CT metal artifact removal method and system, readable storage medium and computer

Abstract

The invention provides a CT (Computed Tomography) metal artifact removal method and system, a readable storage medium and a computer. The method comprises the following steps: performing image analysis on each medical image to obtain an image data pair; inputting the image data pair into an image output model for training to obtain an image output training model; inputting the image data pair into an image processing model for simulation degradation to obtain a degraded image; and carrying out model optimization on the image processing model by utilizing the degraded image, carrying out model fusion on the optimized image processing model and the image output training model to construct an unsupervised metal artifact removal network model, and carrying out metal artifact removal on the medical image to be processed by utilizing the unsupervised metal artifact removal network model. According to the method, through continuous training and optimization, the model can adapt to different types of metal implants, different scanning parameters and individual differences of different patients, and the method has high universality and adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method, system, readable storage medium and computer for removing CT metal artifacts. Background Art

[0002] In medical imaging diagnosis, CT scanning is a commonly used examination method, which can provide high-resolution images of the internal structure of the human body. However, when patients have metal implants (such as dentures, orthopedic fixtures, pacemakers, etc.), CT scan images often have metal artifacts. These metal artifacts mainly appear as radial or striped light and dark areas, which seriously affect the quality of the image and the accuracy of the diagnosis. When doctors interpret CT images containing metal artifacts, they may misjudge the location, size and shape of the lesion, thus affecting the formulation of treatment plans.

[0003] At present, there are mainly the following methods to eliminate CT metal artifacts: 1. Software correction method: The collected raw data is processed through a specific algorithm to try to reduce the impact of metal artifacts. However, the effect of this method is limited, and it is often difficult to completely eliminate complex metal artifacts.

[0004] 2. Hardware improvement method: For example, adopting special detector design or adjusting scanning parameters. However, this method usually requires large-scale modification of existing equipment, which is costly and may not be applicable to all types of CT scanners.

[0005] 3. Dual-energy CT technology: It uses X-rays of different energies to scan and distinguish metal from soft tissue by analyzing the difference in the absorption of X-rays of different energies by the material, thereby reducing metal artifacts. However, dual-energy CT equipment is expensive and less popular. Summary of the invention

[0006] Based on this, the purpose of the present invention is to provide a CT metal artifact removal method, system, readable storage medium and computer to solve the deficiencies in the above-mentioned technology.

[0007] The present invention provides a CT metal artifact removal method, comprising: Acquire a plurality of medical image images, and perform image analysis on each of the medical image images to obtain a corresponding image data pair, wherein metal artifact image data in the image data pair does not match normal image data; Constructing an image output model, and inputting the metal artifact image data in the image data pair into the image output model for training, so as to obtain an image output training model; Constructing an image processing model, and inputting the image data pair into the image processing model for simulated degradation to obtain a degraded image; Optimizing the image processing model using the degraded image, and fusing the optimized image processing model with the image output training model to construct an unsupervised metal artifact removal network model; A medical image to be processed is acquired, and metal artifacts are removed from the medical image to be processed using the unsupervised metal artifact removal network model.

[0008] Furthermore, the steps of constructing an image output model and inputting the metal artifact image data in the image data pair into the image output model for training to obtain an image output training model include: Extracting metal artifact image data and normal image data from the image data pair, and inputting the metal artifact image data into a reconstructor of the image output model to obtain corresponding clear image data; Using the kernel estimation network of the image output model to estimate the artifact kernel of the metal artifact image, and using the artifact kernel to perform a degradation operation on the normal image data to obtain a degraded image; Training a reconstructor of the image output model, inputting the metal artifact image into the trained reconstructor, and inputting output data of the trained reconstructor into a degradation information generation module of the image output model for processing to obtain corresponding degradation information; The clear image and the degradation information are input into a generator of the image output model to obtain a second metal artifact image.

[0009] Furthermore, in the degradation operation, by maintaining the consistency of the metal artifact image and the degraded image in the principal components, an L1 norm loss is introduced to constrain the generator of the image output model. The specific calculation formula is: ; In the formula, represents the artifact kernel estimation process, represents an artifact operation, The kernel size is The Gaussian filter operator is is the weight of the corresponding scale.

[0010] Furthermore, the step of training the reconstructor of the image output model includes: Initializing a reconstructor of the image output model, fixing parameters of the reconstructor, and guiding a generator of the image output model to learn artifact degradation information; The generator, discriminator, and reconstructor of the image output model are retrained until the reconstructor converges to determine an optimized reconstructor loss.

[0011] Furthermore, the expression of the optimized reconstructor loss is: ; In the formula, Rec represents the reconstructor before optimization, represents the synthesized pseudo metal artifact image, express The corresponding image after artifact removal, y represents the real metal artifact image, represents the image after processing the real metal artifact image y, Represents the original loss function of the reconstructor before optimization.

[0012] The present invention also provides a CT metal artifact removal system, comprising: An image analysis module, used for acquiring a plurality of medical image images, and performing image analysis on each of the medical image images to obtain a corresponding image data pair, wherein the metal artifact image data in the image data pair does not match the normal image data; A first model building module is used to build an image output model, and input the metal artifact image data in the image data pair into the image output model for training to obtain an image output training model; A second model building module is used to build an image processing model and input the image data pair into the image processing model to simulate degradation to obtain a degraded image; A model optimization module, used to optimize the image processing model using the degraded image, and to fuse the optimized image processing model with the image output training model to construct an unsupervised metal artifact removal network model; The artifact removal module is used to obtain the medical image to be processed and use the unsupervised metal artifact removal network model to remove metal artifacts from the medical image to be processed.

[0013] Furthermore, the first model building module includes: a data extraction unit, used for extracting metal artifact image data and normal image data from the image data pair, and inputting the metal artifact image data into a reconstructor of the image output model to obtain corresponding clear image data; A degradation operation unit, configured to estimate an artifact kernel of the metal artifact image by using a kernel estimation network of the image output model, and perform a degradation operation on the normal image data by using the artifact kernel to obtain a degraded image; A model training unit, used for training a reconstructor of the image output model, inputting the metal artifact image into the trained reconstructor, and inputting the output data of the trained reconstructor into a degradation information generation module of the image output model for processing to obtain corresponding degradation information; An image processing unit is used to input the clear image and the degradation information into a generator of the image output model to obtain a second metal artifact image.

[0014] Furthermore, the model training unit is specifically used for: Initializing a reconstructor of the image output model, fixing parameters of the reconstructor, and guiding a generator of the image output model to learn artifact degradation information; The generator, discriminator, and reconstructor of the image output model are retrained until the reconstructor converges to determine an optimized reconstructor loss.

[0015] The present invention also provides a readable storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned CT metal artifact removal method is implemented.

[0016] The present invention also provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned CT metal artifact removal method when executing the computer program.

[0017] The CT metal artifact removal method, system, readable storage medium and computer of the present invention obtain an image output training model by inputting data pairs obtained by image analysis of medical images into an image output model for training, and adopting a deep learning algorithm to automatically learn features and patterns in a large amount of CT image data containing metal artifacts and artifacts, so as to more accurately identify and eliminate metal artifacts; the image output training model is simulated degraded using metal artifact images and artifact-free images, and the image output training model is optimized using the degraded images obtained by simulated degradation, so as to construct an unsupervised metal artifact removal network model. Through continuous training and optimization, the model can adapt to different types of metal implants, different scanning parameters and individual differences of different patients, and has strong versatility and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of a CT metal artifact removal method in a first embodiment of the present invention; Figure 2 for Figure 1 Detailed flow chart of step S102; Figure 3A diagram showing specific implementation steps of the image output model in the first embodiment of the present invention; Figure 4 This is a schematic diagram of the network structure of the DGIG module in the first embodiment of the present invention; Figure 5 This is a schematic diagram of the network structure of a reconstructor in the first embodiment of the present invention; Figure 6 This is a diagram of specific implementation steps of the unsupervised metal artifact removal network model in the first embodiment of the present invention; Figure 7 It is a structural block diagram of a CT metal artifact removal system in a second embodiment of the present invention; Figure 8 FIG. 4 is a structural block diagram of a computer in a third embodiment of the present invention.

[0019] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0020] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0022] Embodiment 1 See also Figure 1 , which shows a CT metal artifact removal method in a first embodiment of the present invention, and the method specifically includes steps S101 to S105: S101, acquiring a plurality of medical images, and performing image analysis on each of the medical images to obtain a corresponding image data pair, wherein metal artifact image data in the image data pair does not match normal image data; In the specific implementation, due to the radiation risk of CT scanning, it is impossible to obtain a completely matched data pair. The complete match specifically refers to the consistency of the real spatial position of the image matrix. Therefore, a certain number of CT metal artifact sequences and a certain number of normal CT sequences are selected to ensure that the scanning parts of the two sequences are consistent.

[0023] Specifically, a number of medical images are acquired, and a value distribution histogram of the medical images is analyzed and normalized to obtain mismatched CT metal artifact / normal CT image data pairs.

[0024] S102, constructing an image output model, and inputting the metal artifact image data in the image data pair into the image output model for training, so as to obtain an image output training model; For further information, see Figure 2 , the step S102 specifically includes steps S1021 to S1024: S1021, extracting metal artifact image data and normal image data from the image data pair, and inputting the metal artifact image data into a reconstructor of the image output model to obtain corresponding clear image data; S1022, estimating an artifact kernel of the metal artifact image by using the kernel estimation network of the image output model, and performing a degradation operation on the normal image data by using the artifact kernel to obtain a degraded image; S1023, training a reconstructor of the image output model, inputting the metal artifact image into the trained reconstructor, and inputting output data of the trained reconstructor into a degradation information generation module of the image output model for processing to obtain corresponding degradation information; S1024: Input the clear image and the degradation information into a generator of the image output model to obtain a second metal artifact image.

[0025] In this example, see Figure 3 The image output model includes a generator, a discriminator and a reconstructor based on a residual network. The generator uses unpaired CT images containing metal artifacts and artifact-free CT images to synthesize a pseudo-paired dataset to provide data support for removing metal artifacts; the discriminator is used to distinguish between synthesized pseudo-metal artifact images and real metal artifact images, and improve the quality of the generator's synthesized images through adversarial generation training; the reconstructor is responsible for converting the synthesized pseudo-metal artifact images into de-artifacted images.

[0026] Specifically, the structure of the generator includes a sharp feature extraction module and multiple residual blocks. The sharp feature extraction module is used to learn the features of the CT image without metal artifacts, and the residual blocks are used to simulate the generation process of metal artifacts and convert the artifact-free image into an image with metal artifacts.

[0027] Further, extracting the metal artifact image data and the normal image data from the above image data pair, inputting the metal artifact image data into the reconstructor of the image output model to obtain the corresponding clear image data, estimating the artifact kernel of the metal artifact image using the kernel estimation network of the image output model, and performing a degradation operation on the above normal image data using the artifact kernel to obtain a degraded image; In this embodiment, in order to synthesize a more realistic metal artifact image, a DGIG module (specific structure is as follows: Figure 4 As shown in Figure 1). This module uses a U-Net structure to learn the features of metal artifact images and guide the image output training model to synthesize pseudo metal artifact images that are closer to the real situation. During the training process, the generator and discriminator of the model are optimized through adversarial loss, so that the generated pseudo metal artifact images are visually similar to the real metal artifact images.

[0028] Furthermore, in order to reduce the influence of noise and other interference factors in the process of synthesizing artifact images and improve the similarity between the synthesized images and the real metal artifact images, the degradation principal component consistency loss DPC (degradation principal component consistency) is designed. Specifically, the kernel estimation network is first used to estimate the artifact kernel of the real metal artifact image, and then the artifact kernel is used to degrade the artifact-free image to obtain a degraded image. By maintaining the consistency of the principal components of the synthesized metal artifact image and the re-degraded image, the L1 norm loss is introduced to constrain the learning of the generator. The specific formula is: ; In the formula, represents the artifact kernel estimation process, represents an artifact operation, The kernel size is The Gaussian filter operator is is the weight of the corresponding scale.

[0029] By introducing the degraded principal component consistency loss, the generator can be guided to simulate the degradation process of metal artifacts more accurately, thereby improving the quality of the pseudo-paired dataset, which in turn helps the reconstructor learn an effective de-metal artifact mapping.

[0030] In some optional embodiments, in the above steps, the step of training the reconstructor of the image output model includes: Initializing a reconstructor of the image output model, fixing parameters of the reconstructor, and guiding a generator of the image output model to learn artifact degradation information; The generator, discriminator, and reconstructor of the image output model are retrained until the reconstructor converges to determine an optimized reconstructor loss.

[0031] In the specific implementation, in order to improve its performance without changing the reconstructor structure core of the model and increasing the reasoning complexity of the original network, a self-enhancement strategy is constructed in this embodiment. Specifically, the initial reconstructor Rec1 of the model is first trained, and the parameters of Rec1 are fixed to guide the generator of the model to learn more accurate artifact degradation information and generate more realistic artifact metal images.

[0032] Secondly, retrain the generator, discriminator (the discriminator adopts the PatchGAN structure) and reconstructor until the reconstructor converges. Repeat the steps of guiding the generator learning of the model, gradually update and enhance the reconstructor, until the best performance reconstructor Recn is obtained (the specific structure is as follows Figure 5 In this way, the reconstructor can use the results of the previous stage as feedback information to continuously improve its own performance, thereby achieving more effective removal of metal artifacts.

[0033] Furthermore, the expression of the optimized reconstructor loss is: ; In the formula, Rec represents the reconstructor before optimization, represents the synthesized pseudo metal artifact image, express The corresponding image after artifact removal, y represents the real metal artifact image, represents the image after processing the real metal artifact image y, Represents the original loss function of the reconstructor before optimization.

[0034] S103, constructing an image processing model, and inputting the image data pair into the image processing model for simulated degradation to obtain a degraded image; S104, optimizing the image processing model using the degraded image, and fusing the optimized image processing model with the image output training model to construct an unsupervised metal artifact removal network model; In the specific implementation, the constructed unsupervised metal artifact removal network model mainly includes the backbone module B (backbone) and three generator complementary constraint modules GCCU (generator complementary constraint unit). Among them, the forward reasoning process FP (forward propagation) of different modules is: 1. Main module: ; ; 2. : ; 3. : ; ; 4. : ; Furthermore, the loss function of the model is: ; ; ; After the above steps, all loss functions are combined, that is, adversarial loss, degraded principal component consistency loss, self-enhanced reconstructor loss (optimized reconstructor loss), to obtain a total loss function, and the image output training model is iteratively optimized using the total loss function to construct an unsupervised metal artifact removal network model. In this embodiment, each loss function has a corresponding hyperparameter to balance the effects of different loss terms in the training process, thereby optimizing the performance of the entire model. The total loss function obtained is:

[0035] S105, obtaining a medical image to be processed, and using the unsupervised metal artifact removal network model to remove metal artifacts from the medical image to be processed.

[0036] In this embodiment, the goal of training the image training model is to minimize the difference between the de-artifacted image and the original artifact-free image, and the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) loss functions are used to constrain the output of the reconstructor to make it as close to the real artifact-free image as possible.

[0037] See also Figure 6 , which is a detailed step diagram of the CT metal artifact removal method in this embodiment, has the effect of efficient artifact removal and strong adaptability compared to the existing software correction method. Specifically, the deep learning algorithm can automatically learn the features and patterns in a large number of CT image data containing metal artifacts and artifact-free, so as to more accurately identify and eliminate metal artifacts. Compared with the traditional software correction method, its elimination effect is more significant. Even in the case of large metal implants or complex shapes, it can effectively restore the real structure of the image, greatly improving the image quality and diagnostic accuracy; through continuous training and optimization, the deep learning model can adapt to different types of metal implants, different scanning parameters and individual differences of different patients, and has strong versatility and adaptability.

[0038] Compared with the existing hardware improvement method, the CT metal artifact removal method in this embodiment has the characteristics of low cost and no impact on the stability and compatibility of the equipment. Specifically, the method based on deep learning does not require large-scale hardware modification of existing CT equipment, but only requires the deployment and update of algorithms at the software level, which greatly reduces the cost. For medical institutions, the quality of CT images can be improved without increasing a large amount of capital investment; since no hardware modification is involved, the stability and compatibility of CT equipment will not be affected, and it can be seamlessly used with various existing models of CT scanners without worrying about the risk of equipment failure caused by hardware modification.

[0039] Compared with the existing dual-energy CT technology, the CT metal artifact removal method in this embodiment is economical and easy to implement. Specifically, compared with expensive dual-energy CT equipment, the metal artifact removal method based on deep learning is extremely low in cost and can be easily afforded by both small medical institutions and large hospitals, which is conducive to the wide promotion and application of this technology. The method in this embodiment does not require special equipment, but only requires the addition of a deep learning algorithm processing link to the existing CT scanning process, which allows medical institutions to quickly deploy the technology without the need for complex equipment installation and debugging.

[0040] In summary, the CT metal artifact removal method in the above-mentioned embodiment of the present invention obtains an image output training model by inputting the data pairs obtained by image analysis of medical images into the image output model for training, and adopts a deep learning algorithm to automatically learn the features and patterns in a large amount of CT image data containing metal artifacts and artifacts, so as to more accurately identify and eliminate metal artifacts; the image output training model is simulated to be degraded using metal artifact images and artifact-free images, and the image output training model is optimized using the degraded images obtained by simulated degradation, so as to construct an unsupervised metal artifact removal network model. Through continuous training and optimization, the model can adapt to different types of metal implants, different scanning parameters and individual differences of different patients, and has strong versatility and adaptability.

[0041] Embodiment 2 Another aspect of the present invention is to provide a CT metal artifact removal system. Figure 7 , which is a CT metal artifact removal system in a second embodiment of the present invention, and the system comprises: An image analysis module 11, used for acquiring a plurality of medical images, and performing image analysis on each of the medical images to obtain a corresponding image data pair, wherein the metal artifact image data in the image data pair does not match the normal image data; A first model building module 12 is used to build an image output model, and input the metal artifact image data in the image data pair into the image output model for training to obtain an image output training model; Furthermore, the first model building module 12 includes: a data extraction unit, used for extracting metal artifact image data and normal image data from the image data pair, and inputting the metal artifact image data into a reconstructor of the image output model to obtain corresponding clear image data; A degradation operation unit, configured to estimate an artifact kernel of the metal artifact image by using a kernel estimation network of the image output model, and perform a degradation operation on the normal image data by using the artifact kernel to obtain a degraded image; A model training unit, used for training a reconstructor of the image output model, inputting the metal artifact image into the trained reconstructor, and inputting the output data of the trained reconstructor into a degradation information generation module of the image output model for processing to obtain corresponding degradation information; Furthermore, the model training unit is specifically used for: Initializing a reconstructor of the image output model, fixing parameters of the reconstructor, and guiding a generator of the image output model to learn artifact degradation information; The generator, discriminator, and reconstructor of the image output model are retrained until the reconstructor converges to determine an optimized reconstructor loss.

[0042] An image processing unit is used to input the clear image and the degradation information into a generator of the image output model to obtain a second metal artifact image.

[0043] A second model building module 13 is used to build an image processing model and input the image data pair into the image processing model to simulate degradation to obtain a degraded image; A model optimization module 14 is used to optimize the image processing model using the degraded image, and to fuse the optimized image processing model with the image output training model to construct an unsupervised metal artifact removal network model; The artifact removal module 15 is used to obtain the medical image to be processed, and use the unsupervised metal artifact removal network model to remove metal artifacts from the medical image to be processed.

[0044] The functions or operation steps implemented when the above modules and units are executed are generally the same as those in the above method embodiments, and will not be repeated here.

[0045] The CT metal artifact removal system provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the system embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0046] Embodiment 3 The present invention also provides a computer, see Figure 8 , shown is a computer in the third embodiment of the present invention, including a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the computer program 30, the above-mentioned CT metal artifact removal method is implemented.

[0047] The memory 10 includes at least one type of readable storage medium, which includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 10 may be an internal storage unit of a computer, such as a hard disk of the computer. In other embodiments, the memory 10 may also be an external storage device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. Further, the memory 10 may also include both an internal storage unit of the computer and an external storage device. The memory 10 may be used not only to store application software and various types of data installed in the computer, but also to temporarily store data that has been output or is to be output.

[0048] Among them, in some embodiments, the processor 20 can be an electronic control unit (Electronic Control Unit, abbreviated as ECU, also known as a vehicle computer), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in the memory 10 or process data, such as executing access restriction programs, etc.

[0049] It should be pointed out that Figure 8 The structure shown does not constitute a limitation on the computer. In other embodiments, the computer may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0050] The embodiment of the present invention further provides a readable storage medium on which a computer program is stored. When the program is executed by a processor, the CT metal artifact removal method as described above is implemented.

[0051] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0052] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0053] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0054] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0055] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for removing CT metal artifacts, characterized in that: include: Acquire a plurality of medical image images, and perform image analysis on each of the medical image images to obtain a corresponding image data pair, wherein metal artifact image data in the image data pair does not match normal image data; Constructing an image output model, and inputting the metal artifact image data in the image data pair into the image output model for training, so as to obtain an image output training model; Constructing an image processing model, and inputting the image data pair into the image processing model for simulated degradation to obtain a degraded image; Optimizing the image processing model using the degraded image, and fusing the optimized image processing model with the image output training model to construct an unsupervised metal artifact removal network model; A medical image to be processed is acquired, and metal artifacts are removed from the medical image to be processed using the unsupervised metal artifact removal network model.

2. The CT metal artifact removal method according to claim 1, characterized in that: The steps of constructing an image output model and inputting the metal artifact image data in the image data pair into the image output model for training to obtain an image output training model include: Extracting metal artifact image data and normal image data from the image data pair, and inputting the metal artifact image data into a reconstructor of the image output model to obtain corresponding clear image data; Using the kernel estimation network of the image output model to estimate the artifact kernel of the metal artifact image, and using the artifact kernel to perform a degradation operation on the normal image data to obtain a degraded image; Training a reconstructor of the image output model, inputting the metal artifact image into the trained reconstructor, and inputting output data of the trained reconstructor into a degradation information generation module of the image output model for processing to obtain corresponding degradation information; The clear image and the degradation information are input into a generator of the image output model to obtain a second metal artifact image.

3. The CT metal artifact removal method according to claim 2, characterized in that: In the degradation operation, by maintaining the consistency of the metal artifact image and the degraded image in the principal components, an L1 norm loss is introduced to constrain the generator of the image output model. The specific calculation formula is: ； In the formula, represents the artifact kernel estimation process, represents an artifact operation, The kernel size is The Gaussian filter operator is is the weight of the corresponding scale.

4. The CT metal artifact removal method according to claim 2, characterized in that: The step of training the reconstructor of the image output model comprises: Initializing a reconstructor of the image output model, fixing parameters of the reconstructor, and guiding a generator of the image output model to learn artifact degradation information; The generator, discriminator, and reconstructor of the image output model are retrained until the reconstructor converges to determine an optimized reconstructor loss.

5. The CT metal artifact removal method according to claim 4, characterized in that: The expression of the optimized reconstructor loss is: ； In the formula, Rec represents the reconstructor before optimization, represents the synthesized pseudo metal artifact image, express The corresponding image after artifact removal, y represents the real metal artifact image, represents the image after processing the real metal artifact image y, Represents the original loss function of the reconstructor before optimization.

6. A CT metal artifact removal system, characterized in that: include: An image analysis module, used for acquiring a plurality of medical image images, and performing image analysis on each of the medical image images to obtain a corresponding image data pair, wherein the metal artifact image data in the image data pair does not match the normal image data; A first model building module is used to build an image output model, and input the metal artifact image data in the image data pair into the image output model for training to obtain an image output training model; A second model building module is used to build an image processing model and input the image data pair into the image processing model to simulate degradation to obtain a degraded image; A model optimization module, used to optimize the image processing model using the degraded image, and to fuse the optimized image processing model with the image output training model to construct an unsupervised metal artifact removal network model; The artifact removal module is used to obtain the medical image to be processed and use the unsupervised metal artifact removal network model to remove metal artifacts from the medical image to be processed.

7. The CT metal artifact removal system according to claim 6, characterized in that: The first model building module includes: a data extraction unit, used for extracting metal artifact image data and normal image data from the image data pair, and inputting the metal artifact image data into a reconstructor of the image output model to obtain corresponding clear image data; A degradation operation unit, configured to estimate an artifact kernel of the metal artifact image by using a kernel estimation network of the image output model, and perform a degradation operation on the normal image data by using the artifact kernel to obtain a degraded image; A model training unit, used for training a reconstructor of the image output model, inputting the metal artifact image into the trained reconstructor, and inputting the output data of the trained reconstructor into a degradation information generation module of the image output model for processing to obtain corresponding degradation information; An image processing unit is used to input the clear image and the degradation information into a generator of the image output model to obtain a second metal artifact image.

8. The CT metal artifact removal system according to claim 6, characterized in that: The model training unit is specifically used for: Initializing a reconstructor of the image output model, fixing parameters of the reconstructor, and guiding a generator of the image output model to learn artifact degradation information; The generator, discriminator, and reconstructor of the image output model are retrained until the reconstructor converges to determine an optimized reconstructor loss.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the CT metal artifact removal method according to any one of claims 1 to 5 is implemented.

10. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the CT metal artifact removal method according to any one of claims 1 to 5 is implemented.

Images