A high-precision image generation method and system based on CT image layer thickness reconstruction

Abstract

The present invention provides a high-precision image generation method and system based on CT image layer thickness reconstruction. The method obtains each scan image of a preset scan layer thickness, and reconstructs each of the scan images to obtain a first scan image and a second scan image with different layer thicknesses; constructs a neural network model, the neural network model includes reconstruction networks of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are composed of a convolution layer, a grid module, a feature fusion module, and an upsampling output module; inputs the scan image of a training database into the neural network model for training, and under the action of a pre-constructed deep learning network loss, a target neural network model is obtained by training; inputs the scan image of a preset thickness into the target neural network model, and outputs the scan image of the target thickness, so as to obtain a high-precision thin-layer CT image, reduce storage requirements, facilitate file transmission, and reduce user workload.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-precision image generation based on CT image layer thickness reconstruction, and in particular relates to a high-precision image generation method and system based on CT image layer thickness reconstruction. Background Art

[0002] Computed tomography (CT) can be obtained by specifying a range of scanning slice thickness and reconstruction slice thickness (reconstruction slice thickness is not less than the scanning slice thickness) to obtain thin-slice CT (1.5 mm to 5 mm slice thickness) and thick-slice CT (5 mm slice thickness or more). Compared with thick-slice CT, thin-slice CT has higher spatial resolution. There is a growing demand for high-resolution (HR) thin-slice CT in clinical applications to identify early lung cancer with sub-centimeter ground-glass opacities, characterize pulmonary fibrosis, determine the relationship between lung lesions and their surrounding structures, and provide quantitative measurements.

[0003] Image resolution and slice thickness are inevitably balanced against the file size of medical images, which affects storage costs, the speed of file exchange over intranets or across institutions, and the human resources required for interpretation. Although raw data from modern scanners can reconstruct thin-slice CT images, most images in picture archiving and communication systems (PACS) are not thin-slice CT images, that is, they have low spatial resolution in the Y and Z planes (the two interslice planes). Summary of the invention

[0004] Based on this, an embodiment of the present invention provides a high-precision image generation method and system based on CT image layer thickness reconstruction, aiming to obtain high-precision thin-layer CT images, reduce storage requirements, facilitate file transfer, and reduce user workload.

[0005] A first aspect of an embodiment of the present invention provides a high-precision image generation method based on CT image layer thickness reconstruction, the method comprising:

[0006] Obtaining each scan image of a preset scan layer thickness, and reconstructing each of the scan images to obtain a corresponding first scan image and a second scan image matching the first scan image, so as to establish a training database, wherein the layer thickness in the second scan image is greater than the layer thickness in the first scan image;

[0007] Constructing a neural network model, the neural network model includes reconstruction networks of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are all composed of convolution layers, grid modules, feature fusion modules, and upsampling output modules, wherein the second scanned image is used as the input of the neural network model, and the matched first scanned image is outputted, specifically, the second scanned image is inputted into two convolution layers, and the outputs after the two convolution layers are sequentially connected to the four grid modules, and the outputs of the four grid modules are also connected to the feature fusion module, and the output of the feature fusion module and the other output after the two convolution layers are inputted into the upsampling output module through residual, and the output of the upsampling output module is the first scanned image;

[0008] Constructing a deep learning network loss, wherein the deep learning network loss is composed of one or more of pixel feature loss, image feature loss, adversarial generation loss, and k-space feature loss;

[0009] Inputting the scanned image of the training database into the neural network model for training, and obtaining the target neural network model through training under the effect of the deep learning network loss;

[0010] A scanned image of a preset thickness is input into the target neural network model, and a scanned image of a target thickness is output.

[0011] Furthermore, the second scan image can be represented by a 3D array matrix, wherein the three planes X, Y, and Z correspond to three array matrices, and the values ​​in the array matrices are all normalized results.

[0012] Furthermore, the grid module includes two alternately connected combination layers, wherein the combination layer consists of a convolutional layer and an activation layer.

[0013] Furthermore, in the feature fusion module, the four grid modules are respectively input into the sequentially connected convolutional layers and activation layers, wherein the outputs of the third grid module and the fourth grid module are superimposed through the first superposition module after passing through the sequentially connected convolutional layers and activation layers, the output of the first superposition module is superimposed with the result of the output of the second grid module passing through the sequentially connected convolutional layers and activation layers through the second superposition module after passing through the sequentially connected convolutional layers and activation layers, and the output of the second superposition module is superimposed with the result of the output of the first grid module passing through the sequentially connected convolutional layers and activation layers through the third superposition module and finally output.

[0014] Furthermore, the upsampling output module includes two interpolation layers and one convolution layer.

[0015] Furthermore, the scanned image of the training database is input into the neural network model for training. Under the action of the deep learning network loss, in the step of training to obtain the target neural network model, iterative updates are performed through the Adam optimizer with a learning rate of 0.001, wherein the network weights are updated for 200 cycles.

[0016] A second aspect of an embodiment of the present invention provides a high-precision image generation system based on CT image layer thickness reconstruction, the system comprising:

[0017] A reconstruction module, used to obtain each scan image with a preset scan layer thickness, and reconstruct each of the scan images to obtain a corresponding first scan image and a second scan image matching the first scan image, so as to establish a training database, wherein the layer thickness in the second scan image is greater than the layer thickness in the first scan image;

[0018] A first construction module is used to construct a neural network model, wherein the neural network model includes reconstruction networks of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are composed of convolutional layers, grid modules, feature fusion modules, and upsampling output modules, wherein the second scanned image is used as the input of the neural network model, and the matched first scanned image is outputted, specifically, the second scanned image is inputted into two convolutional layers, and the outputs after the two convolutional layers are sequentially connected to the four grid modules, and the outputs of the four grid modules are also connected to the feature fusion module, and the output of the feature fusion module and the other output after the two convolutional layers are inputted into the upsampling output module through residuals, and the output of the upsampling output module is the first scanned image;

[0019] A second construction module is used to construct a deep learning network loss, wherein the deep learning network loss is composed of one or more of pixel feature loss, image feature loss, adversarial generation loss, and k-space feature loss;

[0020] A training module, used for inputting the scanned images of the training database into the neural network model for training, and obtaining a target neural network model through training under the effect of the deep learning network loss;

[0021] An input module is used to input a scan image of a preset thickness into the target neural network model and output a scan image of a target thickness.

[0022] A third aspect of an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the high-precision image generation method based on CT image layer thickness reconstruction provided in the first aspect.

[0023] The fourth aspect of an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the high-precision image generation method based on CT image layer thickness reconstruction provided in the first aspect is implemented.

[0024] A high-precision image generation method and system based on CT image layer thickness reconstruction is provided in an embodiment of the present invention. The method obtains each scan image of a preset scan layer thickness, and reconstructs each of the scan images to obtain a first scan image and a second scan image with different layer thicknesses; constructs a neural network model, the neural network model includes a reconstruction network of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are composed of a convolutional layer, a grid module, a feature fusion module, and an upsampling output module; the scan images of the training database are input into the neural network model for training, and a target neural network model is obtained by training under the action of a pre-constructed deep learning network loss; the scan image of the preset thickness is input into the target neural network model, and the scan image of the target thickness is output to obtain a high-precision thin-layer CT image. Since the storage is mainly thick-layer CT images, the storage requirements can be reduced, file transfer is convenient, and the user workload is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flowchart of a method for generating a high-precision image based on CT image layer thickness reconstruction provided in the first embodiment of the present invention;

[0026] Figure 2 It is a schematic diagram of the neural network model;

[0027] Figure 3 is a schematic diagram of the grid module;

[0028] Figure 4 It is a schematic diagram of the feature fusion module;

[0029] Figure 5 A structural block diagram of a high-precision image generation system based on CT image layer thickness reconstruction provided in Embodiment 2 of the present invention;

[0030] Figure 6 This is a structural block diagram of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0031] 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.

[0032] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0033] 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.

[0034] Embodiment 1

[0035] See also Figure 1 , Figure 1 A flowchart of a high-precision image generation method based on CT image layer thickness reconstruction provided in the first embodiment of the present invention is shown, which specifically includes steps S01 to S05.

[0036] Step S01, obtaining each scan image with a preset scanning layer thickness, and reconstructing each of the scan images to obtain a corresponding first scan image and a second scan image matching the first scan image, so as to establish a training database, wherein the layer thickness in the second scan image is greater than the layer thickness in the first scan image.

[0037] Specifically, CT (Computed Tomography) is a technology that uses X-rays to scan the human body and reconstructs the image using a computer. During the CT scan, the detector captures X-ray attenuation data at different levels, which are then used to reconstruct the tomographic image.

[0038] In this embodiment, in order to obtain matching first and second scans, that is, thin-slice CT and thick-slice CT, specifically, the same patient and the same part are scanned once to obtain a CT scan with a lower scanning layer thickness, and reconstructed twice to obtain CT reconstructions with different layer thicknesses, thereby obtaining matching thin-slice CT and thick-slice CT. It should be noted that the scanning layer thickness shall not be greater than the reconstruction layer thickness, that is, the scanning layer thickness is 0.5mm, so the optional reconstruction layer thickness should be greater than 0.5mm, and the thin-slice CT (first scan) is usually between 1.5mm and 5mm thick, and the thick-slice CT (second scan) is usually between 8mm and 10mm.

[0039] In addition, reconstruction methods include multi-planar reconstruction (MPR) and interpolation algorithm reconstruction. Multi-planar reconstruction is the most basic image reconstruction method, which can process thin-layer images in the transverse section through a computer to generate images of any angle and orientation such as the coronal and sagittal planes. This method is suitable for observing structural imaging in any plane and observing normal tissues, organs or lesions at any angle; interpolation algorithm reconstruction is the process of image post-processing, in which interpolation algorithms (such as linear interpolation, cubic interpolation, etc.) can be used to generate images of different layer thicknesses. These algorithms interpolate the original data of adjacent layer thicknesses to generate images with higher or lower resolution.

[0040] It should be noted that the second scan image can be represented by a 3D array matrix, wherein the three planes X, Y, and Z correspond to three array matrices, and the values ​​in the array matrices are all normalized results. It can be understood that the scan image is a CT scan sequence, and illustratively, a CT scan sequence consists of 128 CT scan images with a length × width of 512 × 512, then the corresponding 3D array matrix is ​​a corresponding array of 128 × 512 × 512.

[0041] Step S02, constructing a neural network model, wherein the neural network model includes reconstruction networks of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are composed of a convolutional layer, a grid module, a feature fusion module, and an upsampling output module.

[0042] See also Figure 2 , is a schematic diagram of a neural network model, wherein pb1 to pb4 represent feature maps of different dimensions from low to high. Specifically, the second scanned image is used as the input of the neural network model, and the output is the matched first scanned image. Specifically, the second scanned image is input into two convolutional layers, and the output after the two convolutional layers is sequentially connected to four grid modules, and the outputs of the four grid modules are also connected to the feature fusion module. The output of the feature fusion module and the other output after the two convolutional layers are input into the upsampling output module through the residual, and the output of the upsampling output module is the first scanned image.

[0043] The grid module is an attention mechanism structure, see Figure 3 , is a schematic diagram of the grid module, where pb1 to pb4 represent feature maps of different dimensions from low to high. Specifically, and They are all composed of convolutional layers and activation layers. If the input of the combined layer is x, the output of the module is the corresponding H(x) or G(x). The expression is:

[0044] ;

[0045] ;

[0046] ;

[0047] ;

[0048] in, Expressed as Input, and Represented as a learnable weight coefficient, Expressed as The output, Expressed as passing The first lower branch output, Expressed as passing and The second lower branch output.

[0049] See also Figure 4 , which is a schematic diagram of the feature fusion module. In the feature fusion module, the four grid modules are respectively input into the sequentially connected convolutional layers and activation layers, wherein the outputs of the third grid module and the fourth grid module are superimposed through the first superposition module after passing through the sequentially connected convolutional layers and activation layers, the output of the first superposition module is superimposed with the result of the output of the second grid module passing through the sequentially connected convolutional layers and activation layers through the second superposition module after passing through the sequentially connected convolutional layers and activation layers, and the output of the second superposition module is superimposed with the result of the output of the first grid module passing through the sequentially connected convolutional layers and activation layers through the third superposition module, and finally output.

[0050] The upsampling output module consists of two interpolation layers and one convolution layer.

[0051] Step S03, constructing a deep learning network loss, wherein the deep learning network loss is composed of one or more of pixel feature loss, image feature loss, adversarial generation loss, and k-space feature loss.

[0052] Specifically, pixel features include L1loss, L2loss, and SmoothL1loss; image feature losses include SSIMloss, PSNR loss, and losses using public pre-trained network feature maps, including but not limited to neural networks trained based on CNN, Transformer, and Diffusion; adversarial generation losses are composed of trainable discriminator networks and GAN losses; the discriminator network can be composed of different VGG networks, or basic FCN networks, or U-net networks based on spectral normalization and batch normalization, or Transformer-based VIT segmentation networks; GAN losses can be basic GAN losses, WGAN losses, or RGAN losses.

[0053] Step S04: input the scanned image of the training database into the neural network model for training, and obtain the target neural network model through training under the action of the deep learning network loss.

[0054] Specifically, iterative updates are performed using the Adam optimizer with a learning rate of 0.001, where the network weights are updated for 200 cycles to obtain the target neural network model.

[0055] Step S05, inputting the scanned image of the preset thickness into the target neural network model, and outputting the scanned image of the target thickness.

[0056] Among them, the thicker layer scans are input into the target neural network model to obtain high-precision thin layer scans.

[0057] In summary, the high-precision image generation method based on CT image layer thickness reconstruction in the above-mentioned embodiments of the present invention obtains each scan image of a preset scan layer thickness, and reconstructs each of the scan images to obtain a first scan image and a second scan image with different layer thicknesses; constructs a neural network model, the neural network model includes a reconstruction network of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are all composed of a convolutional layer, a grid module, a feature fusion module, and an upsampling output module; the scan images of the training database are input into the neural network model for training, and the target neural network model is trained under the action of a pre-constructed deep learning network loss; the scan image of the preset thickness is input into the target neural network model, and the scan image of the target thickness is output to obtain a high-precision thin-layer CT image, reduce storage requirements, facilitate file transfer, and reduce user workload.

[0058] Embodiment 2

[0059] See also Figure 5 , Figure 51 is a structural block diagram of a high-precision image generation system based on CT image layer thickness reconstruction provided in Embodiment 2 of the present invention. The high-precision image generation system 200 based on CT image layer thickness reconstruction includes: a reconstruction module 21, a first construction module 22, a second construction module 23, a training module 24 and an input module 25, wherein:

[0060] A reconstruction module 21 is used to obtain each scan image of a preset scan layer thickness, and reconstruct each scan image to obtain a corresponding first scan image and a second scan image matching the first scan image, so as to establish a training database, wherein the layer thickness in the second scan image is greater than the layer thickness in the first scan image, and the second scan image can be represented by a 3D array matrix, wherein the three planes X, Y, and Z correspond to three array matrices, and the values ​​in the array matrices are all normalized results;

[0061] The first construction module 22 is used to construct a neural network model, which includes reconstruction networks of three planes X, Y, and Z. The reconstruction networks of the three planes X, Y, and Z are composed of convolutional layers, grid modules, feature fusion modules, and upsampling output modules. The second scanned image is used as the input of the neural network model, and the matched first scanned image is output. Specifically, the second scanned image is input into two convolutional layers, and the outputs after the two convolutional layers are connected to the four grid modules in sequence, and the outputs of the four grid modules are also connected to the feature fusion module. The output of the feature fusion module and the other output after the two convolutional layers are input into the upsampling output module through residuals. The output of the upsampling output module is the first scanned image. In addition, in the feature fusion module , the four grid modules are respectively input into the sequentially connected convolutional layers and activation layers, wherein the outputs of the third grid module and the fourth grid module are superimposed through the first superposition module after passing through the sequentially connected convolutional layers and activation layers, the output of the first superposition module is superimposed with the result of the output of the second grid module passing through the sequentially connected convolutional layers and activation layers through the second superposition module after passing through the sequentially connected convolutional layers and activation layers, the output of the second superposition module is superimposed with the result of the output of the first grid module passing through the sequentially connected convolutional layers and activation layers through the third superposition module, and finally output, the upsampling output module includes two interpolation layers and one convolutional layer, the grid module includes two alternately connected combination layers, wherein the combination layer is composed of a convolutional layer and an activation layer;

[0062] A second construction module 23 is used to construct a deep learning network loss, wherein the deep learning network loss is composed of one or more of pixel feature loss, image feature loss, adversarial generation loss, and k-space feature loss;

[0063] A training module 24 is used to input the scanned image of the training database into the neural network model for training. Under the action of the deep learning network loss, the target neural network model is trained and iteratively updated by an Adam optimizer with a learning rate of 0.001, wherein the network weight is updated for 200 cycles;

[0064] The input module 25 is used to input the scanned image of the preset thickness into the target neural network model and output the scanned image of the target thickness.

[0065] Embodiment 3

[0066] Another aspect of the present invention provides an electronic device, see Figure 6 , shown is an electronic device in Embodiment 3 of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, the high-precision image generation method based on CT image layer thickness reconstruction as described above is implemented.

[0067] In some embodiments, the processor 10 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 20, such as executing access restriction programs.

[0068] Among them, the memory 20 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 20 can be an internal storage unit of an electronic device, such as a hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of an electronic device, such as a plug-in hard disk equipped on the electronic device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Further, the memory 20 can also include both an internal storage unit of the electronic device and an external storage device. The memory 20 can not only be used to store application software and various types of data of the electronic device, but also can be used to temporarily store data that has been output or is to be output.

[0069] It should be pointed out that Figure 6 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than those shown in the figure, or combine certain components, or arrange the components differently.

[0070] The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the high-precision image generation method based on CT image layer thickness reconstruction as described above is implemented.

[0071] 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.

[0072] 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.

[0073] 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.

[0074] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0075] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A high-precision image generation method based on CT image layer thickness reconstruction, characterized in that: The method comprises: Obtaining each scan image of a preset scan layer thickness, and reconstructing each of the scan images to obtain a corresponding first scan image and a second scan image matching the first scan image, so as to establish a training database, wherein the layer thickness in the second scan image is greater than the layer thickness in the first scan image; Constructing a neural network model, the neural network model includes reconstruction networks of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are all composed of convolution layers, grid modules, feature fusion modules, and upsampling output modules, wherein the second scanned image is used as the input of the neural network model, and the matched first scanned image is outputted, specifically, the second scanned image is inputted into two convolution layers, and the outputs after the two convolution layers are sequentially connected to the four grid modules, and the outputs of the four grid modules are also connected to the feature fusion module, and the output of the feature fusion module and the other output after the two convolution layers are inputted into the upsampling output module through residual, and the output of the upsampling output module is the first scanned image; Constructing a deep learning network loss, wherein the deep learning network loss is composed of one or more of pixel feature loss, image feature loss, adversarial generation loss, and k-space feature loss; Inputting the scanned image of the training database into the neural network model for training, and obtaining the target neural network model through training under the effect of the deep learning network loss; Inputting a scanned image of a preset thickness into the target neural network model, and outputting a scanned image of a target thickness; The grid module includes two alternately connected combination layers, wherein the combination layer consists of a convolution layer and an activation layer; In the feature fusion module, the four grid modules are respectively input into the sequentially connected convolutional layers and activation layers, wherein the outputs of the third grid module and the fourth grid module are superimposed through the first superposition module after passing through the sequentially connected convolutional layers and activation layers, the output of the first superposition module is superimposed with the result of the output of the second grid module passing through the sequentially connected convolutional layers and activation layers through the second superposition module after passing through the sequentially connected convolutional layers and activation layers, and the output of the second superposition module is superimposed with the result of the output of the first grid module passing through the sequentially connected convolutional layers and activation layers through the third superposition module, and finally output; The upsampling output module includes two interpolation layers and one convolution layer.

2. The high-precision image generation method based on CT image layer thickness reconstruction according to claim 1, characterized in that: The second scan image can be represented by a 3D array matrix, wherein the three planes X, Y, and Z correspond to three array matrices, and the values ​​in the array matrices are all normalized results.

3. The high-precision image generation method based on CT image layer thickness reconstruction according to claim 1, characterized in that: The scanned image of the training database is input into the neural network model for training. Under the action of the deep learning network loss, in the step of training to obtain the target neural network model, iterative updates are performed through the Adam optimizer with a learning rate of 0.001, wherein the network weights are updated for 200 cycles.

4. A high-precision image generation system based on CT image layer thickness reconstruction, characterized in that: A method for generating a high-precision image based on CT image layer thickness reconstruction as claimed in any one of claims 1 to 3, the system comprising: A reconstruction module, used to obtain each scan image with a preset scan layer thickness, and reconstruct each of the scan images to obtain a corresponding first scan image and a second scan image matching the first scan image, so as to establish a training database, wherein the layer thickness in the second scan image is greater than the layer thickness in the first scan image; A first construction module is used to construct a neural network model, wherein the neural network model includes reconstruction networks of three planes X, Y, and Z, and the reconstruction networks of the three planes X, Y, and Z are composed of convolutional layers, grid modules, feature fusion modules, and upsampling output modules, wherein the second scanned image is used as the input of the neural network model, and the matched first scanned image is outputted, specifically, the second scanned image is inputted into two convolutional layers, and the outputs after the two convolutional layers are sequentially connected to the four grid modules, and the outputs of the four grid modules are also connected to the feature fusion module, and the output of the feature fusion module and the other output after the two convolutional layers are inputted into the upsampling output module through residuals, and the output of the upsampling output module is the first scanned image; A second construction module is used to construct a deep learning network loss, wherein the deep learning network loss is composed of one or more of pixel feature loss, image feature loss, adversarial generation loss, and k-space feature loss; A training module, used for inputting the scanned images of the training database into the neural network model for training, and obtaining a target neural network model through training under the effect of the deep learning network loss; An input module is used to input a scan image of a preset thickness into the target neural network model and output a scan image of a target thickness.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the high-precision image generation method based on CT image layer thickness reconstruction as described in any one of claims 1 to 3 is implemented.

6. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for generating a high-precision image based on CT image layer thickness reconstruction as described in any one of claims 1 to 3 is implemented.

Images