chest.

[0034] In another specific embodiment provided in the embodiments of this specification, the registration object information can also be a specific organ. Still taking the medical registration scenario as an example, an image of a certain organ can be extracted from the CT image, such as the stomach, lungs, liver, etc. Then the registration object information can be the stomach, lungs, liver, etc. In the embodiments of this specification, the specific content of the registration object information is not limited and is subject to actual applications.

[0035] The image type information can be understood as the information determining the relationship between the fixed image and the floating image. In the actual application scenario of image registration, the fixed image and the floating image can be images of the same target at different time points, or images of multiple targets. Still taking the medical scenario as an example, the fixed image and the floating image can be CT images of user A at different time points, or CT image 1 of user A and CT image 2 of user B. Therefore, in the method provided in the embodiments of this specification, the image type information is used to determine the relationship information between the fixed image and the floating image, and the image type information is determined by two parameters, inter and intra. Among them, inter represents images between different patients, and intra represents images of the same patient. For example, when the image type information is 10, it means that the two images are between different patients; when the image type information is 01, it means that the two images are of the same patient.

[0036] In practical applications, the registration object information and / or the image type information can be carried from an image registration task, or can be determined by the fixed image and the floating image. Specifically, determining the registration object information and / or the image type information according to the fixed image and the floating image includes: Obtaining first image attribute information of the fixed image and obtaining second image attribute information of the floating image; Determining the registration object information and / or the image type information according to the first image attribute information and the second image attribute information.

[0037] Among them, the first image attribute information can be understood as the image attribute information of the fixed image, and the second image attribute information can be understood as the image attribute information of the floating image. The image attribute information refers to information related to the attributes of the object corresponding to the image. For example, the first image attribute information of the fixed image includes the detected object information of the image, the attribution role information of the detected object, and so on. The registration object information and / or the image type information can be further determined through the first image attribute information and the second image attribute information.

[0038] In a specific embodiment provided in this specification, determining the registration object information according to the first image attribute information and the second image attribute information includes: Obtaining a first detected object in the first image attribute information and obtaining a second detected object in the second image attribute information; When the first detected object is the same as the second detected object, determining the first detected object as the registration object information.

[0039] In this embodiment, taking the determination of the registration object information as an example for explanation, in the process of determining the registration object information, it is necessary to obtain the first detected object of the first image attribute information and obtain the second detected object in the second image attribute information. The detected object can be understood as the content included in the image. For example, if the image is a picture of the lungs, the detected object is the lungs. If the image is a picture of the chest and abdomen, the detected object is the chest and abdomen.

[0040] If the first detected object and the second detected object are the same, it can be determined that the first detected object is the registration object information. As shown in the above example, if the fixed image is a head and neck CT, the first detected object is the head and neck. The floating image is also a head and neck CT, then the second detected object is the head and neck. At this time, the first detected object and the second detected object are the same. Then the registration object information is the head and neck.

[0041] For another example, if the fixed image captures the lungs, the first detection object is the lungs. If the floating image captures the lungs, the second detection object is the lungs. At this time, the first detection object and the second detection object are the same. Then the registration object information is the lungs.

[0042] If the first detection object and the second detection object are not the same, it means that registration cannot be performed between these two images, and subsequent processing methods cannot be carried out. Therefore, in a specific embodiment provided in this specification, it is also necessary to further determine whether the first detection object and the second detection object are the same. If they are not the same, image registration cannot be performed. Only when the first detection object and the second detection object are the same can the subsequent processing process be carried out.

[0043] In another specific embodiment provided in this specification, determining the image type information according to the first image attribute information and the second image attribute information includes: Obtain the first role information in the first image attribute information, and obtain the second role information in the second image attribute information; When the first role information is the same as the second role information, determine that the image type information is intra-role registration; When the first role information is different from the second role information, determine that the image type information is inter-role registration.

[0044] In this embodiment, it further explains in detail how to determine the image type information. Obtain the first role information from the first image attribute information and the second role information from the second image attribute information. The role information can be understood as the belonging role of the registration object information. For example, if the fixed image is a CT image of the stomach of user A, the corresponding first role information is user A. For another example, if the fixed image is a CT image of the head and neck of user B, the corresponding first role information is user B.

[0045] If the first role information is the same as the second role information, it means that the fixed image and the floating image belong to the same role, that is, the image type information is intra-role registration; if the first role information is different from the second role information, it means that the fixed image and the floating image belong to different roles, that is, the image type information is inter-role registration.

[0046] Step 106: Input the fixed image, the floating image, and the registration task information into an image registration model to obtain the image adjustment information output by the image registration model.

[0047] After obtaining the fixed image, the floating image, and the registration task information, these three types of information can be input into the image registration model for processing to obtain the image adjustment information output by the image registration model.

[0048] In the method provided by the embodiments of this specification, an image registration model is pre-trained. The image registration model is trained to generate image adjustment information for guiding the spatial adjustment of the floating image according to the input fixed image, floating image, and registration task information.

[0049] The image adjustment information can be understood as the guiding information for the spatial adjustment of the floating image. Subsequently, the floating image can be rotationally adjusted according to this image adjustment information.

[0050] In the method provided by the embodiments of this specification, the pre-trained image registration model is a machine learning model. Specifically, the image registration model includes an embedding layer, an encoder, a decoder, and a convolutional controller; Inputting the fixed image, the floating image, and the registration task information into the image registration model to obtain the image adjustment information output by the image registration model includes: Inputting the fixed image, the floating image, and the registration task information into the embedding layer to obtain fixed image features, floating image features, and registration task information features; After splicing the fixed image features and floating image features, inputting them into the encoder to obtain encoded image features; Splicing the encoded image features and the registration task information features to obtain spliced feature information, and inputting the spliced feature information into the convolutional controller to obtain feature convolution information; Inputting the encoded image features into the decoder to obtain decoded image features; Performing convolution processing on the decoded image features according to the feature convolution information to generate image adjustment information.

[0051] Among them, the role of the embedding layer is to convert the image into feature information that the machine can recognize and process. In practical applications, when the image registration model processes an image, it actually processes the image feature information of the image and cannot process the image itself. Therefore, after inputting the fixed image and the floating image into the image registration model, the fixed image and the floating image need to be first converted into feature information that the image registration model can process. Specifically, the embedding layer in the image registration model performs embedding processing on each fixed image and floating image to obtain the fixed image features corresponding to each fixed image and the floating image features corresponding to the floating image.

[0052] In the embedding layer provided by the embodiments of this specification, text information can also be converted into corresponding feature information. Specifically, after inputting the registration task information into the embedding layer for processing, the registration task information can also be converted into corresponding registration task information features.

[0053] See Figure 2 , Figure 2The figure shows a schematic diagram of data flow in the image registration model provided by an embodiment of this specification. As Figure 2 shown, after the fixed image and the floating image are processed by the embedding layer, fixed image features and floating image features are obtained; the registration task information also undergoes embedding processing to obtain registration task information features. It should be noted that the registration object information in the registration task information itself is a parameter identifier of 0 or 1 and can be recognized by the image registration model. Therefore, the registration object information can be directly recognized in its original representation form.

[0054] After obtaining the fixed image features and the floating image features, the two can be concatenated and then input into the encoder for feature extraction. In the method provided by the embodiment of this specification, the backbone network of the image registration network can use a model with an encoder-decoder structure. In the image registration task, it is usually necessary to perform rotation adjustment on the image. Therefore, usually, the fixed image and the floating image can be 3D images. Therefore, in the method provided by the embodiment of this specification, a feature processing model based on the U-Net model can be used. In the encoder, a downsampling convolution method is adopted to perform convolution processing on the concatenated fixed image features and floating image features to obtain the encoded image features output by the encoder. Refer to Figure 2 for the process of concatenating the fixed image features and the floating image features and then inputting them into the encoder for processing to obtain the encoded image features.

[0055] After obtaining the encoded image features, they will be processed in two branches. The first branch will be concatenated with the registration task information features, so as to fuse the registration task information into the encoded image features to guide subsequent data processing; the second branch is decoding, that is, inputting the encoded image features into the decoder for decoding processing to obtain the corresponding decoded image features.

[0056] In the image registration model provided by the embodiment of this specification, a convolution controller is also provided. The purpose of the convolution controller is to convert the encoded image features fused with the registration task information in the first branch into feature convolution information for the registration task, so that in the subsequent processing process, the decoded image features are processed to generate the corresponding deformation field (i.e., image adjustment information). The convolution controller will dynamically generate different feature convolution information according to different input registration task information, so as to generate a suitable registration solution for each registration task, so that the image registration model can generate the corresponding convolution task according to the registration task information, and then generate the corresponding image adjustment information. Through the convolution controller, the image registration model can be adapted to different registration task information, and thus the image registration model has better generalization.

[0057] In a specific implementation manner provided by this specification, the convolution controller includes convolution kernel parameters; Input the spliced feature information into the convolutional controller to obtain feature convolution information, including: Input the spliced feature information into the convolutional controller, and generate at least one convolution kernel information according to the convolution kernel parameters in the convolutional controller; Generate feature convolution information according to each convolution kernel information.

[0058] In practical applications, the performance of traditional CNNs is related to their learnable kernel weights, and the learning kernels of convolutional layers are fixed after training. Networks optimized for specific tasks may perform poorly when applied to other tasks without retraining, making it impossible to use a single network for different tasks. The dynamic filter unit in the convolutional controller provided in the embodiments of this specification generates dynamic learning kernel parameters according to the input information. Generate corresponding learning kernel weights for different tasks through the convolutional controller, so that in subsequent processing, it can be more targeted.

[0059] The convolutional controller also includes a dynamic registration unit to assign customized kernels to each task associated with a specific registration task. The dynamic registration unit includes three stacked convolutional layers with 1*1*1 kernels. The convolution kernel parameters in the three layers are dynamically generated by the convolutional controller. The dynamic registration unit and the dynamic filter unit are the convolution kernel parameters in the convolutional controller.

[0060] See Figure 2 For example Figure 2 As shown, after the encoded image features output by the encoder are spliced with the feature information and image type information of the registration object information, they are input into the convolutional controller, and convolutional feature information is generated in the convolutional controller. In addition, the encoded image features are also input into the decoder for decoding to obtain decoded image features. Then, the decoded image features are convolved with the convolutional feature information to obtain the final image adjustment information.

[0061] To better splice the encoded image features with the features of the registration object information and the image type information, the convolution kernel in the encoder can be adjusted to output one-dimensional encoded image features, and then spliced with the above information. After being processed by the convolutional controller, the dynamic filter unit assigns the learning kernel weights for the registration task information, and the dynamic registration unit assigns the parameter configurations in the three convolutional layers, so that three-layer feature convolution information for the registration task information will be generated.

[0062] After obtaining the feature convolution information, the feature convolution information can be used to perform convolution processing on the decoded image features to generate image adjustment information.

[0063] Step 108: Adjust the floating image according to the image adjustment information to obtain a registered image, where the registered image and the fixed image are spatially aligned.

[0064] After obtaining the image adjustment information, the floating image can be adjusted according to the image adjustment information, so as to obtain a registered image corresponding to the floating image. The corresponding points in the registered image are spatially aligned with the corresponding points in the fixed image. Thus, the operation of registering the floating image with the fixed image is realized.

[0065] After the image registration is completed, some marker information on the fixed image can be transferred to the registered image, or some marker information in the registered image can be transferred to the fixed image.

[0066] In the method provided in the embodiments of this specification, a general image registration method is provided. By inputting registration task information into the image registration model, the image registration model can perform multiple types of image registration tasks according to a single model framework, reducing the necessity of training a corresponding registration model for each registration task. Significantly reducing the dependence on model development expertise and reducing the resource consumption caused by training the model. Saving time and computational overhead.

[0067] In addition, a convolutional controller is designed in the image registration model provided in the embodiments of this specification, which can encode different registration tasks (registration object information, image type information) and allocate different feature convolution information, so that the image registration model can adapt to multiple image registration tasks. Thus, the image registration model is more flexible.

[0068] See Figure 3 , Figure 3 shows a flowchart of a method for training an image registration model according to an embodiment of this specification, which specifically includes the following steps.

[0069] Step 302: Obtain a sample fixed image and a sample floating image, and determine sample registration task information according to the sample fixed image and the sample floating image.

[0070] In the method for training an image registration model provided in the embodiments of this specification, the image registration model is trained using pre-set training samples. The training samples include a sample fixed image and a sample floating image. It should be noted that marker information is set on both the sample fixed image and the sample floating image, that is, sample reference marker information is included in the sample fixed image, and sample registration marker information is included in the sample floating image.

[0071] The marked information refers to the information pre-marked by technicians on the fixed sample image and the floating sample image. For example, when the fixed sample image and the floating sample image are CT images, technicians can pre-mark the mask information of different organs in the fixed sample image and the floating sample image. The marked information is used to provide reference during the subsequent model training process.

[0072] In this embodiment, the method for obtaining the fixed sample image and the floating sample image is similar to the method for obtaining the fixed image and the floating image in the above embodiment. For the specific obtaining method, refer to the relevant description in the above embodiment, which will not be elaborated here.

[0073] After obtaining the fixed sample image and the floating sample image, the sample registration task information can be further determined based on the fixed sample image and the floating sample image. Similarly, the specific implementation method for determining the sample registration task information is similar to the implementation method for determining the registration task information based on the fixed image and the floating image in the above embodiment of the image registration method. For the specific determination method, refer to the relevant description in the above embodiment, which will not be elaborated here.

[0074] Step 304: Input the fixed sample image, the floating sample image, and the sample registration task information into the image registration model to obtain the predicted image adjustment information output by the image registration model.

[0075] After obtaining the fixed sample image, the floating sample image, and the sample registration task information, these information can be input into the image registration model for processing. At this time, the image registration model is still an untrained model, and the predicted image adjustment information output by the image registration model at this time is not the final predicted information, and further adjustment is required according to the subsequent model parameter adjustment method.

[0076] The model structure of the image registration model, as well as the data processing method of the fixed sample image, the floating sample image, and the sample registration task information in the image registration model, are the same as those in the above embodiment. For the detailed content of the model structure and the data processing method, refer to the relevant description in the above embodiment, which will not be elaborated here.

[0077] Step 306: Adjust the floating sample image according to the predicted image adjustment information to obtain the adjusted registration image.

[0078] After obtaining the predicted image adjustment information, the floating sample image can be adjusted according to the predicted image adjustment information, so as to obtain the adjusted registration image. The adjustment in this step is the same as the processing method of adjusting the floating image according to the image adjustment information to obtain the registered image in the above embodiment, which will not be elaborated here.

[0079] Step 308: Calculate the model loss value according to the sample fixed image and the adjusted registration image.

[0080] Since the image registration model at this time is not yet trained, the predicted image adjustment information generated by it may not be the final adjustment information. In order to further adjust the image registration model, it is also necessary to calculate the model loss value according to the sample fixed image and the adjusted registration image. In the method provided in the embodiments of this specification, there are many methods for calculating the model loss value. For the similarity loss of the images in the registration network, common ones include the correlation coefficient (CC), the normalized correlation coefficient (NCC), the mutual information (MI), the mean-square error (MSE), etc. When the images have similar gray value distributions, MSE is usually used to evaluate the similarity of gray values; while in multi-modal registration, the NCC and MI metrics are more appropriate. In this specification, the specific manner of the loss function is not limited and shall be subject to actual applications.

[0081] As mentioned in the above steps, the sample fixed image includes sample reference marker information, and the sample floating image includes sample registration marker information; correspondingly, in a specific implementation manner provided in this specification, calculating the model loss value according to the sample fixed image and the adjusted registration image includes: Mark the sample fixed image according to the sample registration marker information in the adjusted registration image to obtain a predicted registration image, where the predicted registration image includes predicted marker information; Calculate a first loss value according to the sample fixed image and the adjusted registration image; Calculate a second loss value according to the predicted marker information corresponding to the predicted registration image and the sample reference marker information in the sample fixed image; Determine the model loss value according to the first loss value and the second loss value.

[0082] In the method provided in the embodiments of this specification, calculating the model loss value according to the sample fixed image and the adjusted registration image can calculate the loss value from two dimensions. The first dimension is whether the adjusted registration image after adjustment can be spatially aligned with the sample fixed image, and the second dimension is to judge from the marker information on the image. Map the marker information in the adjusted registration image to the sample fixed image to generate predicted marker information, and then compare whether the predicted marker information in the sample fixed image is aligned with the sample reference marker information.

[0083] Based on this, the first loss value can be calculated according to the sample fixed image and the adjusted registration image. The first loss value is the loss value for verifying whether the adjusted registration image is spatially aligned with the sample fixed image.

[0084] In addition, since the adjusted registration image is obtained by spatial transformation of the sample floating image, the sample registration mark information in the sample floating image is also included in the adjusted registration image. By mapping the sample registration mark information to the sample fixed image, predicted mark information can be generated in the sample fixed image. Then, it is compared with the sample reference mark information in the sample fixed image to calculate the second loss value. The second loss value is the loss value for whether the mark information in the sample fixed image is aligned.

[0085] After obtaining the first loss value and the second loss value, the model loss value can be obtained by adding these two loss values.

[0086] In another specific implementation manner provided in this specification, the method further includes: Calculating a regularization loss value according to a preset regularization parameter; Correspondingly, determining the model loss value according to the first loss value and the second loss value includes: Determining the model loss value according to the first loss value, the second loss value, and the regularization loss value.

[0087] The loss of the registration network generally includes two parts. One is the similarity loss of the image, and the other is the smooth regularization term loss of the deformation field. The similarity loss of the image can be obtained through the first loss value and the second loss value. It is also possible to calculate the smooth regularization term loss by introducing regularization prior knowledge for calibration, so as to better perform model training.

[0088] Based on this, in another specific implementation manner provided in the embodiments of this specification, a preset regularization parameter related to the registration task can also be set in advance. The preset regularization parameter participates in the model training process as a hyperparameter and is used to generate a regularization loss value. The preset regularization parameter is used to match the registration task. During the model training process, the corresponding preset regularization parameter will be selected according to the specific content of the registration task to ensure that the image registration model knows the registration task it needs to perform, so that the model is more adaptable to different registration tasks, thereby improving the generalization of the image registration model.

[0089] In the method provided in the embodiments of this specification, the first loss value, the second loss value, and the regularization loss value can also be added to determine the final model loss value.

[0090] In the solution provided in the embodiments of this specification, the model loss value is generated from a first loss value, a second loss value, and a regularization loss value. The first loss value and the second loss value are determined from a sample fixed image and an adjusted registration image, and the regularization loss value is generated from a preset regularization parameter, which is determined according to the sample registration task information.

[0091] Step 310: Adjust the model parameters of the image registration model according to the model loss value, and continue to train the image registration model until the model training stop condition is reached.

[0092] After obtaining the model loss value, the model parameters of the image registration model can be adjusted according to the model loss value. Specifically, the model loss value is propagated backward in the image registration model to adjust the model parameters of the image registration model.

[0093] After adjusting the model parameters, continue to repeat the above model training operation until the model training stop condition is reached.

[0094] In the method provided in the embodiments of this specification, the model training stop condition can be that the model loss value is less than or equal to a preset loss value threshold and / or the model training reaches a preset number of training epochs. In the method provided in the embodiments of this specification, no specific limitation is imposed on the model training stop condition, and it can be set according to the actual situation.

[0095] In the training method of the image registration model provided in the embodiments of this specification, during the model processing, in addition to inputting two images to be registered, registration task information is introduced to inform the image registration model of the registration task to be processed, so that the image registration model can adapt to multiple tasks.

[0096] In addition, during the process of calculating the loss value in model training, loss values in two dimensions are calculated, so that the accuracy of model training is higher. In addition, a preset regularization parameter is introduced, and different preset regularization parameters are adjusted to match different registration tasks, so that the image registration model can better adapt to different registration tasks and further improve the generalization of the image registration model.

[0097] See Figure 4 , Figure 4 shows a flowchart of a CT image registration method provided in an embodiment of this specification, which specifically includes the following steps: Step 402: Obtain a fixed CT image and a floating CT image.

[0098] Step 404: Determine the registration task information according to the fixed CT image and the floating CT image.

[0099] Step 406: Input the fixed CT image, the floating CT image, and the registration task information into the CT image registration model to obtain the CT image adjustment information output by the CT image registration model.

[0100] Step 408: Adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, where the registered CT image and the fixed CT image are spatially aligned.

[0101] It should be noted that the implementation manners of steps 402 to 408 are the same as those of steps 102 to 108 above, and will not be elaborated in this embodiment of the present specification.

[0102] Applying the method of this embodiment of the present specification, the CT image registration model is the image registration model in the above embodiment. The model structure of the CT image registration model is the same as that of the image registration model in the above embodiment, which will not be elaborated here. Taking the fixed CT image and the floating CT image of the user's head and neck CT as an example for explanation, the registration task information determined according to the fixed CT image and the floating CT image is head and neck CT registration and inter-role registration. Input the fixed CT image, the floating CT image, head and neck CT registration, and inter-role registration into the CT image registration model, so as to obtain the CT image adjustment information output by the CT image registration model, and adjust the floating CT image based on the CT image adjustment information to obtain a registered CT image. Thus, the operation of registering the floating CT image with the fixed CT image is realized.

[0103] In the method provided in this embodiment of the present specification, a general CT image registration method is provided. By inputting the registration task information into the CT image registration model, the image registration model can execute various types of CT image registration tasks according to a single model framework, thereby reducing the necessity of registration models corresponding to multiple registration tasks. Significantly reducing the dependence on model development expertise and reducing the resource consumption caused by training models. Saving time and computational overhead.

[0104] In addition, a convolutional controller is designed in the CT image registration model provided in this embodiment of the present specification, which can encode different registration tasks (registration object information, image type information) and allocate different feature convolution information, so that the CT image registration model can adapt to various image registration tasks. Thus, the CT image registration model is more flexible.

[0105] See Figure 5 , Figure 5 shows a flowchart of a CT image registration method applied to a cloud-side device provided in an embodiment of the present specification, which specifically includes the following steps: Step 502: Obtain the fixed CT image and the floating CT image sent by the terminal device.

[0106] Step 504: Determine the registration task information according to the fixed CT image and the floating CT image.

[0107] Step 506: Input the fixed CT image, the floating CT image and the registration task information into the CT image registration model to obtain the CT image adjustment information output by the CT image registration model.

[0108] Step 508: Adjust the floating CT image according to the CT image adjustment information to obtain the registered CT image, where the registered CT image and the fixed CT image are spatially aligned.

[0109] Step 510: Send the registered CT image to the terminal device.

[0110] It should be noted that the implementation manners of steps 502 to 508 are the same as those of steps 102 to 108 above, and will not be elaborated in this embodiment of the specification.

[0111] In practical applications, the CT image registration method can also be deployed in the cloud device, and the processing speed of the CT image registration method is ensured by the rich computing resources in the cloud device. Thereby reducing the hardware configuration requirements of the terminal device and making the CT image registration method more universal.

[0112] See Figure 6 , Figure 6 shows a flowchart of a computer-aided diagnosis method provided by an embodiment of this specification, which specifically includes the following steps: Step 602: Obtain the fixed CT image and the floating CT image for the target detection area, where the floating CT image includes initial marking information for the target detection area, and determine the registration task information according to the fixed CT image and the floating CT image.

[0113] Step 604: Input the fixed CT image, the floating CT image and the registration task information into the CT image registration model to obtain the CT image adjustment information output by the CT image registration model.

[0114] Step 606: Adjust the floating CT image according to the CT image adjustment information to obtain the registered CT image, where the registered CT image and the fixed CT image are spatially aligned.

[0115] Step 608: Mark the fixed CT image according to the initial marking information in the registered CT image to obtain the target marking information corresponding to the target detection area in the fixed CT image.

[0116] Step 610: Detect the target detection area according to the target marker information in the fixed CT image to obtain the detection result corresponding to the target detection area.

[0117] In this embodiment, it involves an implementation method of mapping the marker information in the floating CT image to the fixed CT image after CT image registration to obtain the target marker information corresponding to the target detection area in the fixed CT image. Furthermore, computer-aided diagnosis is performed through the target marker information in the fixed CT image. Thereby, the role of computer technology in medical diagnosis is enhanced to assist doctors in diagnostic analysis.

[0118] For example, it is explained by taking the fixed CT image as the plain scan CT image of user A and the floating CT image as the enhanced CT image of user A. There is marker information of the doctor for the target detection area in the enhanced CT image. Through the method provided in the embodiments of this specification, the enhanced CT image and the plain scan CT image can be registered, so as to map the marker information marked in the enhanced CT image to the plain scan CT image. The doctor can then use the plain scan CT image with marker information for subsequent processing work.

[0119] See Figure 7 , Figure 7 which shows the architecture diagram of a CT image registration system provided by an embodiment of this specification. The CT image registration system may include a client 100 and a server 200; The client 100 is configured to send the fixed CT image and the floating CT image to the server 200; The server 200 is configured to determine registration task information according to the fixed CT image and the floating CT image; input the fixed CT image, the floating CT image, and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, wherein the registered CT image and the fixed CT image are spatially aligned; and send the registered CT image to the client 100. The client 100 is further configured to receive the registered CT image sent by the server 200.

[0120] The CT image registration system may include multiple clients 100 and a server 200. Among them, the clients 100 can be referred to as end-side devices, and the server 200 can be referred to as cloud-side devices. Communication connections can be established among the multiple clients 100 through the server 200. In the CT image registration scenario, the server 200 is used to provide CT image registration services among the multiple clients 100. The multiple clients 100 can be used as senders or receivers respectively to achieve communication through the server 200.

[0121] Users can interact with the server 200 through the client 100 to receive data sent by other clients 100, or send data to other clients 100, etc. In the CT image registration scenario, it can be that the user publishes a data stream to the server 200 through the client 100, and the server 200 generates a registered CT image according to the data stream and pushes the registered CT image to other communicating clients.

[0122] Among them, a connection is established between the client 100 and the server 200 through a network. The network provides a medium for the communication link between the client 100 and the server 200. The network can include various connection types, such as wired, wireless communication links, or fiber optic cables, etc. The data transmitted by the client 100 may need to be processed such as encoded, transcoded, compressed, etc. before being published to the server 200.

[0123] The client 100 can be a browser, an APP (Application), or a web application such as an H5 (HyperText Markup Language 5) application, or a light application (also known as a mini-program, a lightweight application program), or a cloud application, etc. The client 100 can be developed based on the software development kit (SDK, Software Development Kit) provided by the server 200 for the corresponding service, such as developed based on the real-time communication (RTC, Real Time Communication) SDK. The client 100 can be deployed in an electronic device and needs to rely on the device or certain APPs in the device to run, etc. The electronic device can, for example, have a display screen and support information browsing, etc., such as a personal mobile terminal such as a mobile phone, a tablet computer, a personal computer, etc. Various other types of applications can usually be configured in the electronic device, such as human-computer dialogue applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0124] The server 200 may include servers that provide various services, such as a server that provides communication services for multiple clients, or a server for background training that supports models used on the client, or a server that processes data sent by the client, etc. It should be noted that the server 200 can be implemented as a distributed server cluster composed of multiple servers, or as a single server. The server can also be a server of a distributed system, or a server combined with a blockchain. The server can also be a cloud server of basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs, Content Delivery Network), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0125] It is worth noting that the CT image registration method provided in the embodiments of this specification is generally executed by the server. However, in other embodiments of this specification, the client can also have a similar function as the server, so as to execute the CT image registration method provided in the embodiments of this specification. In other embodiments, the CT image registration method provided in the embodiments of this specification can also be jointly executed by the client and the server.

[0126] Corresponding to the above method embodiments, this specification also provides an embodiment of an image registration device. Figure 8 The structural schematic diagram of an image registration device provided by an embodiment of this specification is shown. As Figure 8 shown, the device includes: An acquisition module 802, configured to acquire a fixed image and a floating image; A determination module 804, configured to determine registration task information according to the fixed image and the floating image; A model processing module 806, configured to input the fixed image, the floating image, and the registration task information into an image registration model, and obtain image adjustment information output by the image registration model; An adjustment module 808, configured to adjust the floating image according to the image adjustment information to obtain a registered image, where the registered image and the fixed image are spatially aligned.

[0127] The above is a schematic solution of an image registration device in this embodiment. It should be noted that the technical solution of this image registration device and the technical solution of the above image registration method belong to the same concept. For the details not described in the technical solution of the image registration device, reference can be made to the description of the technical solution of the above image registration method.

[0128] Figure 9 FIG. Figure 9 shows a block diagram of an electronic device 900 according to an embodiment of the present application. The components of the electronic device 900 include, but are not limited to, a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0129] The electronic device 900 further includes an access device 940, which enables the electronic device 900 to communicate via one or more networks 960. Examples of these networks include the Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or a combination of communication networks such as the Internet. The access device 940 may include one or more of any type of wired or wireless network interfaces (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0130] In an embodiment of the present application, the above components of the electronic device 900 and Figure 9 other components not shown may also be connected to each other, for example, via a bus. It should be understood that Figure 9 the block diagram of the electronic device shown is for illustrative purposes only and is not a limitation on the scope of the present application. Those skilled in the art can add or replace other components as needed.

[0131] The electronic device 900 can be any type of stationary or mobile electronic device, including a mobile computer or mobile electronic device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable electronic device (e.g., a smart watch, smart glasses, etc.) or other types of mobile devices, or a stationary electronic device such as a desktop computer or a personal computer (PC). The electronic device 900 can also be a mobile or stationary server.

[0132] Among them, the processor 920 is used to execute the following computer program / instructions, and when the computer program / instructions are executed by the processor, the steps of the above-mentioned image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method are implemented.

[0133] The above is a schematic solution of an electronic device according to this embodiment. It should be noted that the technical solution of this electronic device and the technical solutions of the above-mentioned image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method belong to the same concept. For the details not described in detail in the technical solution of the electronic device, reference can be made to the descriptions of the technical solutions of the above-mentioned image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method.

[0134] An embodiment of this specification also provides a computer-readable storage medium, which stores computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method are implemented.

[0135] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiment of the computer-readable storage medium, since it is basically similar to the embodiments of the image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method, the description is relatively simple, and for the relevant parts, reference can be made to the partial descriptions of the embodiments of the image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method.

[0136] An embodiment of this specification also provides a computer program product, including computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned image registration method, the training method of the image registration model, CT image registration, and the computer-aided diagnosis method are implemented.

[0137] The above is a schematic solution of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solutions of the above image registration method, image registration model training method, CT image registration, and computer-aided diagnosis method belong to the same concept. For the details not described in detail in the technical solution of the computer program product, reference can be made to the descriptions of the technical solutions of the above image registration method, image registration model training method, CT image registration, and computer-aided diagnosis method.

[0138] The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0139] The computer instructions include computer program code, which may be in source code form, object code form, executable file, or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0140] It should be noted that the above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments of this specification.

[0141] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0142] The preferred embodiments of the present specification disclosed above are only used to help explain the present specification. The alternative embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, according to the content of the embodiments of the present specification, many modifications and variations can be made. These embodiments are selected and specifically described in the present specification to better explain the principles and practical applications of the embodiments of the present specification, so that those skilled in the art can well understand and utilize the present specification. The present specification is only limited by the claims and their full scope and equivalents.

Claims

1. An image registration method, comprising: Get fixed and floating images; determining registration task information according to the fixed image and the floating image; Inputting the fixed image, the floating image and the registration task information into an image registration model to obtain image adjustment information output by the image registration model; The floating image is adjusted according to the image adjustment information to obtain a registered image, wherein the registered image and the fixed image are spatially aligned.

2. The method of claim 1, determining the registration task information according to the fixed image and the floating image, comprising: Registration object information and / or image type information is determined based on the fixed image and the floating image.

3. The method of claim 2, determining registration object information and / or image type information according to the fixed image and the floating image, comprising: Acquire first image attribute information of the fixed image, and acquire second image attribute information of the floating image; Registration object information and / or image type information is determined according to the first image attribute information and the second image attribute information.

4. The method according to claim 3, determining the registration object information according to the first image attribute information and the second image attribute information, comprising: Acquire a first detection object in the first image attribute information, and acquire a second detection object in the second image attribute information; In the case that the first detection object is the same as the second detection object, the first detection object is determined as the registration object information.

5. The method according to claim 3, determining the image type information according to the first image attribute information and the second image attribute information, comprising: Acquire the first role information in the first image attribute information, and acquire the second role information in the second image attribute information; When the first role information and the second role information are the same, determining the image type information as intra-role registration; In a case where the first role information is different from the second role information, the image type information is determined to be inter-role registration.

6. The method of claim 1, wherein the image registration model comprises an embedding layer, an encoder, a decoder, and a convolution controller; Inputting the fixed image, the floating image and the registration task information into an image registration model to obtain image adjustment information output by the image registration model includes: Inputting the fixed image, the floating image and the registration task information into the embedding layer to obtain fixed image features, floating image features and registration task information features; splicing the fixed image features and the floating image features and inputting them into the encoder to obtain encoded image features; Splicing the encoded image features and the registration task information features to obtain splicing feature information, and inputting the splicing feature information into the convolution controller to obtain feature convolution information; Inputting the encoded image features into the decoder to obtain decoded image features; The decoded image features are convolved according to the feature convolution information to generate image adjustment information.

7. The method of claim 6, wherein the convolution controller includes convolution kernel parameters; Inputting the splicing feature information into the convolution controller to obtain feature convolution information includes: Inputting the splicing feature information into the convolution controller, and generating at least one convolution kernel information according to the convolution kernel parameters in the convolution controller; Generate feature convolution information based on each convolution kernel information.

8. A method for training an image registration model, comprising: Obtaining a sample fixed image and a sample floating image, and determining sample registration task information according to the sample fixed image and the sample floating image; Inputting the sample fixed image, the sample floating image and the sample registration task information into an image registration model to obtain predicted image adjustment information output by the image registration model; Adjust the sample floating image according to the predicted image adjustment information to obtain an adjusted registration image; Calculate a model loss value according to the sample fixed image and the adjusted registration image; The model parameters of the image registration model are adjusted according to the model loss value, and the image registration model is continuously trained until a model training stop condition is reached.

9. The method of claim 8, wherein the sample fixed image includes sample reference mark information, and the sample floating image includes sample registration mark information; Calculating a model loss value according to the sample fixed image and the adjusted registration image includes: Marking the sample fixed image according to the sample registration mark information in the adjusted registration image to obtain a predicted registration image, wherein the predicted registration image includes the predicted mark information; Calculate a first loss value according to the sample fixed image and the adjusted registration image; Calculate a second loss value according to the predicted mark information corresponding to the predicted registration image and the sample reference mark information in the sample fixed image; A model loss value is determined according to the first loss value and the second loss value.

10. The method of claim 9, further comprising: Calculate the regularization loss value according to the preset regularization parameter; Correspondingly, determining a model loss value according to the first loss value and the second loss value includes: A model loss value is determined according to the first loss value, the second loss value, and the regularization loss value.

11. A CT image registration method, comprising: Acquire fixed CT images and floating CT images; Determining registration task information according to the fixed CT image and the floating CT image; Inputting the fixed CT image, the floating CT image and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; The floating CT image is adjusted according to the CT image adjustment information to obtain a registered CT image, wherein the registered CT image and the fixed CT image are aligned in space.

12. A CT image registration method, applied to a cloud-side device, comprising: Obtain fixed CT images and floating CT images sent by the end-side device; Determining registration task information according to the fixed CT image and the floating CT image; Inputting the fixed CT image, the floating CT image and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; adjusting the floating CT image according to the CT image adjustment information to obtain a registered CT image, wherein the registered CT image and the fixed CT image are spatially aligned; The registered CT image is sent to the terminal device.

13. A computer-aided diagnosis method, comprising: Acquire a fixed CT image and a floating CT image for a target detection region, wherein the floating CT image includes initial marking information for the target detection region, and determine registration task information according to the fixed CT image and the floating CT image; Inputting the fixed CT image, the floating CT image and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; adjusting the floating CT image according to the CT image adjustment information to obtain a registered CT image, wherein the registered CT image and the fixed CT image are spatially aligned; Mark the fixed CT image according to the initial marking information in the registered CT image to obtain target marking information corresponding to the target detection area in the fixed CT image; The target detection area is detected according to the target mark information in the fixed CT image to obtain a detection result corresponding to the target detection area.

14. A CT image registration system, comprising a client and a server; The client is configured to send a fixed CT image and a floating CT image to the server; The server is configured to determine registration task information according to the fixed CT image and the floating CT image; input the fixed CT image, the floating CT image and the registration task information into a CT image registration model to obtain CT image adjustment information output by the CT image registration model; adjust the floating CT image according to the CT image adjustment information to obtain a registered CT image, wherein: The registered CT image and the fixed CT image are spatially aligned; and the registered CT image is sent to the client.

15. An electronic device, comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method described in any one of claims 1 to 13 are implemented.

16. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13.

17. A computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13.

Images