Neural network model for CT image restoration and training method, device and equipment

Abstract

The invention provides a neural network model for CT image restoration and a training method, device and equipment, and relates to the technical field of deep learning, and the training method for the neural network model for CT image restoration comprises the steps: obtaining a sample image set; training a preset neural network model by using the sample image set to obtain a gradient value corresponding to a loss function; weight parameters of the preset neural network model are updated based on the reverse direction of the gradient value, and iterative training is carried out until a loss function corresponding to an image output by the preset neural network model reaches a preset value; and determining the corresponding preset neural network model when the loss function reaches a preset value as a target neural network model, and using the target neural network to complete restoration of the to-be-restored image. According to the method, the neural model is reasonably optimized, so that the efficiency and the accuracy of CT image restoration are improved.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a neural network model and training method, device and equipment for CT image restoration. Background Art

[0002] CT images are the first digital images to appear, bringing medical imaging into the era of digital images. New digital imaging technologies such as nuclear magnetic resonance imaging have subsequently emerged, allowing doctors to obtain disease-related information more efficiently. CT imaging technology uses X-rays for tomography, uses sensors instead of film to receive transmitted X-rays, and combines computer processing to generate cross-sectional images, which can provide doctors with detailed anatomical information and help clinical diagnosis and treatment planning. When performing CT scanning imaging, higher radiation doses can provide clearer images, but they increase the risk of radiation exposure to patients, so doctors need to weigh the relationship between radiation dose and imaging quality; when scanning, the patient's posture and movement or bones and metal objects will also affect the quality of the image, causing image distortion or artifacts in a certain area. Therefore, in practice, it is often necessary to use corresponding image processing technology to improve the imaging effect of these specific area images.

[0003] Although current medical image restoration algorithms can accomplish the corresponding tasks, their restoration quality often varies greatly when faced with complex real images. To achieve a better restoration effect, the algorithm often needs to be adjusted multiple times according to the image, which reduces the efficiency of image restoration and makes it difficult to ensure the final restoration quality. Summary of the invention

[0004] The purpose of the present invention is to solve the problems of low restoration efficiency and poor restoration quality of current medical image restoration algorithms, and to provide a neural network model and training method, device and equipment for CT image restoration.

[0005] The technical solution of the embodiment of the present application is implemented as follows: A first aspect of an embodiment of the present application provides a method for training a neural network model for CT image restoration, comprising: Get a sample image set; Using the sample image set to train a preset neural network model to obtain a gradient value corresponding to a loss function; Based on the reverse direction of the gradient value, the weight parameters of the preset neural network model are updated and iterative training is performed until the loss function corresponding to the image output by the preset neural network model reaches a preset value; The preset neural network model corresponding to when the loss function reaches a preset value is determined as the target neural network model.

[0006] Optionally, the sample image set includes a training set and a label set, and the data in the training set is obtained by randomly labeling the data in the label set; and the using the sample image set to train a preset neural network model to obtain a gradient value corresponding to a loss function includes: Inputting the training set into the preset neural network model to obtain an output image; determining a loss function based on the output image and the label set; The back propagation algorithm is used to determine the gradient value corresponding to the loss function.

[0007] Optionally, the sample image set and the label set include labeled images in which preset areas of the images are labeled; and the using the sample image set to train a preset neural network model to obtain a gradient value corresponding to a loss function further includes: Inputting the marked image into the preset neural network model to obtain an output area image corresponding to a preset area in the marked image; A loss function is determined based on the output region image and a corresponding preset region in the label set.

[0008] Optionally, updating the weight parameters of the preset neural network model based on the reverse direction of the gradient value and performing iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value includes: Using a gradient descent algorithm to update the weight parameters of the preset neural network model, and obtaining the loss function of the preset neural network model after each update; If the loss function reaches a preset value, the updating operation is stopped.

[0009] Optionally, the loss function includes: a restoration loss function and a latent space loss function; the restoration loss function characterizes the restoration loss of image input and output; the latent space loss function characterizes the difference between the latent space distribution and the standard normal distribution.

[0010] Optionally, the method further includes: inputting the image to be repaired into the target neural network model, and outputting a target image.

[0011] The second aspect of the embodiment of the present application provides a training device for a neural network model of CT image restoration, comprising: an acquisition module, a training module, an iteration module and a determination module, wherein: The acquisition module is configured to acquire a sample image set; The training module is configured to train a preset neural network model using the sample image set to obtain a gradient value corresponding to a loss function; The iteration module is configured to update the weight parameters of the preset neural network model based on the reverse direction of the gradient value and perform iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value; The determination module is configured to determine the preset neural network model corresponding to when the loss function reaches a preset value as the target neural network model.

[0012] A third aspect of an embodiment of the present application provides a neural network model for CT image restoration, wherein the neural network model is a variational autoencoder including an attention module, and the variational autoencoder is trained based on the training method of the neural network model for CT image restoration described in the first aspect.

[0013] A fourth aspect of an embodiment of the present application provides an electronic device, comprising a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the training method of the neural network model for CT image restoration described in the first aspect.

[0014] A fifth aspect of an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0015] Compared with the prior art, the technical solution provided by this application has the following beneficial effects: The present invention provides a neural network model and training method, device and equipment for CT image restoration. The method obtains a sample image set, uses the sample image set to train a preset neural network model, obtains a gradient value corresponding to a loss function, updates a weight parameter of the preset neural network model based on the reverse direction of the gradient value and performs iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value, determines the preset neural network model corresponding to the loss function reaching the preset value as a target neural network model, and uses the target neural network to complete the restoration of the image to be restored, thereby improving the efficiency and quality of CT image restoration. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a training method for a neural network model for CT image restoration provided in an embodiment of the present application; Figure 2 A schematic diagram of a restoration result of a medical CT scan image provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a training device for a neural network model for CT image restoration provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of a neural network model for CT image restoration is provided for an embodiment of the present application; Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present application. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present application. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present application. The terms used herein are only for describing specific embodiments and are not intended to limit the present application. The terms "include", "comprising", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components. All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0018] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some blocks or combinations thereof in the block diagrams and / or flow charts may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions, when executed by the processor, may create a device for implementing the functions / operations described in these block diagrams and / or flow charts.

[0019] In some embodiments, see Figure 1 , Figure 1 A flow chart of a training method for a neural network model for CT image restoration provided in an embodiment of the present application; the training method for a neural network model for CT image restoration provided in an embodiment of the present application includes: S110, obtaining a sample image set.

[0020] Here, the sample image set may be an image of a type similar to that of the image to be restored. For example, if the image to be restored is a CT image, the sample image set may include a large number of similar CT images.

[0021] S120, using the sample image set to train the preset neural network model to obtain the gradient value corresponding to the loss function.

[0022] The preset neural network can be composed of two sub-networks, namely an encoder network and a decoder network. The attention mechanism is used in the two sub-networks to process the preset features of the sample image set. By inputting the images in the sample image set into the preset neural network model in batches, the loss function corresponding to the input image and the gradient value corresponding to the loss function can be obtained. Here, the loss function represents the difference between the output image of the preset neural network model and the real image, and the gradient value represents the optimization direction and optimization rate of the network weight parameters.

[0023] S130, updating the weight parameters of the preset neural network model based on the reverse direction of the gradient value and performing iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value.

[0024] The loss function is back-propagated, and the loss function can be reduced after the network is optimized. It can be understood that after each loss function is obtained, the weight parameters of the preset neural network model are adjusted and updated based on the loss function to reduce the loss function. As the number of iterations increases, the loss function gradually decreases. When the loss function tends to be stable, or when the loss function reaches the preset value, it means that the preset neural network model has met the image processing requirements and the iteration can be stopped. Here, the specific size of the preset value can be determined according to the image accuracy requirements, and there is no restriction on this.

[0025] S140, determining the preset neural network model corresponding to when the loss function reaches a preset value as the target neural network model.

[0026] Here, the preset value is a value close to 0. When the loss function reaches the preset value, the preset neural network model meets the image processing requirements. The weight parameters corresponding to the current preset neural network model are recorded and used as the weight parameters of the target neural network model.

[0027] In some embodiments, the sample image set includes a training set and a label set, and the data in the training set is obtained by randomly labeling the data in the label set; S120, using the sample image set to train a preset neural network model to obtain a gradient value corresponding to the loss function, including: Input the training set into the preset neural network model to obtain the output image; Determining a loss function based on the output image and the label set; The back propagation algorithm is used to determine the gradient value corresponding to the loss function.

[0028] In this embodiment, the sample image set is CT image data as an example for illustration. The preset neural network model may include an encoder network and a decoder network. The encoder network is used to extract the features of the input CT image data and encode it into a latent space distribution. The decoder network is used to sample from the latent space distribution, and then decode to generate the repaired CT image data. The loss function is used to calculate the difference between the complete CT image data and the corresponding data in the label set, and then the back propagation algorithm is used to calculate the gradient of each weight parameter in the encoder network and the decoder network, and then each weight parameter is updated according to the gradient, so that the complete CT image data output by the neural network is close to the original CT image data in the label set.

[0029] In some embodiments, the sample image set and the label set include a labeled image in which a preset area of ​​the image is labeled; S120, using the sample image set to train a preset neural network model to obtain a gradient value corresponding to the loss function, further comprising: Input the marked image into a preset neural network model to obtain an output region image corresponding to a preset region in the marked image; The loss function is determined based on the output region image and the corresponding preset region in the label set.

[0030] In this embodiment, the attention mechanism can be applied to complete the training of the preset neural network model. During the training process, the number of times the preset neural network model needs to loop each time to generate a complete image is determined according to the set hyperparameters. In each loop, the encoder selectively reads part of the structure according to the characteristics of the training set image data, and encodes this part of the structure into a latent space. The decoder samples from this latent space to obtain a latent vector to construct this part of the training set image data. After completing all the loops set in the hyperparameters, the complete output image data can be obtained. Through local sampling and reorganization, the ability of the preset neural network model to perform local repair on the image to be repaired can be improved, and the generalization performance of the preset neural network model can be improved.

[0031] In some embodiments, S130, updating the weight parameters of the preset neural network model based on the reverse direction of the gradient value and performing iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value includes: Use the gradient descent algorithm to update the weight parameters of the preset neural network model, and obtain the loss function after each update of the preset neural network model; If the loss function reaches the preset value, the update operation is stopped.

[0032] The gradient represents the rate of change of the loss function. By calculating the gradient, we can clearly understand the direction of parameter change and its impact on the loss function. In the loss function, the contribution of each network parameter can be measured by its gradient. By using the chain rule, the gradient can be calculated layer by layer until the gradient of the output layer is obtained. Here, by continuously calculating the gradient of the loss function and updating the network parameters, the loss value is gradually reduced.

[0033] In some embodiments, the loss function includes: a restoration loss function and a latent space loss function; the restoration loss function represents the restoration loss of the image input and output; the latent space loss function represents the difference between the latent space distribution and the standard normal distribution.

[0034] During the training process, the recovery loss function can be used to guide the preset neural network model to learn the characteristics of the sample image set. Represents the output of the preset neural network model, Represents the image data corresponding to the label set, then the loss function is restored It can be expressed by the following formula: ; The latent space loss function can be used to measure the difference between the latent space distribution and the standard normal distribution, and is used to constrain the output of the encoder network. Assume that the latent space distribution obtained by the encoder network is , the normal distribution is , then the latent space loss function It can be expressed as: ; Neural network presets the overall loss function of the neural network model It can be expressed as: ; in and It is the calculation coefficient of the two losses, which is used to adjust the degree of constraint on network optimization.

[0035] In some embodiments, after determining the target neural network model, the method further includes: inputting the image to be repaired into the target neural network model, and outputting the target image.

[0036] The image to be restored is an image of the same or similar type as the image in the sample image set. By inputting the image to be restored into the target neural network model, a high-quality restored image can be obtained.

[0037] In one instance, see Figure 2 , Figure 2A schematic diagram of the restoration result of a medical CT scan image provided in an embodiment of the present application. Here, a CT scanned colon image is used as experimental data, and the image size is . Figure 2 (a) is a schematic diagram of the area to be restored randomly selected by this method, indicating that the area was disturbed during imaging. Figure 2 (b) is a schematic diagram of the original CT scan images, which are used as the label set during training. Figure 2 (c) is a schematic diagram of a CT scan image after the relevant area information has been deleted, which can provide the neural network with relevant information around the area to be restored. Figure 2 (d) is a CT scan colon image restored by the trained neural network. Through image comparison, it can be seen that the target neural network model can generate image information of the area to be restored based on the information provided by the surrounding area. Although the image information generated by the target neural network model is not clear enough and has obvious differences from the surrounding areas, the basic structure of the image can be well restored.

[0038] The embodiment of the present application obtains a sample image set, uses the sample image set to train a preset neural network model, obtains the gradient value corresponding to the loss function, updates the weight parameters of the preset neural network model in the opposite direction of the gradient value and performs iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value, determines the preset neural network corresponding to the preset value of the loss function as the target neural network, and uses the target neural network to complete the restoration of the image to be restored, thereby improving the efficiency and quality of image processing.

[0039] In some embodiments, see Figure 3 , Figure 3 A structural diagram of a training device for a neural network model for CT image restoration provided in an embodiment of the present application. The training device 300 for a neural network model for CT image restoration provided in an embodiment of the present application includes: an acquisition module 310, a training module 320, an iteration module 330 and a determination module 340, wherein: An acquisition module 310 is configured to acquire a sample image set; The training module 320 is configured to train the preset neural network model using the sample image set to obtain the gradient value corresponding to the loss function; An iteration module 330 is configured to update the weight parameters of the preset neural network model based on the reverse direction of the gradient value and perform iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value; The determination module 340 is configured to determine the preset neural network model corresponding to when the loss function reaches a preset value as the target neural network model.

[0040] In some embodiments, the sample image set includes a training set and a label set; the training module 320 is specifically configured as follows: Input the training set into the preset neural network model to obtain the output image; Calculate the loss function based on the output image and label set; The back propagation algorithm is used to determine the gradient value corresponding to the loss function.

[0041] In some embodiments, the sample image set and the label set include labeled images in which the preset areas of the images are labeled; the training module 320 is further configured as follows: Input the marked image into a preset neural network model to obtain an output region image corresponding to a preset region in the marked image; The loss function is determined based on the output region image and the corresponding preset region in the label set.

[0042] In some embodiments, the iteration module 330 is specifically configured as follows: Use the gradient descent algorithm to update the weight parameters of the preset neural network model, and obtain the loss function after each update of the preset neural network model; If the loss function reaches the preset value, the update operation is stopped.

[0043] In some embodiments, the loss function includes: a restoration loss function and a latent space loss function; the restoration loss function represents the restoration loss of the image input and output; the latent space loss function represents the difference between the latent space distribution and the standard normal distribution.

[0044] In some embodiments, an input-output module is also included; the input-output module is configured to: input the image to be repaired into the target neural network model, and output the target image.

[0045] The training device for the neural network model for CT image restoration provided in the embodiment of the present application can implement the various processes in the embodiments corresponding to the training method for the neural network model for CT image restoration mentioned above, and will not be described again here to avoid repetition.

[0046] It should be noted that the training device for the neural network model for CT image restoration provided in the embodiment of the present application and the training method for the neural network model for CT image restoration provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned training method for the neural network model for CT image restoration, and the repeated parts will not be repeated.

[0047] In some embodiments, see Figure 4 , Figure 4A structural schematic diagram of a neural network model for CT image restoration is provided in an embodiment of the present application; an embodiment of the present application provides a neural network model for CT image restoration, which is a variational autoencoder including an attention module, and the variational autoencoder is trained based on the training method of the neural network model for CT image restoration mentioned above.

[0048] In this embodiment, the original high-quality CT images can be labeled, 80% of which are used as a training set, 20% as a test set, and the original high-quality CT images as a label set. Figure 4 Build a neural network model and set hyperparameters, wherein the variational autoencoder architecture includes an attention module and is combined with a long short-term memory model LSTM to extract the features of medical CT images. The neural network model is trained using the training method of the CT image repair neural network model mentioned in the above embodiment. In each cycle, the variational encoder selectively reads part of the structure according to the characteristics of the medical CT image, and encodes this part of the structure into a latent space. The decoder samples from this latent space to obtain a latent vector to construct this part of the CT image data. After completing all the cycles set in the hyperparameters, the complete CT image data can be obtained. On this basis, the weight parameters of the neural network model are updated based on the gradient value corresponding to the loss function and iterative training is performed based on the reverse direction of the gradient value until the loss function corresponding to the image output by the preset neural network model reaches the preset value, and it is determined that the neural network model at this time meets the requirements. That is, the image restoration efficiency and image restoration quality of the neural network model meet expectations.

[0049] In some embodiments, see Figure 5 , Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. An electronic device 500 provided in an embodiment of the present application includes a processor 510 and a memory 520; the memory 520 stores a computer program, wherein the computer program implements the training method of the neural network model for CT image restoration when executed by the processor. Specifically, the processor 510 may include, for example, a general-purpose microprocessor, an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 510 may also include an onboard memory for cache purposes. The processor 510 may be a single processing unit or multiple processing units for executing different actions of the method flow according to the embodiment of the present application. The memory 520 may be any medium capable of containing, storing, conveying, propagating or transmitting instructions. For example, the memory 520 may include, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device, component or propagation medium. Specific examples of the memory 520 include: a magnetic storage device, such as a magnetic tape or a hard disk (HDD); an optical storage device, such as a compact disk (CD-ROM); a random access memory (RAM) or a flash memory; and / or a wired / wireless communication link. The present application also provides a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, the training method of the neural network model for CT image restoration is implemented. The computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently without being assembled into the device / apparatus / system. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present application is implemented. According to an embodiment of the present application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, optical cable, radio frequency signal, etc., or any suitable combination of the above. Those skilled in the art will appreciate that the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments and / or claims of the present application may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present application. Therefore, the scope of the present application should not be limited to the above-described embodiments, but should be determined not only by the appended claims, but also by the equivalents of the appended claims.

Claims

1. A training method for a neural network model for CT image restoration, It is characterized in that include: Get a sample image set; Using the sample image set to train a preset neural network model to obtain a gradient value corresponding to a loss function; Based on the reverse direction of the gradient value, the weight parameters of the preset neural network model are updated and iterative training is performed until the loss function corresponding to the image output by the preset neural network model reaches a preset value; The preset neural network model corresponding to when the loss function reaches a preset value is determined as the target neural network model.

2. The training method of the neural network model for CT image restoration according to claim 1, It is characterized in that The sample image set includes a training set and a label set, and the data in the training set is obtained by randomly labeling the data in the label set; and the using the sample image set to train the preset neural network model to obtain the gradient value corresponding to the loss function includes: Inputting the training set into the preset neural network model to obtain an output image; Determining a loss function based on the output image and the label set; The back propagation algorithm is used to determine the gradient value corresponding to the loss function.

3. The training method of the neural network model for CT image restoration according to claim 2, It is characterized in that The sample image set and the label set include labeled images in which preset areas of the images are labeled; and the using the sample image set to train a preset neural network model to obtain a gradient value corresponding to a loss function also includes: Inputting the marked image into the preset neural network model to obtain an output area image corresponding to a preset area in the marked image; A loss function is determined based on the output region image and a corresponding preset region in the label set.

4. The training method of the neural network model for CT image restoration according to claim 1, It is characterized in that The step of updating the weight parameters of the preset neural network model based on the reverse direction of the gradient value and performing iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value includes: Using a gradient descent algorithm to update the weight parameters of the preset neural network model, and obtaining the loss function of the preset neural network model after each update; If the loss function reaches a preset value, the updating operation is stopped.

5. The training method of the neural network model for CT image restoration according to claim 1, It is characterized in that The loss function includes: a restoration loss function and a latent space loss function; the restoration loss function represents the restoration loss of image input and output; the latent space loss function represents the difference between the latent space distribution and the standard normal distribution.

6. The training method of the neural network model for CT image restoration according to claim 1, It is characterized in that Also includes: The image to be restored is input into the target neural network model, and the target image is output.

7. A training device for a neural network model for CT image restoration, It is characterized in that include: Acquisition module, training module, iteration module and determination module, wherein, The acquisition module is configured to acquire a sample image set; The training module is configured to train a preset neural network model using the sample image set to obtain a gradient value corresponding to a loss function; The iteration module is configured to update the weight parameters of the preset neural network model based on the reverse direction of the gradient value and perform iterative training until the loss function corresponding to the image output by the preset neural network model reaches a preset value; The determination module is configured to determine the preset neural network model corresponding to when the loss function reaches a preset value as the target neural network model.

8. A neural network model for CT image restoration, It is characterized in that The neural network model is a variational autoencoder including an attention module, and the variational autoencoder is trained based on the training method of the neural network model for CT image restoration described in claims 1 to 6.

9. An electronic device comprising a processor and a memory; the memory stores a computer program, in, When the computer program is executed by the processor, the computer program implements the training method of the neural network model for CT image restoration according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images