CN114511666A - A model generation method, image reconstruction method, apparatus, device and medium - Google Patents

Abstract

Embodiments of the present invention disclose a model generation method, an image reconstruction method, a device, a device and a medium, wherein the model generation method includes: acquiring a preset amount of breast imaging data; inputting the breast imaging data into a reconstruction algorithm based on a preset image In the established initial image reconstruction model, the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module. The regular noise reduction module realizes image feature extraction and image noise reduction based on a preset convolutional neural network; until the output When the difference between the reconstructed image and the expected reconstructed image satisfies the preset loss function, the final image reconstruction model is generated. The technical solution of the embodiment of the present invention realizes the use of the mature mathematical theory support of the iterative algorithm and the image optimization capability, and at the same time, the iterative parameters are determined and the iterative algorithm is optimized through the parameter optimization and feature extraction of the convolutional neural network, so that the three-dimensional breast image reconstruction is more accurate. Efficient.

Description

A model generation method, image reconstruction method, apparatus, device and medium

technical field

Embodiments of the present invention relate to the technical field of medical image processing, and in particular, to a model generation method, an image reconstruction method, an apparatus, a device, and a medium.

Background technique

At present, three-dimensional digital breast tomography reconstruction technology is the most commonly used medical method for breast cancer clinical screening. It effectively solves the problem of overlapping breast tissue in two-dimensional medical images, and has high imaging spatial resolution. Point sensitivity, low radiation dose and other advantages.

Most of the 3D digital breast tomography reconstruction methods are developed based on iterative algorithms. The 3D breast tomography reconstruction problem is modeled as a mathematical optimization model, and the purpose of reducing the noise and artifacts of the reconstructed image is achieved by solving this optimization problem. Or, use the method of convolutional neural network to reconstruct 3D digital breast tomography, directly build a neural network to replace the 3D breast reconstruction process, the input is projection data, and the output is a reconstructed image.

However, using iterative algorithm for image reconstruction, the reconstruction speed is slow and the reconstruction time is long, and real-time reconstruction cannot be achieved; the reconstruction method based on convolutional neural network has fast reconstruction speed and strong parameter optimization ability, but lacks mathematical rigor and interpretability. Dependency on data is high.

SUMMARY OF THE INVENTION

Embodiments of the present invention provide a model generation method, an image reconstruction method, an apparatus, a device, and a medium, so as to improve the reconstruction speed of a three-dimensional digital breast tomography image and improve the quality of the reconstructed image.

In a first aspect, an embodiment of the present invention provides a model generation method, which includes:

Acquire a preset amount of breast imaging data;

Inputting the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module, the regular noise reduction module The module realizes image feature extraction and image noise reduction based on preset convolutional neural network;

The model parameters are updated in the process of image reconstruction calculation performed by the initial image reconstruction model, and the final image reconstruction model is generated until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function.

Optionally, simulating a preset amount of breast imaging data according to the geometric features of the breast imaging system includes:

Within a preset projection angle range, one piece of projection data is collected at every preset angle to obtain the breast imaging data.

Optionally, the preset image reconstruction algorithm includes an alternating direction multiplier algorithm.

Optionally, the image reconstruction module is specifically configured to iteratively calculate the value of the previous iterative calculation of the current image reconstruction iterative calculation in the current image reconstruction iterative calculation, so as to correct the difference between the output result and the orthographic breast imaging data.

Optionally, the preset convolutional neural network is a four-layer convolutional operation network, and the convolution kernel of the preset convolutional neural network is a three-dimensional convolution kernel.

In a second aspect, an embodiment of the present invention further provides a three-dimensional breast image reconstruction method, the method comprising:

Acquire the breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to the model generation method in any model generation embodiment;

Inputting the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed.

In a third aspect, an embodiment of the present invention further provides a model generation device, the device comprising:

A sample data simulation module for acquiring a preset amount of breast imaging data;

A model correction module for inputting the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module , the regular noise reduction module realizes image feature extraction and image noise reduction based on a preset convolutional neural network;

The model determination module is used to update the model parameters in the process of image reconstruction calculation performed by the initial image reconstruction model, until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function, and generate the final image reconstruction model .

Optionally, the sample data simulation module is specifically used for:

Within a preset projection angle range, one piece of projection data is collected at every preset angle to obtain the breast imaging data.

Optionally, the preset image reconstruction algorithm includes an alternating direction multiplier algorithm.

Optionally, the image reconstruction module is specifically configured to iteratively calculate the value of the previous iterative calculation of the current image reconstruction iterative calculation in the current image reconstruction iterative calculation, so as to correct the difference between the output result and the orthographic breast imaging data.

Optionally, the preset convolutional neural network is a four-layer convolutional operation network, and the convolution kernel of the preset convolutional neural network is a three-dimensional convolution kernel.

In a fourth aspect, an embodiment of the present invention further provides a three-dimensional breast image reconstruction device, the device comprising:

a data and model acquisition module, configured to acquire breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to the model generation method in any of the model generation embodiments;

The image reconstruction module is configured to input the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed.

In a fifth aspect, an embodiment of the present invention further provides a computer device, the computer device comprising:

one or more processors;

a storage device for storing one or more programs;

When the one or more programs are executed by the one or more processors, the one or more processors implement the model generation method or the three-dimensional breast image reconstruction method described in any one of the embodiments of the present invention.

In a sixth aspect, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the model generation method or the three-dimensional breast cancer described in any of the embodiments of the present invention Image reconstruction method.

In the embodiment of the present invention, the simulated breast imaging data is input into the initial image reconstruction model established based on the preset image reconstruction algorithm, and then the model parameters are updated in the process of image reconstruction calculation performed by the initial image reconstruction model, until the output reconstructed image When the difference from the expected reconstructed image satisfies the preset loss function, the final image reconstruction model is generated. It solves the problem that only the iterative algorithm is used for image reconstruction, the reconstruction speed is slow, the reconstruction time is long, and real-time reconstruction cannot be achieved, and the reconstruction method based on convolutional neural network lacks mathematical rigor and interpretability. In the process of image reconstruction, the image reconstruction module, the regular noise reduction module and the operator update module in the initial image reconstruction model are used to reconstruct the image. Moreover, the regular noise reduction module realizes image feature extraction based on the preset convolutional neural network. Implement image noise reduction. While realizing the mature mathematical theory support and image optimization capability of iterative algorithm, the iterative parameters are determined and the iterative algorithm is optimized through the parameter optimization and feature extraction of the convolutional neural network, so that the reconstruction of 3D breast images is more efficient.

Description of drawings

1 is a flowchart of a model generation method in Embodiment 1 of the present invention;

2 is a schematic diagram of a breast imaging data acquisition system in Embodiment 1 of the present invention;

3 is a structural diagram of a deep neural network based on the alternating direction multiplier algorithm in Embodiment 1 of the present invention;

4 is a network structure diagram of three sub-modules of a deep neural network based on the alternating direction multiplier algorithm in Embodiment 1 of the present invention;

5 is a flowchart of a three-dimensional breast image reconstruction method in Embodiment 2 of the present invention;

6 is a schematic structural diagram of a model generation device in Embodiment 3 of the present invention;

7 is a schematic structural diagram of a three-dimensional breast image reconstruction device in Embodiment 4 of the present invention;

FIG. 8 is a schematic structural diagram of a computer device in Embodiment 5 of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the following will refer to the accompanying drawings in the embodiments of the present invention, and describe the technical solutions of the present invention clearly and completely through the implementation manner. Obviously, the described embodiments are the present invention. Some examples, but not all examples. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. In the following embodiments, optional features and examples are provided in each embodiment at the same time, and the various features described in the embodiments can be combined to form multiple optional solutions, and each numbered embodiment should not be regarded as only for a technical solution.

Example 1

FIG. 1 is a flowchart of a model generation method provided by Embodiment 1 of the present invention. This embodiment is applicable to the case of generating an image reconstruction model. The method can be implemented by a model generation device. The device is configured in a computer device. software and/or hardware in the device.

As shown in Figure 1, the model generation method specifically includes:

S110. Acquire a preset amount of breast imaging data.

Since the size and noise distribution of the simulation data can be artificially determined, the diversity of data samples is greatly increased, the simulation data has no artifacts caused by experimental system errors, and the image resolution is higher, which is more suitable for scientific research. Therefore, in this embodiment, simulated breast imaging data is taken as model training data. In the process of acquiring training data, in order to be closer to the clinic, this patent simulates the geometry of a real breast imaging system. Please refer to the geometric schematic diagram of the data acquisition system shown in FIG. 2 . In FIG. 2 , the target object (Object) of X-ray scanning is a simulated mammary gland membrane compressed by a compression board. During the sampling process, the distance between the X-ray tube and the detection center position of the detector (the black origin) remains unchanged. Among them, the left image in Figure 2 is the front view (Front View) of the data acquisition system, and the right image in Figure 2 is the side view (Side View) of the data acquisition system

In a specific embodiment, the pixel size of the reconstructed 3D breast image may be set to 1024x512x30. The X-ray projection angle is 40 ^° , the number of projection angles is 21 (one breast projection image is collected every 2 degrees), and the number of projection pixels is 1024x512x21. MATLAB software can be used to simulate a preset number (eg 3000) breast imaging data as training samples.

Within a preset projection angle range, one piece of projection data is collected at every preset angle to obtain the breast imaging data.

S120. Input the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, where the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module, and an operator update module, and the regular The noise reduction module implements image feature extraction and image noise reduction based on a preset convolutional neural network.

The preset image reconstruction algorithm may be an iterative algorithm, such as an algebraic reconstruction algorithm or an alternating direction method of multipliers (the alternating direction method of multipliers, ADMM). In the process of establishing the initial image reconstruction model, according to the model characteristics of different algorithms, the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module, and each module is solved to determine the final global parameters. Thus, the final image reconstruction model is obtained. The image reconstruction module is specifically configured to iteratively calculate the value of the previous iterative calculation of the current image reconstruction iterative calculation in the current image reconstruction iterative calculation, so as to correct the difference between the output result and the orthographic breast imaging data. Regular noise reduction modules are generally used to prevent overfitting in the reconstruction process and to optimize image quality. The operator update module is used to update parameter values during each iteration. That is, the alternating direction multiplier algorithm is decomposed into three subproblems.

In a specific example, the preset image reconstruction algorithm is an alternating direction multiplier algorithm. For the network structure of the initial image reconstruction model established based on the alternating direction multiplier algorithm, please refer to the deep neural network structure diagram based on the alternating direction multiplier algorithm shown in FIG. 3 . Among them, the breast tissue projection image is the 3D breast imaging data obtained by simulation, the number of pixels of the data is 1024*512*21, and the reconstructed image of breast tissue is the reconstructed image output by the initial image reconstruction model, and the number of pixels is 512*1024*30 , Iter1-IterN is the number of iterations, indicating the iterative content of the corresponding number of iterations. Exemplarily, in the nth iteration (Iter n), the inputs are x ^n-1 , z ^n-1 , λ ^n-1 and b.

Specifically, the X module corresponds to the image reconstruction module, and the network architecture of the X module is shown in Figure 4(a). Among them, b is the orthographic projection of the three-dimensional breast, λ ^n-1 , z ^n-1 are the initial values corresponding to λ and z, and the superscripts n and k are the number of iterations; by combining λ ^n-1 and z ^{n -1} , for x ^n-1 performs k iterations, and constantly repairs the difference between x ^n-1 and b, thereby reducing the error with the original image. Its mathematical formula is: X ⁿ : x ^n,k =μ ₁ x ^n,k-1 +μ ₂ (z ^n-1 -λ ^n-1 )-τ(A ^T Ax ^n,k-1 -A ^T b ), where x is the breast imaging data that needs to be reconstructed, parameters μ ₁ , μ ₂ , τ are the iterative step size, A is the system matrix, ^AT is the transpose of the system matrix, and the multiplication with A and ^AT is equivalent to the image Do projection and backprojection.

其中，

为迭代步长，n为迭代次数。The Z module corresponds to the regular noise reduction module, and the network architecture of the Z module is shown in Figure 4(b). Typically used to prevent overfitting during reconstruction and to optimize image quality. This module introduces a convolutional neural network, and inputs the input z ^n-1 into a 4-layer convolutional neural network, and the corresponding network layers are 1, 16, 16, and 1, respectively. Since breast tissue has information distribution in three spatial dimensions, a 3-dimensional convolution kernel is used in this embodiment, and the size of the convolution kernel is 3x3x3, and the important features of the image are extracted through the convolutional neural network. R(z) is a nonlinear transformation. Its mathematical expression is: Z ⁿ :z ⁿ =R(z)+αS(z); R(z)=μ ₃ z ^n-1 +μ ₄ (x ⁿ +λ ^n-1 ), where z can be It is understood as the denoised data of x. Under the premise that x=z, μ ₃ and μ ₄ are the iterative step size, S(z) is the convolution transformation, the superscript n is the number of iterations, and α is the iterative step size. λ corresponds to the operator update module, and the network architecture of the λ module is shown in Figure 4(c). Specifically, the Laplacian operator is updated, and the parameter values are updated in each iteration process. The mathematical formula is:

in,

is the iteration step size, and n is the number of iterations.

S130: Update model parameters during the image reconstruction calculation process performed by the initial image reconstruction model, until the difference between the output reconstructed image and the expected reconstructed image satisfies a preset loss function, and generate a final image reconstruction model.

Specifically, iterative calculation is continuously performed in the above-mentioned neural network, and the X, Z, and Λ(λ) module data are updated in turn, until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function, and the final output is generated. The image reconstructs the model, thus ending the model generation process.

In a preferred embodiment, the mammary gland reconstruction algorithm (analytic back-projection process) accelerated by CUDA (Compute Unified Device Architecture) of GPU is directly embedded in the neural network, which can quickly realize the projection domain in the process of network training And the parallel transformation of the back-projection domain to realize the process of image reconstruction, which can improve the speed of image reconstruction.

It should be noted here that the number of pixels input and output of the model, the number of samples, and the size of the convolution kernel in the neural network can all be set according to the actual model generation requirements. It is not limited.

The technical solution of this embodiment is to input the simulated breast imaging data into the initial image reconstruction model established based on the preset image reconstruction algorithm, and then update the model parameters in the process of image reconstruction calculation performed by the initial image reconstruction model, until the output When the difference between the reconstructed image and the expected reconstructed image satisfies the preset loss function, the final image reconstruction model is generated. It solves the problem that only the iterative algorithm is used for image reconstruction, the reconstruction speed is slow, the reconstruction time is long, and real-time reconstruction cannot be achieved, and the reconstruction method based on convolutional neural network lacks mathematical rigor and interpretability. In the process of image reconstruction, the image reconstruction module, the regular noise reduction module and the operator update module in the initial image reconstruction model are used to reconstruct the image. Moreover, the regular noise reduction module realizes image feature extraction based on the preset convolutional neural network. Implement image noise reduction. While realizing the mature mathematical theory support and image optimization capability of iterative algorithm, the iterative parameters are determined and the iterative algorithm is optimized through the parameter optimization and feature extraction of the convolutional neural network, so that the reconstruction of 3D breast images is more efficient.

Embodiment 2

5 is a flowchart of a method for reconstructing a 3D breast image according to Embodiment 2 of the present invention. The embodiment of the present invention can be applied to the situation of reconstructing a 3D breast image. The method can be implemented by a 3D breast image reconstruction device, which is configured in In a computer device, the specific implementation may be implemented by software and/or hardware in the device.

As shown in Figure 5, the three-dimensional breast image reconstruction method specifically includes:

S210: Acquire the breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to the model generation method in any model generation embodiment.

Specifically, the breast imaging data to be reconstructed may be computed tomography scan data obtained from a clinical examination of a patient, or imaging data obtained through simulation by a breast imaging data simulation system. The format of the breast imaging data to be reconstructed is the same as that of the input data of the three-dimensional breast image reconstruction model.

The three-dimensional breast image reconstruction model is an image reconstruction model generated according to the model generation method in any model generation embodiment of the present invention. The 3D breast image reconstruction model is used to directly replace the entire iterative reconstruction process, enabling end-to-end reconstruction. Taking the ADMM iterative algorithm as an example, the large-scale problem of 3D breast tomographic reconstruction is divided into three easy-to-solve sub-problems, X ⁿ , Z ⁿ , and Λ ⁿ . Among them, the X module corresponds to the image reconstruction function, the Z module corresponds to the regular term noise reduction function, and the Λ corresponds to the Laplace operator update, and the parameter values are updated in each iteration process.

S220. Input the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed.

Input the collected breast phantom projection data or patient breast data into the three-dimensional breast image reconstruction model to obtain the final three-dimensional breast reconstruction image.

Preferably, by using the GPU to perform the reconstruction process in this embodiment, the speed of image reconstruction can be further improved, so that a three-dimensional breast image with low noise, few artifacts and high image resolution can be obtained more quickly.

In the technical solution of this embodiment, a reconstructed 3D breast image is obtained by inputting the image data to be reconstructed into a pre-trained 3D breast image reconstruction model, and the 3D breast image reconstruction model uses a deep convolutional neural network to automatically Learning and updating the parameters of the iterative process solves the tedious parameter selection process in the traditional iterative algorithm, improves the resolution of the three-dimensional breast image, has a better effect of removing artifacts, and can improve the quality of the reconstructed three-dimensional breast image.

Embodiment 3

6 is a schematic structural diagram of an apparatus for generating a model according to Embodiment 3 of the present invention. This embodiment is applicable to the case of generating an image reconstruction model. The apparatus may be configured in a computer device, and may be configured by software and/or hardware in the device. to implement.

As shown in FIG. 6 , the model generation apparatus in the embodiment of the present invention includes: a sample data simulation module 310 , a model correction module 320 and a model determination module 330 .

The sample data simulation module 310 is used to obtain a preset amount of breast imaging data; the model correction module 320 is used to input the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein, The initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module, and the regular noise reduction module implements image feature extraction and image noise reduction based on a preset convolutional neural network; the model determination module 330 uses In the process of image reconstruction calculation performed by the initial image reconstruction model, the model parameters are updated until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function, and then the final image reconstruction model is generated.

The technical solution of this embodiment is to input the simulated breast imaging data into the initial image reconstruction model established based on the preset image reconstruction algorithm, and then update the model parameters in the process of image reconstruction calculation performed by the initial image reconstruction model, until the output When the difference between the reconstructed image and the expected reconstructed image satisfies the preset loss function, the final image reconstruction model is generated. It solves the problem that only the iterative algorithm is used for image reconstruction, the reconstruction speed is slow, the reconstruction time is long, and real-time reconstruction cannot be achieved, and the reconstruction method based on convolutional neural network lacks mathematical rigor and interpretability. In the process of image reconstruction, the image reconstruction module, the regular noise reduction module and the operator update module in the initial image reconstruction model are used to reconstruct the image. Moreover, the regular noise reduction module realizes image feature extraction based on the preset convolutional neural network. Implement image noise reduction. While realizing the mature mathematical theory support and image optimization capability of iterative algorithm, the iterative parameters are determined and the iterative algorithm is optimized through the parameter optimization and feature extraction of the convolutional neural network, so that the reconstruction of 3D breast images is more efficient.

Optionally, the sample data simulation module 310 is specifically used for:

Within a preset projection angle range, one piece of projection data is collected at every preset angle to obtain the breast imaging data.

Optionally, the preset image reconstruction algorithm includes an alternating direction multiplier algorithm.

Optionally, the image reconstruction module is specifically configured to iteratively calculate the value of the previous iterative calculation of the current image reconstruction iterative calculation in the current image reconstruction iterative calculation, so as to correct the difference between the output result and the orthographic breast imaging data.

Optionally, the preset convolutional neural network is a four-layer convolutional operation network, and the convolution kernel of the preset convolutional neural network is a three-dimensional convolution kernel. The model generation apparatus provided by the embodiment of the present invention can execute the model generation method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

Embodiment 4

FIG. 7 is a three-dimensional breast image reconstruction device provided in Embodiment 4 of the present invention, which can be applied to the situation of reconstructing three-dimensional breast images. The device can be implemented by a three-dimensional breast image reconstruction device. The device is configured in a computer device. software and/or hardware in the device.

As shown in FIG. 7 , the apparatus includes: a data and model acquisition module 410 and an image reconstruction module 420 .

Among them, the data and model acquisition module 410 is used to acquire the breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to the model generation method in any model generation embodiment; the image reconstruction module 420 is used to The reconstructed breast imaging data is input into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed.

In the technical solution of this embodiment, a reconstructed 3D breast image is obtained by inputting the image data to be reconstructed into a pre-trained 3D breast image reconstruction model, and the 3D breast image reconstruction model uses a deep convolutional neural network to automatically Learning and updating the parameters of the iterative process solves the tedious parameter selection process in the traditional iterative algorithm, improves the resolution of the three-dimensional breast image, has a better effect of removing artifacts, and can improve the quality of the reconstructed three-dimensional breast image.

The three-dimensional breast image reconstruction apparatus provided by the embodiment of the present invention can execute the three-dimensional breast image reconstruction method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

Embodiment 5

8 is a schematic structural diagram of a computer device in Embodiment 3 of the present invention. The computer device is connected to an imaging device (eg, a Mammo breast imaging device) for controlling the imaging device and receiving signals collected by the imaging device. Figure 8 shows a block diagram of an exemplary computer device 812 suitable for use in implementing embodiments of the present invention. The computer device 812 shown in FIG. 8 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present invention.

As shown in FIG. 8, computer device 812 takes the form of a general-purpose computing device. Components of computer device 812 may include, but are not limited to, one or more processors or processing units 814, system memory 828, and a bus 818 connecting various system components including system memory 828 and processing unit 814.

Bus 818 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of a variety of bus structures. By way of example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect ( PCI) bus.

Computer device 812 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 812, including both volatile and nonvolatile media, removable and non-removable media.

System memory 828 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 830 and/or cache memory 832 . Computer device 812 may further include other removable/non-removable, volatile/non-volatile computer system storage media. For example only, storage system 834 may be used to read and write to non-removable, non-volatile magnetic media (not shown in FIG. 8, commonly referred to as a "hard drive"). Although not shown in Figure 8, a disk drive may be provided for reading and writing to removable non-volatile magnetic disks (eg "floppy disks"), as well as removable non-volatile optical disks (eg CD-ROM, DVD-ROM) or other optical media) to read and write optical drives. In these cases, each drive may be connected to bus 818 through one or more data media interfaces. Memory 828 may include at least one program product having a set (eg, at least one) of program modules configured to perform the functions of various embodiments of the present invention.

A program/utility 840 having a set (at least one) of program modules 842, which may be stored, for example, in memory 828, such program modules 842 including, but not limited to, an operating system, one or more application programs, other program modules, and program data , each or some combination of these examples may include an implementation of a network environment. Program modules 842 generally perform the functions and/or methods of the described embodiments of the present invention.

Computer device 812 may also communicate with one or more external devices 814 (eg, keyboard, pointing device, display 824, etc.), may also communicate with one or more devices that enable a user to interact with the computer device 812, and/or communicate with Any device (eg, network card, modem, etc.) that enables the computer device 812 to communicate with one or more other computing devices. Such communication may take place through input/output (I/O) interface 822 . Also, the computer device 812 may communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN), and/or a public network such as the Internet) through a network adapter 820 . As shown, network adapter 820 communicates with other modules of computer device 812 via bus 818 . It should be appreciated that, although not shown in FIG. 8, other hardware and/or software modules may be used in conjunction with computer device 812, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tapes drives and data backup storage systems, etc.

The processing unit 816 executes various functional applications and data processing by running the programs stored in the system memory 828, for example, implementing the model generation method or the three-dimensional breast image reconstruction method provided by the embodiments of the present invention.

Among them, the model generation method includes:

Acquire a preset amount of breast imaging data;

Inputting the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module, the regular noise reduction module The module realizes image feature extraction and image noise reduction based on preset convolutional neural network;

The model parameters are updated in the process of image reconstruction calculation performed by the initial image reconstruction model, and the final image reconstruction model is generated until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function.

Three-dimensional breast image reconstruction methods include:

Acquire the breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to any of the above methods;

Inputting the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed.

Embodiment 6

Embodiment 6 of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the model generation method or the three-dimensional breast image reconstruction method provided by the embodiment of the present invention.

Among them, the model generation method includes:

Acquire a preset amount of breast imaging data;

Inputting the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module, the regular noise reduction module The module realizes image feature extraction and image noise reduction based on preset convolutional neural network;

The model parameters are updated in the process of image reconstruction calculation performed by the initial image reconstruction model, and the final image reconstruction model is generated until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function.

Three-dimensional breast image reconstruction methods include:

Acquire the breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to any of the above methods;

Inputting the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed.

The computer storage medium in the embodiments of the present invention may adopt any combination of one or more computer-readable mediums. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples (a non-exhaustive list) of computer readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), Erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

A computer-readable signal medium may include a propagated data signal in baseband or as part of a carrier wave, with computer-readable program code embodied thereon. 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 foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device .

Program code embodied on a computer readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations of the present invention may be written in one or more programming languages, including object-oriented programming languages such as Java, Smalltalk, C++, and also conventional procedures, or a combination thereof programming languages such as "C" or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or wide area network (WAN), or may be connected to an external computer (eg, through the Internet using an Internet service provider) connect).

Note that the above are only preferred embodiments of the present invention and applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention. The scope is determined by the scope of the appended claims.

Claims (10)

1. a model generation method, is characterized in that, comprises: Acquire a preset amount of breast imaging data; Inputting the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module, the regular noise reduction module The module realizes image feature extraction and image noise reduction based on preset convolutional neural network; The model parameters are updated in the process of image reconstruction calculation performed by the initial image reconstruction model, and the final image reconstruction model is generated until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function. 2 . The method according to claim 1 , wherein the simulating a preset number of breast imaging data according to the geometric features of the breast imaging system comprises: 2 . Within a preset projection angle range, one piece of projection data is collected at every preset angle to obtain the breast imaging data. 3. The method of claim 1, wherein the preset image reconstruction algorithm comprises an alternating direction multiplier algorithm. 4. The method according to any one of claims 1-3, characterized in that, the image reconstruction module is specifically configured to, in the iterative calculation of current image reconstruction, perform a calculation on the numerical value of the previous iterative calculation of the current image reconstruction iterative calculation. Iterative calculations to correct for discrepancies between the output and orthographic mammography data. 5 . The method according to claim 4 , wherein the preset convolutional neural network is a four-layer convolutional operation network, and the convolution kernel of the preset convolutional neural network is a three-dimensional convolution kernel. 6 . 6. A three-dimensional breast image reconstruction method, characterized in that, comprising: Acquire the breast imaging data to be reconstructed, and generate a three-dimensional breast image reconstruction model according to any one of claims 1-5; Inputting the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed. 7. A model generation device, characterized in that, comprising: A sample data simulation module for acquiring a preset amount of breast imaging data; A model correction module for inputting the breast imaging data into an initial image reconstruction model established based on a preset image reconstruction algorithm, wherein the initial image reconstruction model includes an image reconstruction module, a regular noise reduction module and an operator update module , the regular noise reduction module realizes image feature extraction and image noise reduction based on a preset convolutional neural network; The model determination module is used to update the model parameters in the process of image reconstruction calculation performed by the initial image reconstruction model, until the difference between the output reconstructed image and the expected reconstructed image satisfies the preset loss function, and generate the final image reconstruction model . 8. A three-dimensional breast image reconstruction device, comprising: a data and model acquisition module for acquiring breast imaging data to be reconstructed, and generating a three-dimensional breast image reconstruction model according to any one of claims 1-5; The image reconstruction module is configured to input the breast imaging data to be reconstructed into the three-dimensional breast image reconstruction model to obtain a three-dimensional breast reconstruction image corresponding to the breast imaging data to be reconstructed. 9. A computer device, characterized in that the computer device comprises: one or more processors; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the model generation method as claimed in any one of claims 1 to 5 or as claimed in claim 6 3D breast image reconstruction method. 10. A computer-readable storage medium on which a computer program is stored, wherein when the program is executed by a processor, the model generation method according to any one of claims 1-5 or the method according to claim 6 is realized when the program is executed 3D breast image reconstruction method.

Images