WO2021135773A1 - 图像重建方法、装置、设备、系统及计算机可读存储介质 - Google Patents

图像重建方法、装置、设备、系统及计算机可读存储介质 Download PDF

Info

Publication number
WO2021135773A1
WO2021135773A1 PCT/CN2020/132371 CN2020132371W WO2021135773A1 WO 2021135773 A1 WO2021135773 A1 WO 2021135773A1 CN 2020132371 W CN2020132371 W CN 2020132371W WO 2021135773 A1 WO2021135773 A1 WO 2021135773A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
network structure
data set
image reconstruction
reconstructed image
Prior art date
Application number
PCT/CN2020/132371
Other languages
English (en)
French (fr)
Inventor
程冉
韦增培
肖鹏
谢庆国
Original Assignee
苏州瑞派宁科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 苏州瑞派宁科技有限公司 filed Critical 苏州瑞派宁科技有限公司
Publication of WO2021135773A1 publication Critical patent/WO2021135773A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/002D [Two Dimensional] image generation
    • G06T11/003Reconstruction from projections, e.g. tomography
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/70Denoising; Smoothing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10072Tomographic images
    • G06T2207/10081Computed x-ray tomography [CT]

Definitions

  • This application relates to the field of image processing technology, and in particular to an image reconstruction method, device, equipment, system, and computer-readable storage medium.
  • CT Computed Tomography
  • PET Positron Emission Tomography
  • other medical equipment it is usually necessary to reconstruct the scan data.
  • the purpose of the embodiments of the present application is to provide an image reconstruction method, device, equipment, system, and computer-readable storage medium to solve at least one problem in the prior art.
  • an embodiment of the present application provides an image reconstruction method, the image reconstruction method includes:
  • the target network model includes a first network structure with fixed network parameters and a second network structure with multiple convolutional layers
  • the data set to be measured includes images of target objects collected under a plurality of different projection angles.
  • Denoising processing is performed on the first reconstructed image by using the trained second network structure to obtain a second reconstructed image.
  • the step of training the constructed target network model by using the obtained measured sample image includes:
  • Radon transform processing is performed on the acquired measured sample image to obtain a corresponding sample data set.
  • the sample data set includes projection data corresponding to the measured sample image at a plurality of different specific angles, and The sample data set matches the data set to be tested;
  • the step of using the trained second network structure to perform denoising processing on the first reconstructed image to obtain a second reconstructed image includes:
  • the second convolutional layer in the second network structure is used to process the shallow layer characteristic information and the deep layer characteristic information to obtain a second sample image.
  • the preset algorithm includes a direct back projection method, a filtered back projection method, or a Fourier direct transform method.
  • the projection data includes CT image data, PET image data, or PET/CT image data.
  • the embodiment of the present application also provides an image reconstruction device on which a target network model is constructed, and the image reconstruction device includes:
  • the training unit is configured to train the target network model by using the obtained measured sample images, wherein the target network model includes a first network structure with fixed network parameters and a first network structure containing multiple convolutional layers Two network structure,
  • An image reconstruction unit configured to use the first network structure to perform image reconstruction processing on the acquired data set to be measured according to a preset algorithm to obtain a first reconstructed image, the data set to be measured is included in a plurality of different projections Projection data of the target object collected at different angles;
  • a denoising unit configured to perform denoising processing on the first reconstructed image by using the trained second network structure to obtain a second reconstructed image.
  • the training unit is specifically configured as:
  • Radon transform processing is performed on the acquired measured sample image to obtain a corresponding sample data set, and the sample data set includes projection data corresponding to the measured sample image under multiple different projection angles, and The sample data set matches the data set to be tested;
  • the second network structure includes a first convolutional layer, a residual module, and a second convolutional layer that are sequentially connected.
  • the embodiment of the present application also provides a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the above-mentioned image reconstruction method.
  • An embodiment of the present application also provides an image processing system, which includes the above-mentioned computer device and a detection device, wherein the detection device is configured to obtain projection data by scanning a target object and convert the obtained projection The data is provided to the computer equipment.
  • the detection device includes a CT scanner, a PET detector, or a PET/CT device.
  • the embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program can realize the functions corresponding to the above-mentioned image reconstruction method when the computer program is executed.
  • the embodiments of the application use the first network structure in the target network model to perform image reconstruction processing on the projection data set according to a preset algorithm to obtain the first reconstructed image, and then use the target
  • the second network structure in the network model performs denoising processing on the first reconstructed image to obtain a second reconstructed image, which can improve the quality of the reconstructed image.
  • the data processing speed can be improved.
  • Fig. 1 is an application environment diagram of an image reconstruction method in an embodiment of the present application
  • Fig. 2 is a schematic structural diagram of a target network model used in an embodiment of the present application
  • Figure 3 is a schematic diagram of the structure of the sub-modules in the residual module in the target network model
  • FIG. 4 is a schematic flowchart of an image reconstruction method provided by an embodiment of the present application.
  • Figure 5 is a schematic diagram of the acquired data set to be tested
  • Fig. 6 is a schematic structural diagram of an image reconstruction device provided by an embodiment of the present application.
  • FIG. 7 is a schematic diagram of the structure of a computer device in an embodiment of the present application.
  • Fig. 8 is a schematic structural diagram of a computer device in another embodiment of the present application.
  • Fig. 9 is a schematic structural diagram of an image processing system in an embodiment of the present application.
  • connection/connection refers to the presence or addition of features, steps or elements, but does not exclude the presence or addition of one or more other features, steps or elements.
  • connecting/connection refers to the presence or addition of features, steps or elements, but does not exclude the presence or addition of one or more other features, steps or elements.
  • connecting/connection refers to the presence or addition of features, steps or elements, but does not exclude the presence or addition of one or more other features, steps or elements.
  • and/or as used herein includes any and all combinations of one or more of the associated listed items.
  • Fig. 1 is an application environment diagram of an image reconstruction method in an embodiment.
  • the method can be applied to computer equipment.
  • the computer equipment includes a terminal 100 and a server 200 connected through a network.
  • This method can be executed in the terminal 100 or the server 200.
  • the terminal 100 can directly obtain the projection data of the original image of the target object from the detection device, and execute the above method on the terminal side; or, the terminal 100 can also be used in the acquisition of the target object.
  • the original image is sent to the server 200, so that the server 200 obtains the projection data of the original image of the target object and executes the above-mentioned method.
  • the terminal 100 may specifically be a desktop terminal (for example, a desktop computer) or a mobile terminal (for example, a notebook computer).
  • the server 200 may be implemented as an independent server or a server cluster composed of multiple servers.
  • Fig. 2 is a schematic structural diagram of a target network model used in an embodiment of the application.
  • the target network model can be a deep learning model, the main structure of which can be a residual network, and the various network parameters can be determined by training it with a large number of measured sample images, so that the training can be used
  • the latter target network model performs image reconstruction processing on the image to be tested.
  • the target network model may include a first network structure and a second network structure.
  • the first network structure has the function of performing image reconstruction processing on the input image data according to the preset algorithm to obtain low-resolution images, and its corresponding network parameters can be set in advance according to actual needs or empirical data, and remain fixed constant.
  • the second network structure can be used to perform image reconstruction processing on the low-resolution image output by the first network structure to remove noise in the low-resolution image, thereby obtaining a high-resolution image, and each network corresponding to the second network structure
  • the initial value of the parameter is set randomly, and the final value can be determined through training.
  • the second network structure may include a first convolutional layer, a residual module, and a second convolutional layer that are sequentially connected.
  • the first convolutional layer is mainly used to extract the shallow characteristic information in the low-resolution image output by the first network structure.
  • the shallow characteristic information can also be called low-level features, which usually refers to some of the image. Small details, such as edges, corners, colors, pixels, gradients, etc.
  • the residual module can be used to extract deep feature information from the shallow feature information output by the first convolutional layer.
  • the deep feature information can also be called high-level features, which can be used to identify and/or detect target regions in the image
  • the residual module may include multiple (for example, 4) repeated sub-modules, and each sub-module may include multiple (for example, 3) convolutional layers, normalization Layer (BN) and modified linear unit (ReLU), as shown in Figure 3.
  • BN normalization Layer
  • ReLU modified linear unit
  • the size of the convolution kernel shown in FIG. 3 is only an example.
  • the second convolutional layer may be used to process the shallow characteristic information extracted by the first convolutional layer and the deep characteristic information extracted by the residual module to obtain a high-resolution image.
  • this application provides an image reconstruction method, as shown in FIG. 4.
  • the method may specifically include the following steps:
  • the measured sample image may refer to the actual images of various detection objects collected by the detector.
  • the target network model may include a first network structure with fixed parameters and a second network structure with multiple convolutional layers.
  • Radon transformation processing can be performed on each acquired measured sample image to obtain a corresponding sample data set.
  • the measured sample images can be line-integrated according to multiple different specific angles, so that the projection data of the measured sample images at these specific angles can be obtained, and the corresponding projections of each measured sample image at different specific angles
  • the data constitutes a sample data set.
  • the specific value of a specific angle can be set according to actual needs or empirical data, and is not limited here.
  • the obtained sample data set can be used to train the constructed target network model, so as to determine the final network parameters corresponding to the second network structure in the target network model value.
  • This step may specifically include the following sub-steps:
  • S101 Perform image reconstruction processing on the projection data in the sample data set by using the first network structure to obtain a first sample reconstructed image.
  • the first network structure can be used to perform image reconstruction processing on the projection data in the sample data set to obtain a first sample reconstructed image.
  • image reconstruction processing can be performed on the projection data in the sample data set according to a preset algorithm such as direct back-projection method, filtered back-projection method, or Fourier direct transformation method, so as to obtain the corresponding first image. This reconstructed image.
  • S102 Perform image reconstruction processing on the first sample reconstructed image by using the second network structure to obtain a second sample reconstructed image.
  • image reconstruction processing may be performed on the first sample reconstructed image through the second network structure to obtain a second sample reconstructed image.
  • the first convolutional layer in the second network structure can be used to extract the shallow characteristic information in the first reconstructed image; then, the residual module in the second network structure can be used to extract the shallow characteristic information from the Deep feature information with richer semantic information; finally, the second convolutional layer in the second network structure can be used to process the shallow feature information and the deep feature information to obtain a second sample reconstructed image.
  • the noise in the reconstructed image of the first sample can be removed, so that the resolution of the obtained reconstructed image of the second sample can be improved.
  • S103 Construct a loss function between the reconstructed image of the second sample and the corresponding measured sample image, and determine the final network parameter corresponding to the second network structure in the target network model by solving the constructed loss function value.
  • the constructed loss function can be a mean square error (MSE) loss function (as shown in the following equation (1)), an absolute error loss function (as shown in the following equation (2)) or a smooth loss function (as shown in the following equation (3) ), etc., which are related to network parameters.
  • MSE mean square error
  • L represents the loss function
  • f(x) represents the second sample reconstructed image
  • Y represents the measured sample image.
  • the loss function L is related to the network parameters in the target network model, and the specific relationship between the two can be referred to the prior art, and it will not be repeated here.
  • the value is determined as the final value of the network parameter corresponding to the second network structure in the target network model.
  • S2 Perform image reconstruction processing on the acquired data set to be tested by using the first network structure in the trained target network model to obtain a first reconstructed image.
  • the data set to be measured may include projection data of the target object collected by the detector at multiple different projection angles, as shown in f(x,y) in FIG. 5.
  • x and y represent the horizontal and vertical coordinates
  • represents the projection angle, which can be set according to actual needs, for example, it can be 3 degrees or 6 degrees.
  • the projection data may include CT image data, PET image data, or PET/CT image data, etc., but is not limited thereto.
  • the sample data set used in the training matches the test data set, including the content, type and/or projection angle of the two. For example, both are obtained by projecting the CT image of the patient’s lungs Data set.
  • the target object may refer to an organism that needs to be detected, for example, a patient or a pet.
  • the acquired data set to be tested can be processed through the first network structure in the trained target network model to obtain the first reconstructed image.
  • the projection data in the test data set can be imaged according to a preset algorithm (for example, direct back projection method, filtered back projection method, or Fourier direct transform method, but not limited to these methods) in the first network structure
  • the first reconstructed image is obtained by sequentially back-projecting the projection data in the data set to be measured according to different projection angles.
  • the preset algorithm is the filtered back projection method, it is mainly to perform Fourier transform on the projection data in the measured data set at each projection angle, and then multiply the Fourier transformed projection data by the weighting factor and perform the Fourier transform. Inverse leaf transform, and finally perform direct back projection calculation on the projection data obtained after inverse Fourier transform to obtain the first reconstructed image.
  • the first reconstructed image is mainly obtained by performing Fourier transform on the projection data in the data set to be measured.
  • S3 Use the second network structure in the target network model to perform denoising processing on the first reconstructed image to obtain a second reconstructed image.
  • the second network structure may be used to perform denoising processing on the first reconstructed image to obtain a second reconstructed image with high resolution.
  • the first convolutional layer in the second network structure can be used to extract the shallow characteristic information in the first reconstructed image; the residual module in the second network structure can be used to extract the shallow characteristic information from the extracted shallow characteristic information. Deep-layer characteristic information; and using the second convolutional layer in the second network structure to process the shallow-layer characteristic information and the deep-layer characteristic information to obtain a second reconstructed image.
  • the network parameters corresponding to the second network structure are determined by comparing the sample reconstructed image with the real image, by using the trained second network structure to process the first reconstructed image, it is possible to remove the Noise, which can increase the resolution of the second reconstructed image obtained.
  • the embodiment of the present application uses the first network structure in the target network model to perform image reconstruction processing on the projection data set, thereby obtaining a low-resolution first reconstructed image, and then uses the first network structure in the target network model
  • the second network structure performs denoising processing on the first reconstructed image, thereby obtaining a high-resolution second reconstructed image, which can improve the quality of the reconstructed image.
  • the data processing speed can be improved. Relevant experimental data shows that using the technical solutions provided by the embodiments of this application, the reconstruction time of each image is less than 1 second.
  • an image reconstruction device which may include:
  • the training unit 610 may be configured to use the acquired measured sample images to train the constructed target network model, where the target network model includes a first network structure with fixed network parameters and a multi-layer convolutional layer The second network structure;
  • the image reconstruction unit 620 may be configured to use the trained first network structure to perform image reconstruction processing on the acquired data set to be tested according to a preset algorithm to obtain a first reconstructed image.
  • the data set to be tested is included in a plurality of Projection data of the target object collected under different projection angles;
  • the denoising unit 630 may be configured to perform denoising processing on the first reconstructed image by using the trained second network structure to obtain the second reconstructed image.
  • the training unit 610 may be specifically configured to: perform Radon transformation processing on the acquired measured sample image to obtain a corresponding sample data set, and use the obtained sample data set to compare the constructed target network The model is trained.
  • the above device uses the training unit, the image reconstruction unit and the denoising unit to perform image reconstruction processing on the data set to be tested, which can improve the quality of the reconstructed image and also increase the data processing speed.
  • Fig. 7 shows a schematic structural diagram of a computer device in an embodiment.
  • the computer device may specifically be the terminal 100 in FIG. 1.
  • the computer equipment includes a processor, a memory, a network interface, an input device, and a display connected through a system bus.
  • the memory includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium of the computer device stores an operating system and may also store a computer program.
  • the processor can make the processor execute the image reconstruction method described in the foregoing embodiment.
  • the internal memory may also store a computer program, and when the computer program is executed by the processor, it executes the image reconstruction method described in the above embodiment.
  • Fig. 8 shows a schematic structural diagram of a computer device in another embodiment.
  • the computer device may specifically be the server 200 in FIG. 1.
  • the computer device includes a processor, a memory, and a network interface connected through a system bus.
  • the memory includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium of the computer device stores an operating system, and may also store a computer program.
  • the processor can make the processor execute the image reconstruction method described in the foregoing embodiment.
  • the internal memory may also store a computer program, and when the computer program is executed by the processor, it executes the image reconstruction method described in the above embodiment.
  • FIG. 7 and FIG. 8 are only block diagrams of part of the structure related to the solution of the present application, and do not constitute a limitation on the computer equipment to which the solution of the present application is applied.
  • the computer device may include more or fewer components than shown in the figures, or combine certain components, or have a different component configuration.
  • the present application also provides an image processing system.
  • the image processing system may include the computer device in FIG. 7 or FIG. 8 and a detection device connected to it.
  • the detection device can be used To obtain projection data by scanning the target object and provide the obtained projection data to the computer equipment.
  • the detection device may be any device capable of detecting radioactive rays. For example, it may include a CT scanner, a PET detector, or a PET/CT device, etc., but is not limited thereto.
  • the present application also provides a computer-readable storage medium that stores a computer program that can implement the corresponding functions described in the foregoing method embodiment when the computer program is executed.
  • the computer program can also be run on the computer device as shown in FIG. 7 or FIG. 8.
  • the memory of the computer device contains various program modules constituting the device, and the computer program constituted by each program module can realize the functions corresponding to the steps in the image reconstruction method described in the foregoing embodiment when executed.
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
  • SRAM static RAM
  • DRAM dynamic RAM
  • SDRAM synchronous DRAM
  • DDRSDRAM double data rate SDRAM
  • ESDRAM enhanced SDRAM
  • SLDRAM synchronous chain Channel
  • memory bus Radbus direct RAM
  • RDRAM direct memory bus dynamic RAM
  • RDRAM memory bus dynamic RAM

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Image Analysis (AREA)

Abstract

本申请实施例公开了图像重建方法、装置、设备、系统及计算机可读存储介质,该方法包括:利用所获取的已测样本图像对所构建的目标网络模型进行训练,其中,所述目标网络模型包括具有固定网络参数的第一网络结构以及含有多层卷积层的第二网络结构,利用所述第一网络结构按照预设算法对所获取的待测数据集进行图像重建处理以得到第一重建图像,所述待测数据集包括在多个不同投影角度下采集到的目标对象的投影数据;利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像。通过利用本申请实施例提供的技术方案,可以提高重建图像的质量及数据处理速度。

Description

图像重建方法、装置、设备、系统及计算机可读存储介质 技术领域
本申请涉及图像处理技术领域,特别涉及一种图像重建方法、装置、设备、系统及计算机可读存储介质。
背景技术
在临床实践中,在利用计算机断层成像(Computed Tomography,简称CT)、正电子发射断层成像(Positron Emission Tomography,简称PET)等医疗设备对病人进行扫描检查后,通常需要对扫描数据进行图像重建,以获得可供医生查看的图像。
目前,实现图像重建的方法有很多,例如,直接反投影法、滤波反投影法(Filtered Back Projection,简称FBP)或傅立叶直接变换法等解析方法,这些方法的实质是利用所采集的投影数据求解图像矩阵中的像素值以重新构造图像。
在实现本申请过程中,发明人发现现有技术中至少存在如下问题:
(1)FBP等解析方法对投影数据要求较高,并且通过该方法获得的重建图像分辨率较低;
(2)现有的迭代方法对系统要求高,处理过程中的系统响应矩阵较大,收敛速度慢,并且迭代次数有一定的随机性,因而导致数据处理速度较慢。
发明内容
本申请实施例的目的是提供一种图像重建方法、装置、设备、系统及计算机可读存储介质,以解决现有技术中存在的至少一种问题。
为了解决上述技术问题,本申请实施例提供了一种图像重建方法,该图像重建方法包括:
利用所获取的已测样本图像对所构建的目标网络模型进行训练,其中,所述目标网络模型包括具有固定网络参数的第一网络结构以及含有多层卷积 层的第二网络结构,
利用所述第一网络结构按照预设算法对所获取的待测数据集进行图像重建处理以得到第一重建图像,所述待测数据集包括在多个不同投影角度下采集到的目标对象的投影数据;
利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像。
可选地,利用所获取的已测样本图像对所构建的目标网络模型进行训练的步骤包括:
对所获取的所述已测样本图像进行拉东变换处理以获得对应的样本数据集,所述样本数据集包括所述已测样本图像在多个不同特定角度下所对应的投影数据,并且所述样本数据集与所述待测数据集相匹配;
利用所获得的所述样本数据集对所构建的所述目标网络模型进行训练。
可选地,利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像的步骤包括:
利用所述第二网络结构中的第一卷积层提取出所述第一重建图像中的浅层特性信息;
利用所述第二网络结构中的残差模块从所述浅层特性信息中提取出深层特性信息;
利用所述第二网络结构中的第二卷积层对所述浅层特性信息和所述深层特性信息进行处理以得到第二样本图像。
可选地,所述预设算法包括直接反投影法、滤波反投影法或傅里叶直接变换法。
可选地,所述投影数据包括CT图像数据、PET图像数据或PET/CT图像数据。
本申请实施例还提供了一种图像重建装置,其上构建有目标网络模型,该图像重建装置包括:
训练单元,其被配置为利用所获取的已测样本图像对所述目标网络模型进行训练,其中,所述目标网络模型包括具有固定网络参数的第一网络结构以及含有多层卷积层的第二网络结构,
图像重建单元,其被配置为利用所述第一网络结构按照预设算法对所获取的待测数据集进行图像重建处理以得到第一重建图像,所述待测数据集包括在多个不同投影角度下采集到的目标对象的投影数据;以及
去噪单元,其被配置为利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像。
可选地,所述训练单元具体被配置为:
对所获取的所述已测样本图像进行拉东变换处理以获得对应的样本数据集,所述样本数据集包括所述已测样本图像在多个不同投影角度下所对应的投影数据,并且所述样本数据集与所述待测数据集相匹配;
利用所获得的所述样本数据集对所构建的所述目标网络模型进行训练。
可选地,所述第二网络结构包括依次连接的第一卷积层、残差模块和第二卷积层。
本申请实施例还提供了一种计算机设备,该计算机设备包括存储器和处理器,所述存储器存储有计算机程序,在所述计算机程序被所述处理器执行时,所述处理器执行上述图像重建方法。
本申请实施例还提供了一种图像处理系统,该图像处理系统包括上述计算机设备以及探测设备,其中,所述探测设备被配置为通过对目标对象进行扫描而获得投影数据并且将所获得的投影数据提供给所述计算机设备。
可选地,所述探测设备包括CT扫描仪、PET探测器或PET/CT设备。
本申请实施例还提供了一种计算机可读存储介质,该计算机可读存储介质存储有计算机程序,所述计算机程序被执行时能够实现与上述图像重建方法对应的功能。
由以上本申请实施例提供的技术方案可见,本申请实施例通过利用目标网络模型中的第一网络结构按照预设算法对投影数据集进行图像重建处理,从而获得第一重建图像,然后利用目标网络模型中的第二网络结构对第一重建图像进行去噪处理,从而获得第二重建图像,这可以提高重建图像的质量。而且,通过利用目标网络模型对投影数据集进行图像重建处理,可以提高数据处理速度。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请中记载的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动性的前提下,还可以根据这些附图获得其他的附图。
图1是本申请的一个实施例中的图像重建方法的应用环境图;
图2是本申请的一个实施例中所利用的目标网络模型的结构示意图;
图3是目标网络模型中的残差模块中的子模块的结构示意图;
图4是本申请的一个实施例提供的图像重建方法的流程示意图;
图5是所获取的待测数据集的示意图;
图6是本申请的一个实施例提供的一种图像重建装置的结构示意图;
图7是本申请的一个实施例中的计算机设备的结构示意图;
图8是本申请的另一个实施例中的计算机设备的结构示意图;
图9是本申请的一个实施例中的图像处理系统的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是用于解释说明本申请的一部分实施例,而不是全部的实施例,并不希望限制本申请的范围或权利要求书。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其它实施例,都应当属于本申请保护的范围。
需要说明的是,当元件被称为“设置在”另一个元件上,它可以直接设置在另一个元件上或者也可以存在居中的元件。当元件被称为“连接/联接”至另一个元件,它可以是直接连接/联接至另一个元件或者可能同时存在居中元件。本文所使用的术语“连接/联接”可以包括电气和/或机械物理连接/联接。本文所使用的术语“包括/包含”指特征、步骤或元件的存在,但并不排除一个或更多个其它特征、步骤或元件的存在或添加。本文所使用的术语“和/或”包括一个或多个相关所列项目的任意的和所有的组合。
除非另有定义,本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同。本文中所使用的术语只是为了描述具体实施例的目的,而并不是旨在限制本申请。
另外,在本申请的描述中,术语“第一”、“第二”、“第三”等仅用于描述目的和区别类似的对象,两者之间并不存在先后顺序,也不能理解为指示或暗示相对重要性。此外,在本申请的描述中,除非另有说明,“多个”的含义是两个或两个以上。
图1为一个实施例中的图像重建方法的应用环境图。参照图1,该方法可以应用于计算机设备。该计算机设备包括通过网络连接的终端100和服务器200。该方法可以在终端100或服务器200中执行,例如,终端100可直接从探测设备获取目标对象的原始图像的投影数据,并在终端侧执行上述方法;或者,终端100也可在获取目标对象的原始图像后将原始图像发送至服务器200,使得服务器200获取目标对象的原始图像的投影数据并执行上述方法。终端100具体可以是台式终端(例如,台式电脑)或移动终端(例如,笔记本电脑)。服务器200可以用独立的服务器或者是多个服务器组成的服务器集群来实现。
图2为本申请的一个实施例中所利用的目标网络模型的结构示意图。参照图2,目标网络模型可以是一种深度学习模型,其主体结构可以为残差网络,并且可以通过利用大量的已测样本图像对其进行训练来确定其中的各个网络参数,从而可以利用训练后的目标网络模型对待测图像进行图像重建处理。该目标网络模型可以包括第一网络结构以及第二网络结构。第一网络结构具有按照预设算法对所输入的图像数据进行图像重建处理以得到低分辨率图像的功能,并且其所对应的网络参数可以是按照实际需求或经验数据提前设置的,并且保持固定不变。第二网络结构可以用于对第一网络结构输出的低分辨率图像进行图像重建处理,以去除低分辨率图像中的噪声,从而得到高分辨率图像,并且第二网络结构所对应的各个网络参数的初始值是随机设置的,而最终值可以通过训练来确定。
另外,第二网络结构可以包括依次连接的第一卷积层、残差模块和第二卷积层。其中,第一卷积层主要用于提取第一网络结构所输出的低分辨率图 像中的浅层特性信息,该浅层特性信息也可以称为低水平特征,其通常是指图像中的一些小的细节信息,例如,边缘、角、颜色、像素、梯度等。残差模块可以用于从第一卷积层输出的浅层特性信息中提取深层特性信息,该深层特性信息也可以称为高水平特征,其可以用于识别和/或检测图像中的目标区域的形状,具有更丰富的语义信息,该残差模块可以包括多个(例如,4个)重复的子模块,每个子模块均可以包括多个(例如,3个)卷积层、归一化层(BN)以及修正线性单元(ReLU),如图3所示。需要说明的是,图3中示出的卷积核大小仅为示例。第二卷积层可以用于对第一卷积层提取出的浅层特性信息和残差模块提取出的深层特性信息进行处理以得到高分辨率图像。
关于第一卷积层、残差模块和第二卷积层的详细描述,可以参照现有技术,在此不再赘叙。
在一个实施例中,本申请提供了一种图像重建方法,如图4所示。该方法具体可以包括如下步骤:
S1:利用所获取的已测样本图像对所构建的目标网络模型进行训练。
已测样本图像可以是指利用探测器采集到的各种探测对象的实际图像。目标网络模型可以包括具有固定参数的第一网络结构和含有多层卷积层的第二网络结构。
在获取多个已测样本图像之后,可以对所获取的每个已测样本图像进行拉东变换处理以获得对应的样本数据集。具体地,可以按照多个不同特定角度分别对已测样本图像进行线积分,从而可以获得已测样本图像在这些特定角度下的投影数据,每个已测样本图像在不同特定角度下对应的投影数据便构成了一个样本数据集。特定角度的具体数值可以根据实际需求或经验数据来设置,在此不限制。
在现有技术中,由于医学图像比较珍贵,所以比较难以获得较多的样本数据集,而在本申请中,可以通过对少量的已测样本图像进行拉东变换处理来获得大量的样本数据集,这提高了获得样本数据集的便利性,也使得后续对目标网络模型的训练能够顺利进行。
在获得已测样本图像的样本数据集之后,可以利用所获得的样本数据集对所构建的目标网络模型进行训练,从而确定出目标网络模型中的第二网络 结构所对应的各个网络参数的最终值。该步骤具体可以包括以下子步骤:
S101:利用第一网络结构对样本数据集中的投影数据进行图像重建处理以得到第一样本重建图像。
在获得样本数据集之后,可以利用第一网络结构对样本数据集中的投影数据进行图像重建处理以得到第一样本重建图像。具体地,可以在第一网络结构中按照直接反投影法、滤波反投影法或傅里叶直接变换法等预设算法对样本数据集中的投影数据进行图像重建处理,从而获得对应的第一样本重建图像。
关于利用直接反投影法、滤波反投影法或傅里叶直接变换法等方法对投影数据进行图像重建处理的具体过程可以参照现有技术,在此不再赘叙。
S102:利用第二网络结构对第一样本重建图像进行图像重建处理以得到第二样本重建图像。
在通过第一网络结构输出第一样本重建图像之后,可以通过第二网络结构对第一样本重建图像进行图像重建处理以得到第二样本重建图像。具体地,可以利用第二网络结构中的第一卷积层提取出第一重建图像中的浅层特性信息;接着,可以利用第二网络结构中的残差模块从浅层特性信息中提取出具有更丰富语义信息的深层特性信息;最后,可以利用第二网络结构中的第二卷积层对浅层特性信息和深层特性信息进行处理以得到第二样本重建图像。
关于上述卷积层和残差模块的具体处理过程,可以参照现有技术,在此不再赘叙。
通过该步骤,可以去除第一样本重建图像中的噪声,使得可以提高所得到的第二样本重建图像的分辨率。
S103:构建第二样本重建图像与对应的已测样本图像之间的损失函数,并且通过对所构建的损失函数进行求解来确定出目标网络模型中的第二网络结构所对应的网络参数的最终值。
所构建的损失函数可以为均方误差(MSE)损失函数(如下式(1)所示)、绝对误差损失函数(如下式(2)所示)或平滑损失函数(如下式(3)所示)等,其与网络参数有关。
L=|f(x)-Y| 2            (1)
L=|f(x)-Y|                (2)
Figure PCTCN2020132371-appb-000001
上式中,L表示损失函数,f(x)表示第二样本重建图像,Y表示已测样本图像。虽然上式中没有示出,但损失函数L与目标网络模型中的网络参数有关,而关于二者之间的具体关系,可以参照现有技术,在此也不再赘叙。
可以利用向前传播算法计算损失函数,反向传播根据损失函数计算得到的误差,并利用梯度下降法更新网络参数,反复迭代一定次数,最终将损失函数取得最优解时所对应的相关参数的数值确定为目标网络模型中的第二网络结构所对应的网络参数的最终值。关于具体求解过程,可以参照现有技术,在此不再赘叙。
S2:利用训练后的目标网络模型中的第一网络结构对所获取的待测数据集进行图像重建处理以得到第一重建图像。
该待测数据集可以包括探测器在多个不同投影角度下采集到的目标对象的投影数据,如图5中的f(x,y)所示。其中,x和y表示横、纵坐标;θ表示投影角度,其可以根据实际需求来设置,例如,可以为3度或6度等。该投影数据可以包括CT图像数据、PET图像数据或PET/CT图像数据等,但不限于此。训练中采用的样本数据集与该待测数据集相匹配,包括二者的内容、类型和/或投影角度等相匹配,例如,二者都是通过对病人肺部的CT图像进行投影而得到的数据集。目标对象可以是指需要被探测的生物体,例如,病人或宠物等。
在获取探测设备输出的待测数据集之后,可以通过训练后的目标网络模型中的第一网络结构对所获取的待测数据集进行处理以得到第一重建图像。具体地,可以在第一网络结构中按照预设算法(例如,直接反投影法、滤波反投影法或傅里叶直接变换法等,但不限于这些方法)对待测数据集中的投影数据进行图像重建处理以得到第一重建图像L 2=|f(x)-Y| 2
当预设算法为直接反投影法时,主要是通过将待测数据集中的投影数据按照不同投影角度依次反投影来得到第一重建图像。
当预设算法为滤波反投影法时,主要是在每个投影角度对待测数据集中的对投影数据进行傅里叶变换,然后将傅里叶变换后的投影数据乘以权重因子并且进行傅里叶逆变换,最后再对傅里叶逆变换后所得到的投影数据进行直接反投影计算,从而得到第一重建图像。
当预设算法为傅里叶直接变换法时,主要是通过对待测数据集中的投影数据进行傅里叶变换来得到第一重建图像。
S3:利用目标网络模型中的第二网络结构对第一重建图像进行去噪处理以得到第二重建图像。
在第一网络结构输出第一重建图像之后,可以利用第二网络结构对第一重建图像进行去噪处理,以获得具有高分辨率的第二重建图像。具体地,可以利用第二网络结构中的第一卷积层提取出第一重建图像中的浅层特性信息;利用第二网络结构中的残差模块从所提取的浅层特性信息中提取出深层特性信息;以及利用第二网络结构中的第二卷积层对浅层特性信息和深层特性信息进行处理以得到第二重建图像。
由于第二网络结构所对应的网络参数是通过将样本重建图像与真实图像进行对比来确定的,所以通过利用训练后的第二网络结构来处理第一重建图像,可以去除第一重建图像中的噪声,从而可以提高所得到的第二重建图像的分辨率。
通过上述描述可以看出,本申请实施例通过利用目标网络模型中的第一网络结构来对投影数据集进行图像重建处理,从而获得低分辨率的第一重建图像,然后利用目标网络模型中的第二网络结构对第一重建图像进行去噪处理,从而获得高分辨率的第二重建图像,这可以提高重建图像的质量。而且,通过利用目标网络模型对投影数据集进行图像重建处理,可以提高数据处理速度。经相关实验数据表明,利用本申请实施例提供的技术方案,每张图像重建时间小于1s。
如图6所示,本申请实施例还提供了一种图像重建装置,该装置可以包括:
训练单元610,其可以被配置为利用所获取的已测样本图像对所构建的目标网络模型进行训练,其中,该目标网络模型包括具有固定网络参数的第一 网络结构以及含有多层卷积层的第二网络结构;
图像重建单元620,其可以被配置为利用训练后的第一网络结构按照预设算法对所获取的待测数据集进行图像重建处理以得到第一重建图像,该待测数据集包括在多个不同投影角度下采集到的目标对象的投影数据;
去噪单元630,其可以被配置为利用训练后的第二网络结构对第一重建图像进行去噪处理以得到第二重建图像。
在一个实施例中,训练单元610具体可以被配置为:对所获取的已测样本图像进行拉东变换处理以获得对应的样本数据集,并且利用所获得的样本数据集对所构建的目标网络模型进行训练。
关于上述单元的具体描述,可以参照上面方法实施例中对步骤S1-S3的描述,在此不再赘叙。
上述装置通过利用训练单元、图像重建单元和去噪单元来对待测数据集进行图像重建处理,可以提高重建图像的质量,并且也可以提高数据处理速度。
图7示出了一个实施例中的计算机设备的结构示意图。该计算机设备具体可以是图1中的终端100。如图7所示,该计算机设备包括通过系统总线连接的处理器、存储器、网络接口、输入装置和显示器。其中,存储器包括非易失性存储介质和内存储器。该计算机设备的非易失性存储介质存储有操作系统,还可存储有计算机程序,在该计算机程序被处理器执行时,可使得处理器执行上述实施例中描述的图像重建方法。该内存储器中也可储存有计算机程序,该计算机程序被处理器执行时,执行上述实施例中描述的图像重建方法。
图8示出了另一个实施例中的计算机设备的结构示意图。该计算机设备具体可以是图1中的服务器200。如图8所示,该计算机设备包括通过系统总线连接的处理器、存储器以及网络接口。其中,存储器包括非易失性存储介质和内存储器。该计算机设备的非易失性存储介质存储有操作系统,还可存储有计算机程序,该计算机程序被处理器执行时,可使得处理器执行上述实施例中描述的图像重建方法。该内存储器中也可储存有计算机程序,该计算机程序被处理器执行时,执行上述实施例中描述的图像重建方法。
本领域技术人员可以理解,图7和图8中示出的结构,仅仅是与本申请方 案相关的部分结构的框图,并不构成对本申请方案所应用于其上的计算机设备的限定,具体的计算机设备可以包括比图中所示更多或更少的部件,或者组合某些部件,或者具有不同的部件配置。
在一个实施例中,如图9所示,本申请还提供了一种图像处理系统,该图像处理系统可以包括图7或图8中的计算机设备以及与其连接的探测设备,该探测设备可以用于通过对目标对象进行扫描而获得投影数据并且将所获得的投影数据提供给计算机设备。该探测设备可以是能够探测放射性射线的任意设备,例如,可以包括CT扫描仪、PET探测器或PET/CT设备等,但不限于此。
在一个实施例中,本申请还提供了一种计算机可读存储介质,该计算机可读存储介质种存储有计算机程序,该计算机程序被执行时能够实现上述方法实施例中描述对应的功能。该计算机程序还可在如图7或图8所示的计算机设备上运行。该计算机设备的存储器包含组成该装置的各个程序模块,各个程序模块构成的计算机程序被执行时能够实现与上述实施例中描述的图像重建方法中的各个步骤所对应的功能。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的程序可存储于一非易失性计算机可读取存储介质中,该程序在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的各实施例中所使用的对存储器、存储介质、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM以多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
上述实施例阐明的系统、设备、装置、单元等,具体可以由半导体芯片、 计算机芯片和/或实体实现,或者由具有某种功能的产品来实现。为了描述的方便,描述以上装置时以功能分为各种单元分别描述。当然,在实施本申请时可以把各单元的功能在同一个或多个芯片中实现。
虽然本申请提供了如上述实施例或流程图所述的方法操作步骤,但基于常规或者无需创造性的劳动在所述方法中可以包括更多或者更少的操作步骤。在逻辑性上不存在必要因果关系的步骤中,这些步骤的执行顺序不限于本申请实施例提供的执行顺序。
本说明书中的各个实施例均采用递进的方式描述,各个实施例之间相同相似的部分互相参见即可,每个实施例重点说明的都是与其它实施例的不同之处。另外,以上实施例的各技术特征可以进行任意的组合,为使描述简洁,未对上述实施例中的各个技术特征所有可能的组合都进行描述,然而,只要这些技术特征的组合不存在矛盾,都应当认为是本说明书记载的范围。
上述实施例是为便于该技术领域的普通技术人员能够理解和使用本申请而描述的。熟悉本领域技术的人员显然可以容易地对这些实施例做出各种修改,并把在此说明的一般原理应用到其它实施例中而不必经过创造性的劳动。因此,本申请不限于上述实施例,本领域技术人员根据本申请的揭示,不脱离本申请范畴所做出的改进和修改都应该在本申请的保护范围之内。

Claims (12)

  1. 一种图像重建方法,其特征在于,包括:
    利用所获取的已测样本图像对所构建的目标网络模型进行训练,其中,所述目标网络模型包括具有固定网络参数的第一网络结构以及含有多层卷积层的第二网络结构,
    利用所述第一网络结构按照预设算法对所获取的待测数据集进行图像重建处理以得到第一重建图像,所述待测数据集包括在多个不同投影角度下采集到的目标对象的投影数据;
    利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像。
  2. 根据权利要求1所述的图像重建方法,其特征在于,利用所获取的已测样本图像对所构建的目标网络模型进行训练的步骤包括:
    对所获取的所述已测样本图像进行拉东变换处理以获得对应的样本数据集,所述样本数据集包括所述已测样本图像在多个不同特定角度下所对应的投影数据,并且所述样本数据集与所述待测数据集相匹配;
    利用所获得的所述样本数据集对所构建的所述目标网络模型进行训练。
  3. 根据权利要求1或2所述的图像重建方法,其特征在于,利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像的步骤包括:
    利用所述第二网络结构中的第一卷积层提取出所述第一重建图像中的浅层特性信息;
    利用所述第二网络结构中的残差模块从所述浅层特性信息中提取出深层特性信息;
    利用所述第二网络结构中的第二卷积层对所述浅层特性信息和所述深层特性信息进行处理以得到第二重建图像。
  4. 根据权利要求1或2所述的图像重建方法,其特征在于,所述预设算法包括直接反投影法、滤波反投影法或傅里叶直接变换法。
  5. 根据权利要求1或2所述的图像重建方法,其特征在于,所述投影数据包括CT图像数据、PET图像数据或PET/CT图像数据。
  6. 一种图像重建装置,其上构建有目标网络模型,其特征在于,所述图像重建装置包括:
    训练单元,其被配置为利用所获取的已测样本图像对所述目标网络模型进行训练,其中,所述目标网络模型包括具有固定网络参数的第一网络结构以及含有多层卷积层的第二网络结构,
    图像重建单元,其被配置为利用所述第一网络结构按照预设算法对所获取的待测数据集进行图像重建处理以得到第一重建图像,所述待测数据集包括在多个不同投影角度下采集到的目标对象的投影数据;以及
    去噪单元,其被配置为利用训练后的所述第二网络结构对所述第一重建图像进行去噪处理以得到第二重建图像。
  7. 根据权利要求6所述的图像重建装置,其特征在于,所述训练单元具体被配置为:
    对所获取的所述已测样本图像进行拉东变换处理以获得对应的样本数据集,所述样本数据集包括所述已测样本图像在多个不同投影角度下所对应的投影数据,并且所述样本数据集与所述待测数据集相匹配;
    利用所获得的所述样本数据集对所构建的所述目标网络模型进行训练。
  8. 根据权利要求6所述的图像重建装置,其特征在于,所述第二网络结构包括依次连接的第一卷积层、残差模块和第二卷积层。
  9. 一种计算机设备,其特征在于,所述计算机设备包括存储器和处理器,所述存储器存储有计算机程序,在所述计算机程序被所述处理器执行时,所述处理器执行如权利要求1至5中任一项所述的图像重建方法。
  10. 一种图像处理系统,其特征在于,所述图像处理系统包括权利要求9中所述的计算机设备以及探测设备,其中,所述探测设备被配置为通过对目标对象进行扫描而获得投影数据并且将所获得的投影数据提供给所述计算机设备。
  11. 根据权利要求10所述的图像处理系统,其特征在于,所述探测设备包括CT扫描仪、PET探测器或PET/CT设备。
  12. 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质存储有计算机程序,所述计算机程序被执行时能够实现与权利要求1至5中任一项所述的图像重建方法对应的功能。
PCT/CN2020/132371 2020-01-02 2020-11-27 图像重建方法、装置、设备、系统及计算机可读存储介质 WO2021135773A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010001252.1 2020-01-02
CN202010001252.1A CN111223160A (zh) 2020-01-02 2020-01-02 图像重建方法、装置、设备、系统及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2021135773A1 true WO2021135773A1 (zh) 2021-07-08

Family

ID=70832232

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/132371 WO2021135773A1 (zh) 2020-01-02 2020-11-27 图像重建方法、装置、设备、系统及计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN111223160A (zh)
WO (1) WO2021135773A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113761771A (zh) * 2021-09-16 2021-12-07 中国人民解放军国防科技大学 多孔材料吸声性能预测方法、装置、电子设备和存储介质
CN114092330A (zh) * 2021-11-19 2022-02-25 长春理工大学 一种轻量化多尺度的红外图像超分辨率重建方法
CN114155340A (zh) * 2021-10-20 2022-03-08 清华大学 扫描光场数据的重建方法、装置、电子设备及存储介质
CN115034000A (zh) * 2022-05-13 2022-09-09 深圳模德宝科技有限公司 一种工艺设计的方法
CN115115726A (zh) * 2022-05-10 2022-09-27 深圳市元甪科技有限公司 多频电阻抗层析成像图像的重建方法、装置、设备及介质
CN116503506A (zh) * 2023-06-25 2023-07-28 南方医科大学 一种图像重建方法、系统、装置及存储介质
CN115115726B (zh) * 2022-05-10 2024-06-07 深圳市元甪科技有限公司 多频电阻抗层析成像图像的重建方法、装置、设备及介质

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111223160A (zh) * 2020-01-02 2020-06-02 苏州瑞派宁科技有限公司 图像重建方法、装置、设备、系统及计算机可读存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109300167A (zh) * 2017-07-25 2019-02-01 清华大学 重建ct图像的方法和设备以及存储介质
CN109300166A (zh) * 2017-07-25 2019-02-01 同方威视技术股份有限公司 重建ct图像的方法和设备以及存储介质
US20190273948A1 (en) * 2019-01-08 2019-09-05 Intel Corporation Method and system of neural network loop filtering for video coding
CN110544282A (zh) * 2019-08-30 2019-12-06 清华大学 基于神经网络的三维多能谱ct重建方法和设备及存储介质
CN110599409A (zh) * 2019-08-01 2019-12-20 西安理工大学 基于多尺度卷积组与并行的卷积神经网络图像去噪方法
CN111223160A (zh) * 2020-01-02 2020-06-02 苏州瑞派宁科技有限公司 图像重建方法、装置、设备、系统及计算机可读存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109300167A (zh) * 2017-07-25 2019-02-01 清华大学 重建ct图像的方法和设备以及存储介质
CN109300166A (zh) * 2017-07-25 2019-02-01 同方威视技术股份有限公司 重建ct图像的方法和设备以及存储介质
US20190273948A1 (en) * 2019-01-08 2019-09-05 Intel Corporation Method and system of neural network loop filtering for video coding
CN110599409A (zh) * 2019-08-01 2019-12-20 西安理工大学 基于多尺度卷积组与并行的卷积神经网络图像去噪方法
CN110544282A (zh) * 2019-08-30 2019-12-06 清华大学 基于神经网络的三维多能谱ct重建方法和设备及存储介质
CN111223160A (zh) * 2020-01-02 2020-06-02 苏州瑞派宁科技有限公司 图像重建方法、装置、设备、系统及计算机可读存储介质

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113761771A (zh) * 2021-09-16 2021-12-07 中国人民解放军国防科技大学 多孔材料吸声性能预测方法、装置、电子设备和存储介质
CN113761771B (zh) * 2021-09-16 2024-05-28 中国人民解放军国防科技大学 多孔材料吸声性能预测方法、装置、电子设备和存储介质
CN114155340A (zh) * 2021-10-20 2022-03-08 清华大学 扫描光场数据的重建方法、装置、电子设备及存储介质
CN114155340B (zh) * 2021-10-20 2024-05-24 清华大学 扫描光场数据的重建方法、装置、电子设备及存储介质
CN114092330A (zh) * 2021-11-19 2022-02-25 长春理工大学 一种轻量化多尺度的红外图像超分辨率重建方法
CN114092330B (zh) * 2021-11-19 2024-04-30 长春理工大学 一种轻量化多尺度的红外图像超分辨率重建方法
CN115115726A (zh) * 2022-05-10 2022-09-27 深圳市元甪科技有限公司 多频电阻抗层析成像图像的重建方法、装置、设备及介质
CN115115726B (zh) * 2022-05-10 2024-06-07 深圳市元甪科技有限公司 多频电阻抗层析成像图像的重建方法、装置、设备及介质
CN115034000A (zh) * 2022-05-13 2022-09-09 深圳模德宝科技有限公司 一种工艺设计的方法
CN115034000B (zh) * 2022-05-13 2023-12-26 深圳模德宝科技有限公司 一种工艺设计的方法
CN116503506A (zh) * 2023-06-25 2023-07-28 南方医科大学 一种图像重建方法、系统、装置及存储介质
CN116503506B (zh) * 2023-06-25 2024-02-06 南方医科大学 一种图像重建方法、系统、装置及存储介质

Also Published As

Publication number Publication date
CN111223160A (zh) 2020-06-02

Similar Documents

Publication Publication Date Title
WO2021135773A1 (zh) 图像重建方法、装置、设备、系统及计算机可读存储介质
US11610346B2 (en) Image reconstruction using machine learning regularizers
CN110889005B (zh) 搜索医学参考图像
Zhang et al. Image prediction for limited-angle tomography via deep learning with convolutional neural network
WO2021139439A1 (zh) 图像重建方法、装置、设备、系统及计算机可读存储介质
US20170372193A1 (en) Image Correction Using A Deep Generative Machine-Learning Model
Lee et al. Machine friendly machine learning: interpretation of computed tomography without image reconstruction
Legg et al. Improving accuracy and efficiency of mutual information for multi-modal retinal image registration using adaptive probability density estimation
CN111080584B (zh) 医学图像的质控方法、计算机设备和可读存储介质
US10380738B2 (en) Computed tomography image processing and data analysis
WO2022089079A1 (zh) 图像处理方法、装置、设备、系统及计算机可读存储介质
WO2020098134A1 (zh) Ct图像的重建方法、装置、设备及存储介质
US10825149B2 (en) Defective pixel correction using adversarial networks
CN111709897B (zh) 一种基于域变换的正电子发射断层图像的重建方法
CN110084868B (zh) 图像校正方法、装置、计算机设备和可读存储介质
Wang et al. Removing ring artifacts in CBCT images via generative adversarial networks with unidirectional relative total variation loss
Piccolomini et al. A fast total variation-based iterative algorithm for digital breast tomosynthesis image reconstruction
CN111243052A (zh) 图像重建方法、装置、计算机设备和存储介质
Wang et al. Sparse-view cone-beam CT reconstruction by bar-by-bar neural FDK algorithm
KR101412369B1 (ko) 영상의 객체 검출 방법 및 장치
KR101923962B1 (ko) 의료 영상의 열람을 지원하는 방법 및 이를 이용한 장치
Tian et al. Retinal fundus image superresolution generated by optical coherence tomography based on a realistic mixed attention GAN
US11557068B2 (en) Constrained reconstruction model to restore missing wedge from multiple observations with limited range projections
CN113643394A (zh) 散射校正方法、装置、计算机设备和存储介质
Davradou et al. Diabetic foot ulcers monitoring by employing super resolution and noise reduction deep learning techniques

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20910165

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20910165

Country of ref document: EP

Kind code of ref document: A1