Image generation method and device and storage medium

Abstract

The invention provides an image generation method and device and a storage medium, and relates to the technical field of medical treatment. The specific implementation scheme is as follows: acquiring a bone image block and at least one type of soft tissue image block in an original image of a first modal; determining a dictionary matrix corresponding to at least one type of soft tissue image block from pre-stored dictionary matrixes corresponding to a plurality of soft tissue types; based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converting the at least one type of soft tissue image blocks of the first mode into at least one type of soft tissue image blocks of a second mode; and generating a target image based on the at least one type of soft tissue image blocks of the second modality and the bone image blocks in the original image. Therefore, the soft tissue image blocks in the original modal image are converted into the soft tissue image blocks in the new modal, and the target image is generated by combining the bone image blocks in the original modal image, so that the image quality can be effectively improved, and the image registration precision is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of medical technology, and in particular to an image generation method, device and storage medium. Background Art

[0002] In image guided radiation therapy (IGRT), accurate positioning of tumor target area depends on the image registration accuracy of different modal images such as computed tomography (CT) images and cone-beam computed tomography (CBCT) images.

[0003] However, due to the presence of artifacts in some modal images (such as CBCT images), the image quality is poor, resulting in low image registration accuracy. Summary of the invention

[0004] The present disclosure provides an image generation method, device and storage medium, which are used to convert soft tissue image blocks in the original modality image into soft tissue image blocks under a new modality, and then generate a target image in combination with the bone image blocks in the original modality image, which can effectively improve the image quality and thus improve the image registration accuracy.

[0005] According to one aspect of the present disclosure, there is provided an image generation method, the method comprising:

[0006] Acquire a bone image block and at least one type of soft tissue image block in an original image of the first modality; wherein one type of soft tissue image block corresponds to one soft tissue category;

[0007] Determining a dictionary matrix corresponding to the at least one type of soft tissue image block from dictionary matrices corresponding to a plurality of pre-stored soft tissue categories; the dictionary matrix is ​​used to characterize key features of the corresponding soft tissue image block;

[0008] Based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converting the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality;

[0009] A target image is generated based on at least one type of soft tissue image blocks of the second modality and the bone image blocks in the original image.

[0010] In some embodiments, the dictionary matrix includes a first dictionary matrix and a second dictionary matrix; the first dictionary matrix corresponds to the first mode, and the second dictionary matrix corresponds to the second mode; the first dictionary matrix and the second dictionary matrix correspond to the same weight coefficient matrix;

[0011] The method converts the at least one type of soft tissue image block of the first modality into at least one type of soft tissue image block of the second modality based on a dictionary matrix corresponding to the at least one type of soft tissue image block, including:

[0012] Acquire a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality;

[0013] Determine a weight coefficient matrix corresponding to at least one type of soft tissue image blocks of the first modality based on a sampling matrix corresponding to the at least one type of soft tissue image blocks and the first dictionary matrix;

[0014] Based on the weight coefficient matrix and the second dictionary matrix, a sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality is determined to obtain at least one type of soft tissue image blocks of the second modality.

[0015] In some embodiments, before determining the dictionary matrix corresponding to the at least one type of soft tissue image block from the pre-stored dictionary matrices corresponding to the plurality of soft tissue categories, the method further comprises:

[0016] Acquire a first sampling matrix corresponding to at least one type of soft tissue image blocks in a first sample image and a second sampling matrix corresponding to at least one type of soft tissue image blocks in a second sample image; the first sample image is a sample image of a first modality, and the second sample image is a sample image of a second modality;

[0017] Taking the first condition as a constraint condition, performing eigendecomposition on the first sampling matrix and the second sampling matrix respectively to obtain a first dictionary matrix and a second dictionary matrix corresponding to the at least one type of soft tissue image blocks;

[0018] The first dictionary matrix and the second dictionary matrix corresponding to the at least one type of soft tissue image blocks are stored correspondingly.

[0019] In some embodiments, the first condition includes: the first sampling matrix and the first dictionary matrix obtained by decomposing the first sampling matrix satisfy a first correlation condition, the second sampling matrix and the second dictionary matrix obtained by decomposing the second sampling matrix satisfy a second correlation condition, and the weight coefficient matrix obtained by eigendecomposition satisfies a first sparsity condition.

[0020] In some embodiments, the method further comprises:

[0021] Determine a mapping matrix corresponding to the at least one type of soft tissue image block from mapping matrices corresponding to a plurality of pre-stored soft tissue categories; the mapping matrix is ​​used to characterize a mapping relationship between a first weight coefficient matrix and a second weight coefficient matrix; the first weight coefficient matrix corresponds to the first modality, and the second weight coefficient matrix corresponds to the second modality; the first weight coefficient matrix and the second weight coefficient matrix correspond to the same dictionary matrix;

[0022] The method converts the at least one type of soft tissue image block of the first modality into at least one type of soft tissue image block of the second modality based on a dictionary matrix corresponding to the at least one type of soft tissue image block, including:

[0023] Acquire a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality;

[0024] Determining the first weight coefficient matrix based on the sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality and the dictionary matrix;

[0025] Determining the second weight coefficient matrix based on the first weight coefficient matrix and the mapping matrix;

[0026] Based on the second weight coefficient matrix and the dictionary matrix, a sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality is determined to obtain at least one type of soft tissue image blocks of the second modality.

[0027] In some embodiments, before determining the dictionary matrix corresponding to the at least one type of soft tissue image block from the pre-stored dictionary matrices corresponding to the plurality of soft tissue categories, the method further comprises:

[0028] Based on at least one type of soft tissue image block in a first sample image and at least one type of soft tissue image block in a second sample image, a comprehensive sampling matrix is ​​obtained; the first sample image is a sample image of a first modality, and the second sample image is a sample image of a second modality; the comprehensive sampling matrix is ​​used to characterize soft tissue features of the same soft tissue image block in the first sample image and the second sample image;

[0029] Taking the second condition as a constraint condition, performing feature decomposition on the comprehensive sampling matrix to obtain a dictionary matrix corresponding to the at least one type of soft tissue image block;

[0030] The dictionary matrix corresponding to the at least one type of soft tissue image block is stored accordingly.

[0031] In some embodiments, the second condition includes: the comprehensive sampling matrix and the dictionary matrix obtained by decomposing the comprehensive sampling matrix satisfy a third correlation condition, and the weight coefficient matrix obtained by decomposing the comprehensive sampling matrix satisfies a second sparsity condition.

[0032] In some embodiments, after performing eigendecomposition on the comprehensive sampling matrix with the second condition as a constraint condition to obtain a dictionary matrix corresponding to the at least one type of soft tissue image block, the method further includes:

[0033] Acquire a first sampling matrix corresponding to at least one type of soft tissue image blocks in the first sample image and a second sampling matrix corresponding to at least one type of soft tissue image blocks in the second sample image;

[0034] Based on the first sampling matrix and the dictionary matrix, determine a first weight coefficient matrix corresponding to the first sampling matrix; based on the second sampling matrix and the dictionary matrix, determine a second weight coefficient matrix corresponding to the second sampling matrix;

[0035] Determine a mapping matrix corresponding to the at least one type of soft tissue image blocks based on a first weight coefficient matrix corresponding to the first sampling matrix and a second weight coefficient matrix corresponding to the second sampling matrix;

[0036] The mapping matrices corresponding to the at least one type of soft tissue image blocks are stored accordingly.

[0037] In some embodiments, the method further comprises:

[0038] Deformable registration is performed on the first sample image and the second sample image.

[0039] In some embodiments, obtaining a bone image block and at least one type of soft tissue image block in an original image of a first modality includes:

[0040] Performing bone segmentation processing on the original image to obtain bone image blocks and soft tissue image blocks in the original image;

[0041] The soft tissue image blocks are classified to obtain the at least one type of soft tissue image blocks.

[0042] In some embodiments, the original image is a cone beam computed tomography (CBCT) image, and the target image is a pseudo computed tomography (CT) image.

[0043] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0044] at least one processor; and

[0045] a memory communicatively connected to the at least one processor; wherein,

[0046] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the image generating method provided by the present disclosure.

[0047] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the electronic device to execute the image generating method provided by the present disclosure.

[0048] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements the image generating method provided by the present disclosure when the computer program is executed by a processor.

[0049] The technical solution provided by the present disclosure obtains bone image blocks and at least one type of soft tissue image blocks in the original image of the first modality, and based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converts the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality. In this way, a dictionary learning-based image generation method is provided, which can convert the soft tissue image blocks in the original modality image into soft tissue image blocks under a new modality. Furthermore, a target image is generated based on the soft tissue image blocks under the new modality combined with the bone image blocks in the original modality image. In this way, not only the high-quality bone tissue structure information in the original modality image is retained, but also the high-quality soft tissue structure information under the new modality is converted, which can effectively improve the image quality, thereby improving the image registration accuracy.

[0050] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0052] Figure 1 is a schematic diagram of an implementation environment of an image generation method shown in an embodiment of the present disclosure;

[0053] Figure 2 is a flowchart of an image generation method shown in an embodiment of the present disclosure;

[0054] Figure 3 is a schematic diagram of the principle of an image generation method shown in an embodiment of the present disclosure;

[0055] Figure 4 is a schematic diagram of the principle of an image generation method shown in an embodiment of the present disclosure;

[0056] Figure 5 It is a block diagram of an electronic device used to implement the image generating method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0057] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0058] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0059] First, the application scenarios involved in the embodiments of the present disclosure are described. The image generation method provided by the embodiments of the present disclosure can be applied in the field of medical technology, and can be specifically applied in the scenario of radiotherapy. In some embodiments, the image generation method provided by the embodiments of the present disclosure can be applied in the treatment process of image-guided radiotherapy.

[0060] Generally, during the treatment of image-guided radiotherapy, accurate positioning of the tumor target area depends on the image registration accuracy of different modal images such as CT images and CBCT images. However, due to the presence of artifacts in some modal images (such as CBCT images), the image quality is poor, resulting in low image registration accuracy.

[0061] For example, image-guided radiotherapy usually requires scanning the patient's CT images for target area delineation, dose calculation, etc., and also requires scanning the patient's CBCT images for CBCT-CT image registration to achieve accurate positioning of the tumor target area. On the one hand, the presence of artifacts in the imaging area of ​​CBCT images will lead to a decrease in soft tissue resolution, so the quality of CBCT images is relatively poor compared to CT images, resulting in low image registration accuracy. On the other hand, the process is relatively cumbersome, requiring patients to repeatedly collect medical images, which increases the additional radiation dose received by patients.

[0062] In view of this, an embodiment of the present disclosure provides an image generation method, by acquiring bone image blocks and at least one type of soft tissue image blocks in an original image of a first modality, and based on a dictionary matrix corresponding to the at least one type of soft tissue image blocks, converting the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of a second modality. In this way, an image generation method based on dictionary learning is provided, which can convert soft tissue image blocks in the original modality image into soft tissue image blocks under a new modality. Furthermore, a target image is generated based on the soft tissue image blocks under the new modality in combination with the bone image blocks in the original modality image. In this way, not only the bone tissue structure information with higher quality (resolution) in the original modality image is retained, but also the soft tissue structure information with higher quality in the new modality is converted, which can effectively improve the image quality, thereby improving the image registration accuracy.

[0063] In some embodiments, taking the original image as a CBCT image and the target image as a pseudo CT image as an example, a method for generating a pseudo CT image from a CBCT image based on dictionary learning is provided. The method converts the soft tissue image blocks in the CBCT image into the soft tissue image blocks in the pseudo CT image, so that the soft tissue image blocks in the pseudo CT image can obtain a higher soft tissue resolution, and are superior to the CBCT image in terms of noise, artifacts and other indicators. Furthermore, the pseudo CT image is generated based on the soft tissue image blocks in the pseudo CT image combined with the bone image blocks in the CBCT image, so that the pseudo CT image retains the high-resolution bone fine structure unique to the CBCT image. In this way, the newly generated pseudo CT image not only has the high resolution advantage of the CBCT image, but also has the low noise advantage of the CT image, which can effectively improve the image quality, thereby improving the image registration accuracy. In addition, in image-guided radiotherapy, the pseudo CT image can be directly generated by scanning the patient's CBCT image, and the operations such as target area delineation, dose calculation, and patient positioning can be directly performed on the pseudo CT image. In this way, the patient does not need to repeatedly collect medical images, which not only reduces the additional radiation dose received by the patient, but also simplifies the process of image-guided radiotherapy.

[0064] Figure 1 Schematic diagram of an implementation environment of an image generation method shown in an embodiment of the present disclosure. Figure 1 The implementation environment includes an image acquisition device 101, an image server 102, and a storage server 103.

[0065] The image acquisition device 101 is a device for acquiring images of the patient's tumor site and surrounding normal tissues. In some embodiments, the image acquisition device 101 may be a cone beam computed tomography device (ie, a CBCT device).

[0066] See also Figure 1 The image acquisition device 101 may include a frame 1011, a tube 1012, a detector 1013, and a support device 1014. The frame 1011 may be a rotatable frame. The tube 1012 may emit imaging rays, such as KV-level X-rays. The detector 1013 may be a detection plate facing the tube 1012, such as an X-ray detection plate. The support device 1014 is used to support and move the patient, and may be a treatment bed.

[0067] In some embodiments, when the patient is on the support device 1014, the rotation of the rack 1011 can drive the tube 1012 to perform 360-degree projection around the patient. After the imaging ray passes through the patient, it can be projected onto the detector 1013. At this time, the detector 1013 can collect the projection data after the projection of the tube 1012. Then, after collecting the projection data of multiple projections, the medical image of the patient (such as a CBCT image) can be obtained through data reconstruction.

[0068] In the disclosed embodiment, the image acquisition device 101 is used to acquire the original image of the first modality of the patient. Then, the original image of the first modality can be uploaded to the image server 102, so that the image server 102 performs a subsequent image generation process based on the original image of the first modality to obtain a target image.

[0069] The image server 102 is used to provide background services for the above-mentioned image acquisition device 101. In some embodiments, the image server 102 can be an independent physical server, or a server cluster or distributed file system composed of multiple physical servers, or at least one of the cloud servers that provide 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 distribution networks, and big data or artificial intelligence platforms, etc., which is not limited in the embodiments of the present disclosure. In some embodiments, the number of the above-mentioned image servers 102 can be more or less, which is not limited in the embodiments of the present disclosure. Of course, the image server 102 can also include other functions to provide more comprehensive and diversified services.

[0070] In the disclosed embodiment, the image server 102 is used to obtain bone image blocks and at least one type of soft tissue image blocks in an original image of a first modality; wherein one type of soft tissue image block corresponds to one soft tissue category; determine a dictionary matrix corresponding to the at least one type of soft tissue image block from dictionary matrices corresponding to a plurality of pre-stored soft tissue categories; based on the dictionary matrix corresponding to the at least one type of soft tissue image block, convert the at least one type of soft tissue image block of the first modality into at least one type of soft tissue image block of a second modality; and generate a target image based on the at least one type of soft tissue image block of the second modality and the bone image blocks in the original image.

[0071] In some embodiments, the implementation environment may also include a gamma knife treatment head, an accelerator treatment head or other radiotherapy heads, which may be arranged on a rack to emit treatment rays, such as gamma rays and MV-level X-rays. Thus, the implementation environment constitutes a radiotherapy system, and accordingly, the image server 102 is also used to perform an image registration process based on the generated target image, and then perform pre-treatment positioning adjustment or real-time image guidance for the patient during treatment based on the image registration result.

[0072] The storage server 103 is used to store the patient's medical data, such as the patient's case type, radiotherapy plan, radiotherapy dose, etc. In the disclosed embodiment, the storage server 103 can also be used to store the medical images collected by the above-mentioned image acquisition device 101, the target image generated by the above-mentioned image server 102, etc.

[0073] The following is based on Figure 1 The implementation environment shown is used to introduce the method provided by the embodiment of the present disclosure.

[0074] Figure 2 1 is a flow chart of an image generation method shown in an embodiment of the present disclosure. In some embodiments, the image generation method is executed by an electronic device. For example, the electronic device may be the above-mentioned Figure 1 The image server shown. Figure 2 As shown, the method includes the following steps.

[0075] S201: Obtain bone image blocks and at least one type of soft tissue image blocks in an original image of a first modality.

[0076] Wherein, the first modality may be a CBCT modality, and accordingly, the original image of the first modality may be a CBCT image. A bone image block refers to an image block including a bone portion in an original image (such as a CBCT image). A soft tissue image block refers to an image block including a soft tissue portion in an original image (such as a CBCT image). It should be noted that a soft tissue image block is an image block of other tissues in the original image except for an image block of a bone portion.

[0077] One type of soft tissue image block corresponds to one soft tissue category. For example, soft tissue can be divided into tendon category, ligament category, fascia category, connective category, fat category, synovial category, muscle category, and so on. It should be noted that the classification granularity of soft tissue can be fine-grained, that is, fine classification of soft tissue, such as dividing soft tissue into the above seven categories or more. Alternatively, the classification granularity of soft tissue can also be coarse-grained, that is, rough classification of soft tissue, such as dividing soft tissue into three or fewer categories: nerve category, fat category, and muscle category. The disclosed embodiments do not limit the classification of soft tissue.

[0078] In the disclosed embodiment, the soft tissue is divided into different categories of soft tissue image blocks, and then the subsequent modality conversion process is performed for the soft tissue image blocks of different categories from the fine-grained dimension of soft tissue category, so as to convert the soft tissue image blocks in the original modality image into soft tissue image blocks under the new modality. In this way, a fine-grained image block modality conversion method is provided, which can improve the accuracy of image block modality conversion.

[0079] S202: Determine a dictionary matrix corresponding to the at least one type of soft tissue image block from among a plurality of pre-stored dictionary matrices corresponding to soft tissue categories.

[0080] The dictionary matrix is ​​used to characterize the key features of the corresponding soft tissue image block. In the disclosed embodiment, different soft tissue categories correspond to different dictionary matrices.

[0081] Thus, a method for generating images based on dictionary learning is provided, wherein dictionary learning is a method of extracting the most essential features of an object as a dictionary matrix and then using the dictionary matrix to represent the object.

[0082] In some embodiments, the goal of dictionary learning is to decompose the original sample matrix into a dictionary matrix and a weight coefficient matrix (or sparse code matrix) under the premise of satisfying constraints. It should be noted that the dictionary matrix contains the features in the original sample matrix (where each column of the dictionary matrix can be called an atom).

[0083] In this way, by extracting the most essential features of things through dictionary learning, we can use as few resources as possible to represent as much knowledge as possible, that is, use as few features as possible to represent as much image content as possible, which is conducive to improving the computational efficiency of subsequent image generation.

[0084] In the disclosed embodiment, dictionary matrices corresponding to soft tissue image blocks of different categories are pre-stored so that the dictionary matrices corresponding to soft tissue image blocks of the same category can be subsequently used to perform subsequent modality conversion processes based on the fine-grained dimension of soft tissue category to convert soft tissue image blocks in the original modality image into soft tissue image blocks under the new modality.

[0085] S203 . Convert the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality based on a dictionary matrix corresponding to the at least one type of soft tissue image blocks.

[0086] The second modality may be a CT modality. Thus, considering that CT images are superior to CBCT in terms of noise, artifacts and other indicators, a soft tissue image block with a higher soft tissue resolution can be obtained through the modality conversion of the soft tissue image block.

[0087] It should be noted that, in other embodiments, the first modality and the second modality may also be other types of modalities. For example, the first modality may also be a magnetic resonance imaging (MRI) modality or an ultrasound modality, etc. The second modality may also be an emission computed tomography (ECT) modality, etc. The embodiments of the present disclosure will be described exemplarily in the following example in which the first modality is a CBCT modality and the second modality is a CT modality.

[0088] S204: Generate a target image based on at least one type of soft tissue image blocks of the second modality and the bone image blocks in the original image.

[0089] The target image may be a pseudo CT image. In this way, the pseudo CT image not only retains the high-resolution bone fine structure unique to the CBCT image, but also contains soft tissue image blocks with high soft tissue resolution. In other words, the newly generated pseudo CT image not only has the high-resolution advantage of the CBCT image, but also has the low-noise advantage of the CT image, which can effectively improve the image quality and thus improve the image registration accuracy.

[0090] The technical solution provided by the embodiment of the present disclosure obtains the bone image blocks and at least one type of soft tissue image blocks in the original image of the first modality, and based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converts the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality. In this way, a dictionary learning-based image generation method is provided, which can convert the soft tissue image blocks in the original modality image into soft tissue image blocks under a new modality. Furthermore, a target image is generated based on the soft tissue image blocks under the new modality in combination with the bone image blocks in the original modality image. In this way, not only the high-quality bone tissue structure information in the original modality image is retained, but also the high-quality soft tissue structure information under the new modality is converted, which can effectively improve the image quality, thereby improving the image registration accuracy.

[0091] Above Figure 2 This is a simple embodiment of the present disclosure. The image generating method provided by the present disclosure is described below based on a specific embodiment. Figure 3 1 is a schematic diagram of a method for generating an image according to an embodiment of the present disclosure. In some embodiments, the method for generating an image is performed by an electronic device. For example, the electronic device may be the above-mentioned Figure 1 The image server shown. Figure 3 As shown, with the image server as the execution subject, the method includes the following steps.

[0092] S301: An image server obtains bone image blocks and at least one type of soft tissue image blocks in an original image of a first modality.

[0093] In the disclosed embodiment, the first modality may be a CBCT modality, and accordingly, the original image of the first modality may be a CBCT image. It should be understood that after the original image of the first modality is acquired by an image acquisition device, the original image of the first modality may be uploaded to an image server, and the image server may obtain the original image of the first modality. A bone image block refers to an image block including a bone portion in an original image (such as a CBCT image). A soft tissue image block refers to an image block including a soft tissue portion in an original image (such as a CBCT image). It should be noted that a soft tissue image block is an image block of other tissues in the original image except for an image block of a bone portion.

[0094] Among them, a type of soft tissue image block corresponds to a soft tissue category. For example, soft tissue can be divided into tendon category, ligament category, fascia category, connective category, fat category, synovial category, muscle category, etc. For example, see Figure 3 As shown in S301, the at least one type of soft tissue image block may be Figure 3 N types of soft tissue image blocks are shown. Wherein, N is a positive integer greater than or equal to 1. It should be understood that N is a preset number of categories, such as 5, 10 or other values. The embodiment of the present disclosure does not limit the number of categories of the at least one type of soft tissue image blocks.

[0095] In some embodiments, after acquiring the original image of the first modality, the image server performs bone segmentation processing on the original image to obtain bone image blocks and soft tissue image blocks in the original image, and then performs classification processing on the soft tissue image blocks to obtain the at least one type of soft tissue image blocks.

[0096] Regarding the above-mentioned bone segmentation process, for example, in some embodiments, the image server may utilize a bone segmentation algorithm to perform bone segmentation processing on the original image to obtain bone image blocks and soft tissue image blocks in the original image.

[0097] The original image is subjected to bone segmentation processing to obtain a binary image, which includes a sliding window for indicating bony tissue and a sliding window for indicating soft tissue. Furthermore, the original image can be divided into bone image blocks and soft tissue image blocks through sliding window patching.

[0098] Or, for example, in some other embodiments, the image server may use a bone segmentation model to perform bone segmentation processing on the original image to obtain bone image blocks and soft tissue image blocks in the original image.

[0099] The bone segmentation model is used to segment the bone tissue and soft tissue in the image to obtain bone image blocks and soft tissue image blocks. For example, the bone segmentation model can use sample images and binary images of the sample images as training data to iteratively train the initial neural network model.

[0100] It is worth noting that in other embodiments, the image server may also use other implementations to perform bone segmentation processing on the original image. The embodiments of the present disclosure do not limit the specific implementation of the bone segmentation processing.

[0101] In the above embodiment, the bone segmentation of the original image is performed to obtain the bone image blocks and soft tissue image blocks in the original image. Then, for the soft tissue image blocks in the original image, a subsequent modality conversion process is performed to convert the soft tissue image blocks in the original modality image into soft tissue image blocks under the new modality. On the one hand, the high-quality bone tissue structure information in the original modality image is retained. On the other hand, the soft tissue structure information under the new modality with high quality is converted. In this way, the image quality can be effectively improved, thereby improving the image registration accuracy.

[0102] Regarding the above classification process, for example, in some embodiments, the image server may utilize a clustering algorithm to perform classification processing on the soft tissue image blocks to obtain the at least one type of soft tissue image blocks.

[0103] The clustering algorithm is used to find image groups with similar categories to obtain image groups with different categories. For example, the clustering algorithm can be K-means clustering, hierarchical clustering, spectral clustering, density clustering, etc.

[0104] Or, for example, in some other embodiments, the image server may utilize a multi-classification model to classify the soft tissue image block to obtain the at least one category of soft tissue image blocks.

[0105] The multi-classification model is used to classify the soft tissue in the image to obtain soft tissue image blocks of different categories. For example, the multi-classification model can use sample images and category labels of soft tissue in the sample images as training data to iteratively train the initial neural network model.

[0106] It is worth noting that in other embodiments, the image server may also use other implementations to classify the soft tissue image block. The embodiments of the present disclosure do not limit the specific implementation of the classification process.

[0107] In the above embodiment, soft tissue image blocks are classified to obtain soft tissue image blocks of different categories. Then, for soft tissue image blocks of different categories, a subsequent modality conversion process is performed based on the fine-grained dimension of soft tissue category to convert the soft tissue image blocks in the original modality image into soft tissue image blocks under the new modality. In this way, a fine-grained image block modality conversion method is provided, which can improve the accuracy of image block modality conversion.

[0108] S302: The image server determines a dictionary matrix corresponding to the at least one type of soft tissue image block from pre-stored dictionary matrices corresponding to multiple soft tissue categories.

[0109] The dictionary matrix is ​​used to characterize the key features of the corresponding soft tissue image block. In the disclosed embodiment, different soft tissue categories correspond to different dictionary matrices.

[0110] In some embodiments, the image server may determine the dictionary matrix corresponding to the at least one type of soft tissue image block from dictionary matrices corresponding to multiple soft tissue categories pre-stored in a preset storage space.

[0111] The preset storage space is used to store dictionary matrices corresponding to multiple soft tissue categories. For example, the preset storage space may be a storage space built into the image server, or the preset storage space may also be a storage database associated with the image server.

[0112] In this way, by pre-storing the dictionary matrices corresponding to soft tissue image blocks of different categories, the dictionary matrices corresponding to soft tissue image blocks of the same category can be used later to perform the subsequent modality conversion process based on the fine-grained dimension of soft tissue category, so as to convert the soft tissue image blocks in the original modality image into soft tissue image blocks under the new modality.

[0113] In the disclosed embodiment, the dictionary matrix includes a first dictionary matrix and a second dictionary matrix. It should be noted that the first dictionary matrix corresponds to the first mode, and the second dictionary matrix corresponds to the second mode. The first dictionary matrix and the second dictionary matrix correspond to the same weight coefficient matrix.

[0114] For example, see Figure 3 In the content shown in S302, taking N types of soft tissue image blocks as an example, taking the first modality as the CBCT modality as an example, and taking the second modality as the CT modality as an example, the dictionary matrix corresponding to the at least one type of soft tissue image block may include D CB1 and D CT1 , D CB2 and D CT2 ,…,D CBN and D CTN Among them, D CB1 , D CB2…、D CBN is the first dictionary matrix corresponding to the at least one type of soft tissue image block, D CT1 , D CT2 ,…,D CTN is a second dictionary matrix corresponding to the at least one type of soft tissue image blocks.

[0115] Based on the contents shown in S303 to S305, the following describes the process of converting the at least one type of soft tissue image block of the first modality into at least one type of soft tissue image block of the second modality based on the dictionary matrix corresponding to the at least one type of soft tissue image block. The second modality may be a CT modality. In this way, considering that CT images are superior to CBCT in terms of noise, artifacts and other indicators, soft tissue image blocks with higher soft tissue resolution can be obtained through the modality conversion of the soft tissue image blocks.

[0116] S303: The image server obtains a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality.

[0117] The sampling matrix is ​​used to characterize the soft tissue features of the corresponding image block.

[0118] In some embodiments, the image server may perform vectorization processing on the at least one type of soft tissue image blocks to obtain a sampling matrix corresponding to the at least one type of soft tissue image blocks.

[0119] For example, see Figure 3 In the content shown in S303, taking N types of soft tissue image blocks as an example, taking the first modality as the CBCT modality as an example, the sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality (that is, the sampling matrix of the first modality) may include A CB1 , A CB2 , …, A CBN .

[0120] S304: The image server determines a weight coefficient matrix corresponding to at least one type of soft tissue image blocks of the first modality based on the sampling matrix corresponding to the at least one type of soft tissue image blocks and the first dictionary matrix.

[0121] In some embodiments, for each type of soft tissue image blocks in the at least one type of soft tissue image blocks, the image server determines the weight coefficient matrix corresponding to the type of soft tissue image blocks based on the sampling matrix corresponding to the type of soft tissue image blocks, the first dictionary matrix corresponding to the type of soft tissue image blocks, and the following formula (1), thereby obtaining the weight coefficient matrix corresponding to the at least one type of soft tissue image blocks.

[0122] A CB =D CB X (1)

[0123] In the formula, A CB represents the sampling matrix corresponding to the soft tissue image block of the first modality; D CB represents the first dictionary matrix (corresponding to the first mode) corresponding to the soft tissue image block; X represents the weight coefficient matrix corresponding to the soft tissue image block.

[0124] For example, see Figure 3 In the content shown in S304, taking N types of soft tissue image blocks as an example, the weight coefficient matrix corresponding to at least one type of soft tissue image block may include X1, X2, ..., X N .

[0125] In this way, through the above formula (1), by utilizing the mapping relationship between the weight coefficient matrix corresponding to the soft tissue image block and the sampling matrix corresponding to the soft tissue image block and the first dictionary matrix corresponding to the soft tissue image block, the weight coefficient matrix corresponding to the at least one type of soft tissue image block can be quickly and accurately determined, so that the determined weight coefficient matrix can be subsequently used to determine at least one type of soft tissue image block of the second modality.

[0126] S305. The image server determines a sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality based on the weight coefficient matrix and the second dictionary matrix, and obtains at least one type of soft tissue image blocks of the second modality.

[0127] In some embodiments, for each type of soft tissue image blocks in the at least one type of soft tissue image blocks, the image server determines the sampling matrix corresponding to the type of soft tissue image blocks of the second modality based on the weight coefficient matrix corresponding to the type of soft tissue image blocks, the second dictionary matrix corresponding to the type of soft tissue image blocks, and the following formula (2), and then obtains the sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality, thereby obtaining at least one type of soft tissue image blocks of the second modality.

[0128] A CT =D CT X (2)

[0129] In the formula, A CT represents the sampling matrix corresponding to the soft tissue image block of the second modality; D CT represents the second dictionary matrix (corresponding to the second mode) corresponding to the soft tissue image block; X represents the weight coefficient matrix corresponding to the soft tissue image block.

[0130] For example, see Figure 3 In the content shown in S305, taking N types of soft tissue image blocks as an example, taking the second modality as the CT modality as an example, the sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality (that is, the sampling matrix of the second modality) may include A CT1 , A CT2 , …, A CTN.

[0131] In this way, through the above formula (2), by utilizing the mapping relationship between the sampling matrix corresponding to the soft tissue image block and the weight coefficient matrix corresponding to the soft tissue image block and the second dictionary matrix corresponding to the soft tissue image block, the sampling matrix corresponding to at least one type of soft tissue image block of the second modality can be quickly and accurately determined, and at least one type of soft tissue image block of the second modality can be obtained.

[0132] In the disclosed embodiment, since the first dictionary matrix and the second dictionary matrix correspond to the same weight coefficient matrix, the weight coefficient matrix can be determined using the sampling matrix of the soft tissue image block of the first modality and the first dictionary matrix, and then, the sampling matrix of the soft tissue image block of the second modality can be determined using the determined weight coefficient matrix and the second dictionary matrix, that is, at least one type of soft tissue image block of the second modality is obtained.

[0133] It should be noted that before implementing this solution, the image server also needs to use the first sample image and the second sample image to perform dictionary training to obtain the dictionary matrices corresponding to the above-mentioned multiple soft tissue categories, that is, the first dictionary matrix and the second dictionary matrix corresponding to the multiple soft tissue categories. Among them, the first sample image is a sample image of the first modality. The second sample image is a sample image of the second modality.

[0134] Accordingly, the dictionary training process may include: obtaining a first sampling matrix corresponding to at least one type of soft tissue image block in the first sample image and a second sampling matrix corresponding to at least one type of soft tissue image block in the second sample image. Taking the first condition as a constraint condition, the first sampling matrix and the second sampling matrix are subjected to feature decomposition respectively to obtain a first dictionary matrix and a second dictionary matrix corresponding to the at least one type of soft tissue image block. Furthermore, the first dictionary matrix and the second dictionary matrix corresponding to the at least one type of soft tissue image block are stored correspondingly.

[0135] Exemplarily, the first sample image may be a CBCT image. The second sample image may be a CT image. The first sampling matrix is ​​used to refer to the sampling matrix of the sample image of the first modality. The second sampling matrix is ​​used to refer to the sampling matrix of the sample image of the second modality.

[0136] In some embodiments, after acquiring the first sample image and the second sample image, deformation registration is performed on the first sample image and the second sample image. In this way, through deformation registration of the first sample image and the second sample image, it is ensured that the feature structure and spatial position of the first sample image and the second sample image are matched, thereby effectively enhancing the accuracy of the dictionary learning algorithm.

[0137] After performing the above-mentioned deformable registration, the first sample image and the second sample image can be subjected to bone segmentation processing respectively using a bone segmentation algorithm or a bone segmentation model, and then the first sample image and the second sample image can be divided into bone image blocks and soft tissue image blocks by sliding window segmentation. Furthermore, the soft tissue image blocks in the first sample image and the second sample image can be classified using a clustering algorithm or a multi-classification model to obtain at least one type of soft tissue image blocks in the first sample image and the second sample image.

[0138] Among them, the first condition includes: the first sampling matrix and the first dictionary matrix obtained by decomposing the first sampling matrix meet the first correlation condition, the second sampling matrix and the second dictionary matrix obtained by decomposing the second sampling matrix meet the second correlation condition, and the weight coefficient matrix obtained by feature decomposition meets the first sparsity condition.

[0139] In some embodiments, for each type of soft tissue image block in the at least one type of soft tissue image blocks, the first sampling matrix and the second sampling matrix corresponding to the type of soft tissue image blocks are eigen-decomposed by solving the following optimization equation (3) to obtain the first dictionary matrix and the second dictionary matrix corresponding to the type of soft tissue image blocks, thereby obtaining the first dictionary matrix and the second dictionary matrix corresponding to the at least one type of soft tissue image blocks.

[0140]

[0141] In the formula, A CB The sampling matrix corresponding to the soft tissue image block of the first modality is the first sampling matrix; D CB represents the first dictionary matrix (corresponding to the first mode) corresponding to the soft tissue image block; X represents the weight coefficient matrix corresponding to the soft tissue image block; A CT The sampling matrix corresponding to the soft tissue image block of the second modality is the second sampling matrix; D CT represents the second dictionary matrix (corresponding to the second modality) corresponding to the soft tissue image block; α and β represent weight coefficients.

[0142] It should be understood that by performing feature decomposition on the first sampling matrix corresponding to this type of soft tissue image block, the first dictionary matrix and weight coefficient matrix corresponding to this type of soft tissue image block can be obtained, and by performing feature decomposition on the second sampling matrix corresponding to this type of soft tissue image block, the second dictionary matrix and weight coefficient matrix corresponding to this type of soft tissue image block can be obtained.

[0143] It should be noted that in the above optimization equation (3), Indicates that the first sampling matrix and the first dictionary matrix decomposed by the first sampling matrix meet the first correlation condition. The first correlation condition can be used to indicate D CBThe value of X and A CB The value of is equal to or can also be used to indicate D CB The value of X and A CB It should be understood that D CB The value of X and A CB The closer the value is, the better the result of eigendecomposition is.

[0144] Indicates that the second sampling matrix and the second dictionary matrix obtained by decomposing the second sampling matrix meet the second correlation condition. The second correlation condition can be used to indicate D CT The value of X and A CT The value of is equal to or can also be used to indicate D CT The value of X and A CT It should be understood that D CT The value of X and A CT The closer the value is, the better the result of eigendecomposition is.

[0145] ||X||1 indicates that the weight coefficient matrix obtained by eigendecomposition satisfies the first sparsity condition. The first sparsity condition can be used to indicate that the weight coefficient matrix is ​​as sparse as possible. It should be understood that the sparser the weight coefficient matrix is, the better the result of eigendecomposition is.

[0146] For example, in some embodiments, the alternating direction method of multipliers (ADMM) can be used to solve the above optimization equation (3) to obtain the first dictionary matrix, the second dictionary matrix and the weight coefficient matrix after matrix decomposition. It should be noted that in the embodiment of the present disclosure, the first dictionary matrix and the second dictionary matrix after matrix decomposition are obtained, and the obtained first dictionary matrix and the second dictionary matrix are stored correspondingly, such as storing the obtained first dictionary matrix and the second dictionary matrix correspondingly in a preset storage space. In other words, there is no need to store the weight coefficient matrix after matrix decomposition.

[0147] In addition, it should be noted that dictionary training is an iterative training process, and the corresponding process can be: in any iterative training process, first solve the dictionary matrix, and then use the solved dictionary matrix to solve the weight coefficient matrix. Then, based on the solved weight coefficient matrix, the next iterative training process is performed to optimize the dictionary matrix until the iterative training meets the above first condition, and the characteristic decomposition result at the end of the iteration is obtained.

[0148] In the above embodiment, the first dictionary matrix and the second dictionary matrix correspond to the same weight coefficient matrix. In this way, the weight coefficient matrix is ​​kept unchanged, and the modality conversion of the soft tissue image block is realized by updating the dictionary, that is, the soft tissue image block in the original modality image is converted into the soft tissue image block under the new modality through the conversion between the first dictionary matrix and the second dictionary matrix.

[0149] S306: The image server generates a target image based on the at least one type of soft tissue image blocks of the second modality and the bone image blocks in the original image.

[0150] In the disclosed embodiment, the target image may be a pseudo CT image.

[0151] In some embodiments, after determining the sampling matrix corresponding to at least one type of soft tissue image block of the second modality, the sampling matrix corresponding to the at least one type of soft tissue image block is converted into the at least one type of soft tissue image block. Further, the at least one type of soft tissue image block is combined with the bone image block in the original image to obtain a target image.

[0152] The technical solution provided by the embodiment of the present disclosure obtains the bone image blocks and at least one type of soft tissue image blocks in the original image of the first modality, and based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converts the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality. In this way, a dictionary learning-based image generation method is provided, which can convert the soft tissue image blocks in the original modality image into soft tissue image blocks under a new modality. Furthermore, a target image is generated based on the soft tissue image blocks under the new modality in combination with the bone image blocks in the original modality image. In this way, not only the high-quality bone tissue structure information in the original modality image is retained, but also the high-quality soft tissue structure information under the new modality is converted, which can effectively improve the image quality, thereby improving the image registration accuracy.

[0153] Exemplarily, taking the original image of the first modality as a CBCT image and the target image as a pseudo CT image as an example, a method for generating a pseudo CT image from a CBCT image based on dictionary learning is provided. The corresponding process can be replaced by: the image server obtains the bone image block and at least one type of soft tissue image block in the CBCT image. From the dictionary matrices corresponding to the pre-stored multiple soft tissue categories, the dictionary matrix corresponding to the at least one type of soft tissue image block is determined. The sampling matrix corresponding to the at least one type of soft tissue image block of the CBCT modality is obtained. Based on the sampling matrix corresponding to the at least one type of soft tissue image block of the CBCT modality and the first dictionary matrix, the weight coefficient matrix corresponding to the at least one type of soft tissue image block is determined. Based on the weight coefficient matrix and the second dictionary matrix, the sampling matrix corresponding to the at least one type of soft tissue image block of the CT modality is determined to obtain at least one type of soft tissue image block of the CT modality. Based on the at least one type of soft tissue image block of the CT modality and the bone image block in the CBCT image, a pseudo CT image is generated.

[0154] In this way, by converting the soft tissue image blocks in the CBCT image into the soft tissue image blocks in the pseudo CT image, the soft tissue image blocks in the pseudo CT image can obtain a higher soft tissue resolution, which is superior to the CBCT image in terms of noise, artifacts and other indicators. Furthermore, based on the soft tissue image blocks in the pseudo CT image combined with the bone image blocks in the CBCT image, the pseudo CT image retains the high-resolution bone fine structure unique to the CBCT image. In this way, the newly generated pseudo CT image not only has the high resolution advantage of the CBCT image, but also has the low noise advantage of the CT image, which can effectively improve the image quality and thus improve the image registration accuracy. In addition, in image-guided radiotherapy, the pseudo CT image can be directly generated by scanning the patient's CBCT image, and the operations such as target area delineation, dose calculation, and patient positioning can be directly performed on the pseudo CT image. In this way, the patient does not need to repeatedly collect medical images many times, which not only reduces the additional radiation dose received by the patient, but also simplifies the process of image-guided radiotherapy.

[0155] Figure 4 1 is a schematic diagram of a method for generating an image according to an embodiment of the present disclosure. In some embodiments, the method for generating an image is performed by an electronic device. For example, the electronic device may be the above-mentioned Figure 1 The image server shown. Figure 4 As shown, with the image server as the execution subject, the method includes the following steps.

[0156] S401: An image server obtains bone image blocks and at least one type of soft tissue image blocks in an original image of a first modality.

[0157] For example, see Figure 4As shown in S401, the at least one type of soft tissue image block may be Figure 4 N types of soft tissue image blocks are shown.

[0158] It should be noted that the relevant contents of S401 refer to S301 and will not be repeated here.

[0159] S402: The image server determines a dictionary matrix corresponding to the at least one type of soft tissue image block from pre-stored dictionary matrices corresponding to multiple soft tissue categories.

[0160] In the disclosed embodiment, the dictionary matrix is ​​a unified dictionary matrix corresponding to the first modality and the second modality, that is, the dictionary matrix includes both the key features of the soft tissue image blocks in the original image of the first modality and the key features of the soft tissue image blocks in the image of the second modality.

[0161] In some embodiments, the image server may determine the dictionary matrix corresponding to the at least one type of soft tissue image block from dictionary matrices corresponding to multiple soft tissue categories pre-stored in a preset storage space, wherein the preset storage space is used to store a unified dictionary matrix corresponding to multiple soft tissue categories.

[0162] For example, see Figure 4 In the content shown in S402, taking N types of soft tissue image blocks as an example, the dictionary matrix corresponding to at least one type of soft tissue image block may include D1, D2, ..., D N .

[0163] S403: The image server determines a mapping matrix corresponding to the at least one type of soft tissue image block from pre-stored mapping matrices corresponding to a plurality of soft tissue types.

[0164] Wherein, the mapping matrix is ​​used to characterize the mapping relationship between the first weight coefficient matrix and the second weight coefficient matrix. It should be noted that the first weight coefficient matrix corresponds to the first mode. The second weight coefficient matrix corresponds to the second mode. The first weight coefficient matrix and the second weight coefficient matrix correspond to the same dictionary matrix.

[0165] In some embodiments, the image server may determine the mapping matrix corresponding to the at least one type of soft tissue image block from mapping matrices corresponding to multiple soft tissue categories pre-stored in a preset storage space, wherein the preset storage space is also used to store mapping matrices corresponding to multiple soft tissue categories.

[0166] For example, see Figure 4 In the content shown in S403, taking N types of soft tissue image blocks as an example, the mapping matrix corresponding to at least one type of soft tissue image block may include M1, M2, ..., M N .

[0167] Based on the contents shown in S404 to S408, the following describes a process in which the image server converts the at least one type of soft tissue image block of the first modality into at least one type of soft tissue image block of the second modality based on the dictionary matrix corresponding to the at least one type of soft tissue image block.

[0168] S404: The image server obtains a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality.

[0169] For example, see Figure 4 In the content shown in S404, taking N types of soft tissue image blocks as an example, taking the first modality as the CBCT modality as an example, the sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality (that is, the sampling matrix of the first modality) may include A CB1 , A CB2 , …, A CBN .

[0170] It should be noted that the relevant contents of S404 refer to S303 and will not be described in detail.

[0171] S405. The image server determines a first weight coefficient matrix based on a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality and the dictionary matrix.

[0172] In some embodiments, for each type of soft tissue image blocks in the at least one type of soft tissue image blocks, the image server determines the first weight coefficient matrix corresponding to the type of soft tissue image blocks based on the sampling matrix corresponding to the type of soft tissue image blocks, the dictionary matrix corresponding to the type of soft tissue image blocks, and the following formula (4), thereby obtaining the first weight coefficient matrix corresponding to the at least one type of soft tissue image blocks.

[0173] A CB =D u X CB (4)

[0174] In the formula, A CB represents the sampling matrix corresponding to the soft tissue image block of the first modality; D u Represents the dictionary matrix corresponding to the soft tissue image block; X CB Represents the first weight coefficient matrix (corresponding to the first modality) corresponding to the soft tissue image block.

[0175] For example, see Figure 4 In the content shown in S405, taking N types of soft tissue image blocks as an example, the first weight coefficient matrix corresponding to at least one type of soft tissue image blocks may include X CB1 , X CB2 , …, X CBN .

[0176] In this way, through the above formula (4), by utilizing the mapping relationship between the first weight coefficient matrix and the sampling matrix corresponding to the soft tissue image block and the dictionary matrix corresponding to the soft tissue image block, the first weight coefficient matrix can be determined quickly and accurately, so that the determined first weight coefficient matrix can be used to perform the subsequent process of determining the second weight coefficient matrix.

[0177] S406: The image server determines the second weight coefficient matrix based on the first weight coefficient matrix and the mapping matrix.

[0178] In some embodiments, for each type of soft tissue image blocks in the at least one type of soft tissue image blocks, the image server determines the second weight coefficient matrix corresponding to the type of soft tissue image blocks based on the first weight coefficient matrix corresponding to the type of soft tissue image blocks, the mapping matrix corresponding to the type of soft tissue image blocks, and the following formula (5), thereby obtaining the second weight coefficient matrix corresponding to the at least one type of soft tissue image blocks.

[0179] X CT =NX CB (5)

[0180] In the formula, X CT represents the second weight coefficient matrix (corresponding to the second mode) corresponding to the soft tissue image block; M represents the mapping matrix corresponding to the soft tissue image block; X CB Represents the first weight coefficient matrix (corresponding to the first modality) corresponding to the soft tissue image block.

[0181] For example, see Figure 4 In the content shown in S406, taking N types of soft tissue image blocks as an example, the second weight coefficient matrix corresponding to at least one type of soft tissue image blocks may include X CT1 , X CT2 , …, X CTN .

[0182] In this way, through the above formula (5), by utilizing the mapping relationship between the second weight coefficient matrix and the first weight coefficient matrix and the mapping matrix, the second weight coefficient matrix can be determined quickly and accurately, so that the determined second weight coefficient matrix can be subsequently used to determine at least one type of soft tissue image block of the second modality.

[0183] S407: The image server determines a sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality based on the second weight coefficient matrix and the dictionary matrix, and obtains at least one type of soft tissue image blocks of the second modality.

[0184] In some embodiments, for each type of soft tissue image blocks in the at least one type of soft tissue image blocks, the image server determines the sampling matrix corresponding to the type of soft tissue image blocks of the second modality based on the second weight coefficient matrix corresponding to the type of soft tissue image blocks, the dictionary matrix corresponding to the type of soft tissue image blocks, and the following formula (6), and then obtains the sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality, thereby obtaining at least one type of soft tissue image blocks of the second modality.

[0185] A CT =D u X CT (6)

[0186] In the formula, A CT represents the sampling matrix corresponding to the soft tissue image block of the second modality; D u Represents the dictionary matrix corresponding to the soft tissue image block; X CT The second weight coefficient matrix (corresponding to the second modality) corresponding to the soft tissue image block is represented.

[0187] For example, see Figure 4 In the content shown in S407, taking N types of soft tissue image blocks as an example, taking the second modality as the CT modality as an example, the sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality (that is, the sampling matrix of the second modality) may include A CT1 , A CT2 , …, A CTN .

[0188] In this way, through the above formula (6), by utilizing the mapping relationship between the sampling matrix corresponding to the soft tissue image block and the second weight coefficient matrix corresponding to the soft tissue image block and the dictionary matrix corresponding to the soft tissue image block, the sampling matrix corresponding to at least one type of soft tissue image block of the second modality can be quickly and accurately determined, and at least one type of soft tissue image block of the second modality can be obtained.

[0189] In the disclosed embodiment, since the first weight coefficient matrix and the second weight coefficient matrix correspond to the same dictionary matrix, the dictionary matrix can be determined using the sampling matrix of the soft tissue image block of the first modality and the first weight coefficient matrix, and then the sampling matrix of the soft tissue image block of the second modality can be determined using the determined dictionary matrix and the second weight coefficient matrix, that is, at least one type of soft tissue image block of the second modality is obtained.

[0190] It should be noted that, before implementing this solution, the image server also uses the first sample image and the second sample image to perform dictionary training to obtain the above-mentioned dictionary matrix and mapping matrix.

[0191] Accordingly, the process of dictionary training may include: obtaining a comprehensive sampling matrix based on at least one type of soft tissue image block in the first sample image and at least one type of soft tissue image block in the second sample image. Taking the second condition as a constraint condition, eigendecomposing the comprehensive sampling matrix to obtain a dictionary matrix corresponding to the at least one type of soft tissue image block. Further, storing the dictionary matrix corresponding to the at least one type of soft tissue image block accordingly.

[0192] The comprehensive sampling matrix is ​​used to characterize the soft tissue features of the same soft tissue image block in the first sample image and the second sample image. For example, taking the first sample image as a CBCT image and the second sample image as a CT image, the comprehensive sampling matrix may include the soft tissue features of the CBCT image and the CT image.

[0193] In some embodiments, after acquiring the first sample image and the second sample image, deformation registration is performed on the first sample image and the second sample image. In this way, through deformation registration of the first sample image and the second sample image, it is ensured that the feature structure and spatial position of the first sample image and the second sample image are matched, thereby effectively enhancing the accuracy of the dictionary learning algorithm.

[0194] After performing the above-mentioned deformable registration, the first sample image and the second sample image can be subjected to bone segmentation processing respectively using a bone segmentation algorithm or a bone segmentation model, and then the first sample image and the second sample image can be divided into bone image blocks and soft tissue image blocks by sliding window segmentation. Furthermore, the soft tissue image blocks in the first sample image and the second sample image can be classified using a clustering algorithm or a multi-classification model to obtain at least one type of soft tissue image blocks in the first sample image and the second sample image.

[0195] Among them, the second condition includes: the comprehensive sampling matrix and the dictionary matrix obtained by decomposing the comprehensive sampling matrix meet the third correlation condition, and the weight coefficient matrix obtained by decomposing the comprehensive sampling matrix meets the second sparsity condition.

[0196] In some embodiments, for each type of soft tissue image block in the at least one type of soft tissue image blocks, the comprehensive sampling matrix corresponding to the type of soft tissue image blocks is feature decomposed by solving the following optimization equation (7) to obtain a dictionary matrix corresponding to the type of soft tissue image blocks, thereby obtaining a dictionary matrix corresponding to the at least one type of soft tissue image blocks.

[0197]

[0198] In the formula, x i is a column in the weight matrix, ||x i || means taking the square root of each vector component after squaring it. It should be noted that ||x i||0 means that the weight coefficient matrix obtained by decomposing the comprehensive sampling matrix satisfies the second sparsity condition. u Represents the comprehensive sampling matrix corresponding to the soft tissue image block. u represents the dictionary matrix. X represents the weight matrix. It should be noted that Indicates that the comprehensive sampling matrix and the dictionary matrix decomposed from the comprehensive sampling matrix satisfy the third correlation condition. Wherein, ε represents the allowable error. It should be understood that the third correlation condition can be used to indicate A u With Du X The difference between them is less than the preset error.

[0199] For example, in some embodiments, the K-SVD method can be used to solve the above optimization equation (7) to obtain the dictionary matrix and the weight coefficient matrix after matrix decomposition. It should be noted that in the embodiment of the present disclosure, the dictionary matrix after matrix decomposition is obtained, and the obtained dictionary matrix is ​​stored accordingly, such as storing the obtained dictionary matrix in a preset storage space. In other words, there is no need to store the weight coefficient matrix after matrix decomposition.

[0200] Furthermore, after obtaining the dictionary matrix corresponding to the at least one type of soft tissue image block through training, the image server also obtains the first sampling matrix corresponding to the at least one type of soft tissue image block in the first sample image and the second sampling matrix corresponding to the at least one type of soft tissue image block in the second sample image. Based on the first sampling matrix and the dictionary matrix, the first weight coefficient matrix corresponding to the first sampling matrix is ​​determined, and based on the second sampling matrix and the dictionary matrix, the second weight coefficient matrix corresponding to the second sampling matrix is ​​determined. Based on the first weight coefficient matrix corresponding to the first sampling matrix and the second weight coefficient matrix corresponding to the second sampling matrix, the mapping matrix corresponding to the at least one type of soft tissue image block is determined. Furthermore, the mapping matrix corresponding to the at least one type of soft tissue image block is stored accordingly.

[0201] For example, in some embodiments, the image server determines the first weight coefficient matrix corresponding to the first sampling matrix based on the first sampling matrix, the dictionary matrix and the above formula (4). Furthermore, the image server determines the second weight coefficient matrix corresponding to the second sampling matrix based on the second sampling matrix, the dictionary matrix and the above formula (6). Furthermore, the image server determines the mapping matrix corresponding to the at least one type of soft tissue image block based on the first weight coefficient matrix corresponding to the first sampling matrix, the second weight coefficient matrix corresponding to the second sampling matrix and the above formula (5). Furthermore, the mapping matrix can be stored in a preset storage space.

[0202] It should be noted that, in the dictionary training stage, the sampling matrix corresponding to the soft tissue image block of the first modality in the above formula (4) (i.e., ACB ) is also the first sampling matrix. The sampling matrix corresponding to the soft tissue image block of the second modality in the above formula (6) (i.e., A CT ) is also the second sampling matrix.

[0203] In addition, it should be noted that dictionary training is an iterative training process, and the corresponding process can be: in any iterative training process, first solve the dictionary matrix, and then use the solved dictionary matrix to solve the weight coefficient matrix. Then, based on the solved weight coefficient matrix, the next iterative training process is performed to optimize the dictionary matrix until the iterative training meets the above second condition, and the characteristic decomposition result at the end of the iteration is obtained.

[0204] In the above embodiment, the first weight coefficient matrix and the second weight coefficient matrix correspond to the same dictionary matrix. In this way, the dictionary matrix is ​​kept unchanged, and the soft tissue image blocks in the original modality image are converted into soft tissue image blocks under the new modality by updating the weight coefficient matrix, that is, by updating the first weight coefficient matrix and the second weight coefficient matrix.

[0205] S408: The image server generates a target image based on the at least one type of soft tissue image blocks of the second modality and the bone image blocks in the original image.

[0206] It should be noted that the relevant contents of S408 refer to S306 and will not be described in detail.

[0207] The technical solution provided by the embodiment of the present disclosure obtains the bone image blocks and at least one type of soft tissue image blocks in the original image of the first modality, and based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converts the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality. In this way, a dictionary learning-based image generation method is provided, which can convert the soft tissue image blocks in the original modality image into soft tissue image blocks under a new modality. Furthermore, a target image is generated based on the soft tissue image blocks under the new modality in combination with the bone image blocks in the original modality image. In this way, not only the high-quality bone tissue structure information in the original modality image is retained, but also the high-quality soft tissue structure information under the new modality is converted, which can effectively improve the image quality, thereby improving the image registration accuracy.

[0208] Exemplarily, taking the original image of the first modality as a CBCT image and the target image as a pseudo CT image as an example, a method for generating a pseudo CT image from a CBCT image based on dictionary learning is provided. The corresponding process can be replaced by: the image server obtains the bone image block and at least one type of soft tissue image block in the CBCT image. From the dictionary matrices corresponding to the pre-stored multiple soft tissue categories, the dictionary matrix corresponding to the at least one type of soft tissue image block is determined. From the mapping matrices corresponding to the pre-stored multiple soft tissue categories, the mapping matrix corresponding to the at least one type of soft tissue image block is determined. The sampling matrix corresponding to at least one type of soft tissue image block of the CBCT modality is obtained. Based on the sampling matrix corresponding to at least one type of soft tissue image block of the CBCT modality and the dictionary matrix, a first weight coefficient matrix is ​​determined. Based on the first weight coefficient matrix and the mapping matrix, the second weight coefficient matrix is ​​determined. Based on the second weight coefficient matrix and the dictionary matrix, the sampling matrix corresponding to at least one type of soft tissue image block of the CT modality is determined to obtain at least one type of soft tissue image block of the CT modality. A pseudo CT image is generated based on at least one type of soft tissue image blocks of the CT modality and the bone image blocks in the CBCT image.

[0209] In this way, by converting the soft tissue image blocks in the CBCT image into the soft tissue image blocks in the pseudo CT image, the soft tissue image blocks in the pseudo CT image can obtain a higher soft tissue resolution, which is superior to the CBCT image in terms of noise, artifacts and other indicators. Furthermore, based on the soft tissue image blocks in the pseudo CT image combined with the bone image blocks in the CBCT image, the pseudo CT image retains the high-resolution bone fine structure unique to the CBCT image. In this way, the newly generated pseudo CT image not only has the high resolution advantage of the CBCT image, but also has the low noise advantage of the CT image, which can effectively improve the image quality and thus improve the image registration accuracy. In addition, in image-guided radiotherapy, the pseudo CT image can be directly generated by scanning the patient's CBCT image, and the operations such as target area delineation, dose calculation, and patient positioning can be directly performed on the pseudo CT image. In this way, the patient does not need to repeatedly collect medical images many times, which not only reduces the additional radiation dose received by the patient, but also simplifies the process of image-guided radiotherapy.

[0210] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the image generation method provided by the present disclosure.

[0211] According to an embodiment of the present disclosure, the present disclosure further provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable an electronic device to execute the image generating method provided by the present disclosure.

[0212] According to an embodiment of the present disclosure, the present disclosure further provides a computer program product, including a computer program, which implements the image generating method provided by the present disclosure when being executed by a processor.

[0213] In some embodiments, the electronic device may be the above Figure 1 The image server shown in . Figure 5 1 is a block diagram of an electronic device 500 for implementing the image generation method of an embodiment of the present disclosure. The electronic device 500 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 500 may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0214] like Figure 5 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 to a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0215] Multiple components in the electronic device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the electronic device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0216] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSP), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as the image generation method. For example, in some embodiments, the image generation method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the image generation method described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to execute the image generation method in any other appropriate manner (eg, by means of firmware).

[0217] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0218] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0219] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0220] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user, such as a cathode ray tube (CRT) or a liquid crystal display (LCD) monitor; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and the input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0221] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0222] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0223] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of this disclosure can be achieved, and this document is not limited here.

[0224] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An image generation method, characterized in that: include: Acquire a bone image block and at least one type of soft tissue image block in an original image of the first modality; wherein one type of soft tissue image block corresponds to one soft tissue category; Determining a dictionary matrix corresponding to the at least one type of soft tissue image block from dictionary matrices corresponding to a plurality of pre-stored soft tissue categories; the dictionary matrix is ​​used to characterize key features of the corresponding soft tissue image block; Based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks, converting the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality; A target image is generated based on at least one type of soft tissue image blocks of the second modality and the bone image blocks in the original image.

2. The method according to claim 1, characterized in that The dictionary matrix includes a first dictionary matrix and a second dictionary matrix; the first dictionary matrix corresponds to the first mode, and the second dictionary matrix corresponds to the second mode; the first dictionary matrix and the second dictionary matrix correspond to the same weight coefficient matrix; The converting the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks comprises: Acquire a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality; Determine a weight coefficient matrix corresponding to at least one type of soft tissue image blocks of the first modality based on a sampling matrix corresponding to the at least one type of soft tissue image blocks of the first modality and the first dictionary matrix; Based on the weight coefficient matrix and the second dictionary matrix, a sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality is determined to obtain at least one type of soft tissue image blocks of the second modality.

3. The method according to claim 2, characterized in that Before determining the dictionary matrix corresponding to the at least one type of soft tissue image block from the pre-stored dictionary matrices corresponding to the plurality of soft tissue categories, the method further comprises: Acquire a first sampling matrix corresponding to at least one type of soft tissue image blocks in a first sample image and a second sampling matrix corresponding to at least one type of soft tissue image blocks in a second sample image; the first sample image is a sample image of a first modality, and the second sample image is a sample image of a second modality; Taking the first condition as a constraint condition, performing eigendecomposition on the first sampling matrix and the second sampling matrix respectively to obtain a first dictionary matrix and a second dictionary matrix corresponding to the at least one type of soft tissue image blocks; The first dictionary matrix and the second dictionary matrix corresponding to the at least one type of soft tissue image blocks are stored correspondingly.

4. The method according to claim 1, characterized in that: The method further comprises: Determine a mapping matrix corresponding to the at least one type of soft tissue image block from mapping matrices corresponding to a plurality of pre-stored soft tissue categories; the mapping matrix is ​​used to characterize a mapping relationship between a first weight coefficient matrix and a second weight coefficient matrix; the first weight coefficient matrix corresponds to the first modality, and the second weight coefficient matrix corresponds to the second modality; the first weight coefficient matrix and the second weight coefficient matrix correspond to the same dictionary matrix; The converting the at least one type of soft tissue image blocks of the first modality into at least one type of soft tissue image blocks of the second modality based on the dictionary matrix corresponding to the at least one type of soft tissue image blocks comprises: Acquire a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality; Determining the first weight coefficient matrix based on a sampling matrix corresponding to at least one type of soft tissue image blocks of the first modality and the dictionary matrix; Determining the second weight coefficient matrix based on the first weight coefficient matrix and the mapping matrix; Based on the second weight coefficient matrix and the dictionary matrix, a sampling matrix corresponding to at least one type of soft tissue image blocks of the second modality is determined to obtain at least one type of soft tissue image blocks of the second modality.

5. The method according to claim 4, characterized in that Before determining the dictionary matrix corresponding to the at least one type of soft tissue image block from the pre-stored dictionary matrices corresponding to the plurality of soft tissue categories, the method further comprises: Based on at least one type of soft tissue image block in a first sample image and at least one type of soft tissue image block in a second sample image, a comprehensive sampling matrix is ​​obtained; the first sample image is a sample image of a first modality, and the second sample image is a sample image of a second modality; the comprehensive sampling matrix is ​​used to characterize soft tissue features of the same soft tissue image block in the first sample image and the second sample image; Taking the second condition as a constraint condition, performing feature decomposition on the comprehensive sampling matrix to obtain a dictionary matrix corresponding to the at least one type of soft tissue image blocks; The dictionary matrix corresponding to the at least one type of soft tissue image blocks is stored accordingly.

6. The method according to claim 5, characterized in that After performing feature decomposition on the comprehensive sampling matrix using the second condition as a constraint condition to obtain a dictionary matrix corresponding to the at least one type of soft tissue image blocks, the method further includes: Acquire a first sampling matrix corresponding to at least one type of soft tissue image blocks in the first sample image and a second sampling matrix corresponding to at least one type of soft tissue image blocks in the second sample image; Based on the first sampling matrix and the dictionary matrix, determine a first weight coefficient matrix corresponding to the first sampling matrix; based on the second sampling matrix and the dictionary matrix, determine a second weight coefficient matrix corresponding to the second sampling matrix; Determining a mapping matrix corresponding to the at least one type of soft tissue image blocks based on a first weight coefficient matrix corresponding to the first sampling matrix and a second weight coefficient matrix corresponding to the second sampling matrix; The mapping matrices corresponding to the at least one type of soft tissue image blocks are stored accordingly.

7. The method according to claim 3 or 5, characterized in that: The method further comprises: Perform deformation registration on the first sample image and the second sample image.

8. The method according to claim 1, characterized in that The step of acquiring the bone image block and at least one type of soft tissue image block in the original image of the first modality includes: Performing bone segmentation processing on the original image to obtain bone image blocks and soft tissue image blocks in the original image; The soft tissue image blocks are classified to obtain the at least one type of soft tissue image blocks.

9. The method according to any one of claims 1 to 8, characterized in that The original image is a cone beam computed tomography (CBCT) image, and the target image is a pseudo computed tomography (CT) image.

10. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

Images