WO2022120588A1 - 一种低剂量pet图像还原方法、系统、设备和介质 - Google Patents

一种低剂量pet图像还原方法、系统、设备和介质 Download PDF

Info

Publication number
WO2022120588A1
WO2022120588A1 PCT/CN2020/134621 CN2020134621W WO2022120588A1 WO 2022120588 A1 WO2022120588 A1 WO 2022120588A1 CN 2020134621 W CN2020134621 W CN 2020134621W WO 2022120588 A1 WO2022120588 A1 WO 2022120588A1
Authority
WO
WIPO (PCT)
Prior art keywords
dictionary
patch
dose pet
image
low
Prior art date
Application number
PCT/CN2020/134621
Other languages
English (en)
French (fr)
Inventor
胡战利
郑海荣
梁栋
杨永峰
刘新
徐英杰
Original Assignee
深圳先进技术研究院
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 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Priority to US18/265,685 priority Critical patent/US20240046459A1/en
Priority to PCT/CN2020/134621 priority patent/WO2022120588A1/zh
Publication of WO2022120588A1 publication Critical patent/WO2022120588A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/003—Reconstruction from projections, e.g. tomography
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/90—Dynamic range modification of images or parts thereof
    • G06T5/94—Dynamic range modification of images or parts thereof based on local image properties, e.g. for local contrast enhancement
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10104—Positron emission tomography [PET]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • the present invention relates to the field of PET image processing, in particular, to a low-dose PET image restoration method, system, device and medium.
  • PET/MR large-scale functional metabolic and molecular imaging diagnostic equipment combining positron emission tomography (PET) and magnetic resonance imaging (MRI), with both PET and MR inspection functions, with high sensitivity and accuracy It has the advantages of good sex and less radiation, and has a significant effect in the diagnosis of many diseases, especially in the diagnosis of tumors and cardiovascular and cerebrovascular diseases.
  • the PET imaging dose range is 4.43-7.35mSv, with an average of 5.89mSv.
  • the standard dose of PET imaging agent still has a certain amount of radiation to the human body, and the cumulative effect will increase the possibility of various diseases and affect human health. Therefore, reasonably reducing the dose of PET to make it lower than the range of the PET developer dose can reduce the impact of radiation on the human body. Realizing the restoration from low-dose PET images to standard-dose PET images is of great significance to the application of PET/MR technology and medical diagnosis.
  • the restoration of low-dose PET images generally has problems such as poor restoration effect, complex restoration process, and low restoration accuracy, resulting in poor restoration image quality, lengthy restoration time, and lack of restoration image content. Therefore, there is a need for a restoration method for low-dose PET images, which can solve the problems of serious noise and loss of details, and at the same time, can improve the efficiency of image restoration and the accuracy of restored images.
  • the present invention provides a low-dose PET image restoration method, and the specific scheme is as follows:
  • a low-dose PET image restoration method comprising the following steps: S1. Perform block processing on a training image including a low-dose PET image, an MR image and a standard-dose PET image to obtain a first patch, and analyze the first patch Perform a first preprocessing to obtain a second patch; S2. According to the second patch, use sparse coding and dictionary update to obtain a first joint dictionary; S3. Restore the low-dose PET image to a standard according to the first joint dictionary Restored image of dose PET.
  • the S2 also includes the following steps: S21, taking the second patch as a sample to obtain an initialization dictionary including a low-dose PET dictionary, an MR dictionary and a standard-dose PET dictionary; S22, constructing according to the initialization dictionary Initializing a joint dictionary, and constructing a target matrix according to the second patch; S23, obtaining a sparse code according to the target matrix, and iteratively updating the sparse code and the initialized joint dictionary to meet the iterative stop condition, and obtaining the first The first combined dictionary including the low-dose PET dictionary, the first MR dictionary, and the first standard-dose PET dictionary.
  • each iteration includes firstly fixing the dictionary to update the sparse coding, and then fixing the sparse coding to update the dictionary.
  • each iteration randomly selects some samples for sparse coding and dictionary update.
  • S31 performing block processing on the low-dose PET image and the MR image to obtain a third patch, and performing a second preprocessing on the third patch to obtain a fourth patch
  • S32 merge the first low-dose PET dictionary and the first MR dictionary obtained in the S2 into a second joint dictionary, and obtain a second sparse code according to the fourth patch and the second joint dictionary
  • S33 A predicted patch is obtained according to the first standard dose PET dictionary and the second sparse coding obtained in S2, and the predicted patch is restored into a two-dimensional lattice to obtain a restored image of standard dose PET.
  • the first patch is a one-dimensional vector randomly selected from multiple frames of images and extended into a one-dimensional vector, including the first patch of the low-dose PET image, the first patch of the MR image, and the standard The first patch of the dose PET image; the first patch is co-located in multiple frames of images.
  • the third patch covers the entire image of one frame according to the sequence of multiple frames of images.
  • the first preprocessing includes mapping the first patch of the low-dose PET image and the first patch of the MR image to the imaging of the standard-dose PET image through a preset matrix
  • a second patch of the low dose PET image and a second patch of the MR image are spatially derived.
  • the second preprocessing includes mapping the third patch of the low-dose PET image and the third patch of the MR image to the imaging of the standard-dose PET image through a preset matrix
  • a fourth patch of the low dose PET image and a fourth patch of the MR image are spatially derived.
  • the step S21 includes adopting the K-means clustering algorithm, taking the second patch as a sample, obtaining K cluster centers as an initialization dictionary, and performing normalization processing on the initialization dictionary.
  • D represents the initialized joint dictionary
  • Y represents the target matrix
  • Dl represents the low-dose PET dictionary
  • Dr represents the MR dictionary
  • Ds represents the standard-dose PET dictionary
  • Yl represents the second patch of the low-dose PET image
  • Yr represents the second patch of the MR image. Patch
  • Ys represents the second patch of the standard dose PET image.
  • the sparse coding expression includes:
  • X represents sparse coding
  • the dictionary update expression includes:
  • yi is the ith element of Y
  • ⁇ i is the is a diagonal matrix of diagonal elements
  • y i is the ith element of Y
  • ⁇ represents the sparse constraint coefficient
  • the dictionary update expression is solved by the gradient descent method, and the gradient descent method expression is:
  • d q is the qth atom of dictionary D
  • k is the number of iterations
  • a q is The qth column element of
  • b q is The qth column element of
  • a qq is Element at row q and column q.
  • a low-dose PET image restoration system comprising: a sample acquisition unit configured to perform block processing on training images including low-dose PET images, MR images and standard-dose PET images to obtain a first patch, and obtain a first patch for the first patch.
  • a patch performs first preprocessing to obtain a second patch;
  • a joint dictionary obtaining unit is used to obtain a first joint dictionary through sparse coding and dictionary update according to the second patch;
  • an image restoration unit is used to obtain a first joint dictionary according to the first joint
  • the dictionary restores the low-dose PET image to a restored image of the standard-dose PET.
  • the joint dictionary obtaining unit further includes: an initialization unit: used for obtaining an initialization dictionary including a low-dose PET dictionary, an MR dictionary and a standard-dose PET dictionary by using the second patch as a sample; a construction unit: used for according to The initialization dictionary constructs an initialization joint dictionary, and a target matrix is constructed according to the second patch; an iterative unit: used to obtain a sparse code according to the target matrix, and iteratively update the sparse code and the initialization joint dictionary to satisfy the iterative After the stop condition, the first combined dictionary including the first low-dose PET dictionary, the first MR dictionary and the first standard-dose PET dictionary is obtained.
  • an initialization unit used for obtaining an initialization dictionary including a low-dose PET dictionary, an MR dictionary and a standard-dose PET dictionary by using the second patch as a sample
  • a construction unit used for according to The initialization dictionary constructs an initialization joint dictionary, and a target matrix is constructed according to the second patch
  • an iterative unit used to obtain
  • the iterative unit further includes: an iterative update unit: used for updating the sparse coding by fixing the dictionary at each iteration, and then updating the dictionary by fixing the sparse coding; a sample selection unit: used for randomly selecting some samples in each iteration Do sparse coding and dictionary updates.
  • a computer device comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, all The one or more processors implement the low-dose PET image restoration method as described above.
  • the present invention proposes a low-dose PET image restoration method, system, equipment and medium, which solves the problems of poor image restoration effect, complex image restoration process, and image restoration commonly existing in the prior art. Defects such as low restoration accuracy can effectively solve the problem of serious noise and loss of details in low-dose PET images, and at the same time can improve the efficiency of image restoration and the accuracy of restored images.
  • the application of the low-dose PET reduction method to specific systems, computer equipment and computer storage media, and the realization of the method is of great significance to the development of the medical impact field.
  • Fig. 1 is the flow chart of the low-dose PET image restoration method of the present invention
  • Fig. 2 is the specific flow chart of S2 of the low-dose PET image restoration method of the present invention.
  • FIG. 3 is a block diagram of a low-dose PET image restoration system of the present invention.
  • Fig. 4 is the concrete block diagram of the low-dose PET image restoration system of the present invention.
  • FIG. 5 is a schematic diagram of applying the low-dose PET image restoration method of the present invention to a computer device.
  • this embodiment proposes a low-dose PET image restoration method.
  • the specific method steps are shown in Figure 1 of the description, and the specific scheme is as follows:
  • image preprocessing perform block processing on a training image including a low-dose PET image, an MR image and a standard-dose PET image to obtain a first patch, and perform a first preprocessing on the first patch to obtain a second patch;
  • S1 is to obtain the first patch from the training sample, and perform the first preprocessing on the first patch to obtain the second patch.
  • the training images targeted by this embodiment include low-dose PET images, MR images, and standard-dose PET images, where the low-dose PET images are images to be restored.
  • a PET image is a group of continuous human tomographic images, and multiple single-frame images constitute multiple-frame images.
  • the first patch is first obtained from the training image, and the first patch is a one-dimensional vector randomly selected from multiple frames of images and extended into a one-dimensional vector. Since the first patch is randomly selected from multiple frames of images, there will be repetitions, so the repeated parts need to be removed and then used as samples to participate in training.
  • the first preprocessing needs to be performed on the first patch.
  • the first preprocessing includes acquiring the mapping matrix M
  • the first patch of the low-dose PET image and the first patch of the MR image are mapped to the imaging space of the standard-dose PET image to obtain the mapped first patch, that is, the second patch.
  • the mapping matrix can make the mapping both accurate and general in a point-to-point, edge-to-edge, etc. manner.
  • the specific mapping matrix is expressed as:
  • S2 mainly acquires the first joint dictionary.
  • S2 consists of two main modules: sparse coding and dictionary update.
  • the dictionary is obtained by alternately sparse coding and dictionary update.
  • S2 specifically includes the following steps: S21. Obtain an initialization dictionary: take the second patch as a sample to obtain an initialization dictionary including a low-dose PET dictionary, an MR dictionary, and a standard-dose PET dictionary; S22, construct an initialization dictionary and a target matrix: according to the initialization dictionary Construct the initialized joint dictionary, and construct the target matrix according to the second patch; S23, obtain the first joint dictionary: obtain the sparse coding according to the target matrix, and iteratively update the sparse coding and the initialized joint dictionary according to the K-SVD idea until the iteration stop condition is satisfied, Each iteration includes first fixing the dictionary to update the sparse coding, and then fixing the sparse coding to update the dictionary.
  • the acquired first joint dictionary includes a first low-dose PET dictionary, a first MR dictionary, and a first standard-dose PET dictionary.
  • the specific steps of S2 are shown in Figure 2 of the description.
  • the present embodiment adopts Local Coordinate Coding for dictionary learning, and the specific expression is:
  • X represents the sparse coding
  • D represents the feature matrix (dictionary)
  • x i represents the sparse coefficient of the ith sample
  • d represents the feature dimension of the feature dictionary
  • q represents the qth column of the dictionary.
  • S21 takes the second patch as a sample to obtain an initialization dictionary including a low-dose PET dictionary, an MR dictionary, and a standard-dose PET dictionary.
  • an initialization dictionary including a low-dose PET dictionary, an MR dictionary, and a standard-dose PET dictionary.
  • the initialization dictionary needs to be normalized, that is, normalized.
  • S22 constructs an initialization joint dictionary according to the initialization dictionary, and constructs a target matrix according to the second patch.
  • the expression to initialize the joint dictionary D and the target matrix Y is:
  • Dl is the low-dose PET dictionary
  • Dr is the MR dictionary
  • Ds is the standard-dose PET dictionary
  • Yl is the second patch of the low-dose PET image
  • Yr is the second patch of the MR image
  • Ys is the second patch of the standard-dose PET image. patch.
  • S23 iteratively updates the sparse coding and the dictionary according to the K-SVD idea until the iteration stop condition is satisfied.
  • Each iteration includes first fixing the dictionary to update the sparse coding, and then fixing the sparse coding to update the dictionary.
  • the sparse coding expressions for dictionary D and target matrix Y are:
  • X represents sparse coding
  • the MP Motion Pursuits
  • OMP Orthogonal Matching Pursuit
  • LASSO Least absolute shrinkage and selection operator
  • dictionary update expression for dictionary D and target matrix Y is:
  • yi is the ith element of Y
  • ⁇ i is the is a diagonal matrix of diagonal elements
  • y i is the ith element of Y
  • ⁇ represents the sparse constraint coefficient
  • k is the number of iterations
  • a q is The qth column of
  • b q is The qth column element of
  • a qq is Element at row q and column q.
  • This embodiment iteratively updates the sparse coding and the dictionary based on the K-SVD idea until the iteration stop condition is satisfied.
  • Each iteration includes first fixing the dictionary D to update the sparse coding X, and then fixing the sparse coding X to update the dictionary D.
  • the K-SVD idea is a classic dictionary training algorithm. According to the principle of minimum error, the error term is decomposed by SVD, and the decomposition term with the smallest error is selected as the updated dictionary atom and the corresponding atomic coefficient. After continuous iteration, it is optimized. solution.
  • an online learning method is used in this embodiment, that is, N samples are randomly selected for training in each iteration.
  • N samples are randomly selected for training in each iteration.
  • the training speed is greatly improved on the basis of ensuring the training accuracy, and the entire low-dose image restoration is shortened. time. Randomly selecting samples for training ensures that the training accuracy will not be different due to chance. Selecting some samples for training avoids repeated training of samples, greatly shortens the iteration time, and improves training efficiency.
  • S3 restores the low-dose PET image to the restored image of the standard-dose PET according to the joint dictionary.
  • the specific steps of S3 include: S31 performing block processing on the low-dose PET image and the MR image to obtain a third patch, and performing second preprocessing on the third patch to obtain a fourth patch; S32 , using the first low-dose PET dictionary obtained in S2 Merge with the first MR dictionary into a second joint dictionary, and obtain a second sparse code according to the fourth patch and the second joint dictionary; S33 obtain a prediction patch according to the first standard dose PET dictionary and the second sparse code obtained in S2, and use the prediction patch It is restored to a two-dimensional lattice to obtain a restored image of standard dose PET.
  • S31 performs block processing on the low-dose PET image and the MR image to obtain a third patch, and performs second preprocessing on the third patch to obtain a fourth patch.
  • the block method is the same as that of S1.
  • Image blocks are randomly selected from multiple frames of images and extended into a one-dimensional vector to obtain the third patch.
  • the third patch includes the third patch of the low dose PET image and the third patch of the MR image.
  • the selection position of the third patch needs to be in the order of the multiple frames of images, so that it can cover the whole frame of images.
  • there may be overlaps between the third patches which may reduce the blockiness of the result.
  • the third patch needs to be subjected to a second preprocessing.
  • the second preprocessing includes acquiring
  • the mapping matrix M maps the third patch of the low-dose PET image and the third patch of the MR image to the imaging space of the standard-dose PET image to obtain the mapped third patch, that is, the fourth patch.
  • the mapping matrix can make the mapping both accurate and general in a point-to-point, edge-to-edge, etc. manner.
  • S32 merges the first low-dose PET dictionary and the first MR dictionary obtained in S2 into a second joint dictionary, and obtains the corresponding second sparse code according to the second joint dictionary D and the fourth patch, and the specific expression is as follows:
  • D1 represents the first low-dose PET dictionary
  • Dr represents the first MR dictionary
  • Y1 represents the fourth patch of the low-dose PET image
  • Yr represents the fourth patch of the MR image.
  • S33 obtains the predicted patch according to the first standard dose PET dictionary and the second sparse coding, restores the predicted patch into a two-dimensional lattice, and obtains a restored image of the standard dose PET.
  • the corresponding prediction patch is predicted through the first standard dose PET dictionary obtained in S2 and the second sparse coding obtained, and the predicted patch is restored to a two-dimensional lattice to obtain the final restored image of the standard dose PET.
  • the final restored image of the standard dose PET is the restored image of the low dose PET image.
  • the method proposed in this embodiment has compatibility and can also be applied to other types of medical image reconstruction fields, such as CT images, etc.
  • the reconstruction effect of images can be improved by combining with deep learning related methods.
  • the method proposed in this embodiment has a strong advantage in noise reduction, and can also be applied to image denoising related fields.
  • This embodiment provides a low-dose PET image restoration method. Reduction of low-dose PET images via dictionary learning and sparse matrices. After the dictionary is constructed, the corresponding algorithm is used to update the dictionary to obtain a joint dictionary that is more suitable for restoring standard dose images, and at the same time, combining with MR images improves the accuracy of restoration. At the same time, the online learning related method is applied to the sparse dictionary update to speed up the convergence speed, which greatly shortens the time required for the entire image restoration process and improves the efficiency of image restoration.
  • this embodiment provides a low-dose PET image restoration system, which modularizes the low-dose PET image restoration method of Embodiment 1.
  • the specific plans are as follows:
  • a low-dose PET image restoration system includes: a sample acquisition unit, a joint dictionary acquisition unit, and an image restoration unit.
  • the joint dictionary acquisition unit is respectively connected to the sample acquisition unit and the image restoration unit.
  • the system is shown in Figure 3 of the specification.
  • the sample acquisition unit is configured to perform block processing on training images including low-dose PET images, MR images and standard-dose PET images to acquire a first patch, and perform first preprocessing on the first patch to acquire a second patch .
  • the user inputs the low-dose PET image as a sample through the sample acquisition unit, and the sample acquisition unit acquires the first patch according to the training image, performs mapping processing on the first patch, and uses the mapping matrix M to map the first patch of the low-dose PET image and the MR image.
  • the first patch is mapped to the imaging space of the standard dose PET image to obtain the mapped first patch, that is, the second patch. and pass the second patch to the joint dictionary acquisition unit.
  • the joint dictionary obtaining unit is configured to obtain the first joint dictionary according to sparse coding and dictionary update.
  • the joint dictionary acquisition unit mainly includes an initialization unit, a construction unit and an iterative unit.
  • the construction unit is respectively connected to the initialization unit and the iteration unit, as shown in FIG. 4 in the specification.
  • the initialization unit is configured to obtain an initialization dictionary including a low-dose PET dictionary, an MR dictionary, and a standard-dose PET dictionary by taking the second patch as a sample.
  • the initialization unit receives the second patch transmitted from the sample acquisition unit, adopts the K-means clustering algorithm, takes the second patch as a sample, obtains K cluster centers as an initialization dictionary, and normalizes the initialization dictionary.
  • the construction unit is used for constructing an initialization joint dictionary according to the initialization dictionary, and constructing a target matrix according to the second patch, wherein the initial joint dictionary includes a low-dose PET dictionary, an MR dictionary and a standard-dose PET dictionary, and the target matrix includes the second patch of the low-dose PET image, The second patch of the MR image and the second patch of the standard dose PET image; the iterative unit is used to obtain the sparse code according to the target matrix, and iteratively updates the sparse code and the initial joint dictionary according to the K-SVD idea until the iterative stop condition is met, and obtain A first combined dictionary including a first low-dose PET dictionary, a first MR dictionary, and a first standard-dose PET dictionary.
  • the iterative unit further includes an iterative update unit and a sample selection unit.
  • the iterative update unit is used to first fix the dictionary to update the sparse coding in each iteration, and then fix the sparse coding to update the dictionary.
  • the sample selection unit is used to randomly select some samples for sparse coding and dictionary update in each iteration.
  • the iterative unit is the core processing unit, which uses the K-SVD idea to iteratively update the sparse coding and dictionary to ensure the accuracy of image restoration.
  • the iteration unit is also provided with a sample selection unit, which is used to randomly select some samples for sparse coding and dictionary update in each iteration, so as to improve the training efficiency while ensuring the training accuracy.
  • the image restoration unit is configured to restore the low-dose PET image into a restored image of the standard-dose PET according to the first joint dictionary acquired by the joint dictionary acquisition unit.
  • the image restoration unit is the final restoration module.
  • the image restoration unit performs block processing on the low-dose PET image and the MR image to obtain the third patch, and then maps the third patch of the low-dose PET image and the third patch of the MR image to the standard-dose PET image through the mapping matrix M
  • the imaging space of obtains the third patch after mapping, that is, the fourth patch.
  • the low-dose PET dictionary and the MR dictionary obtained in the joint dictionary obtaining unit are combined into a joint dictionary, and the fourth patch is combined to obtain the second sparse code.
  • the corresponding prediction patch is predicted through the standard dose PET dictionary and the obtained second sparse coding, and the predicted patch is restored to a two-dimensional lattice to obtain the final restored image of the standard dose PET.
  • the final restored image of the standard dose PET is the restored image of the low dose PET image.
  • this embodiment proposes a low-dose PET image restoration system.
  • the method of Embodiment 1 can effectively solve the problems of serious noise and loss of details in low-dose PET images. At the same time, it can improve the efficiency of image restoration and improve the accuracy of restored images.
  • FIG. 5 is a schematic structural diagram of a computer device according to Embodiment 3 of the present invention.
  • the computer device 12 shown in FIG. 5 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present invention.
  • computer device 12 takes the form of a general-purpose computing device.
  • Components of computer device 12 may include, but are not limited to, one or more processors or processing units 16 , system memory 28 , and a bus 18 connecting various system components including system memory 28 and processing unit 16 .
  • Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by device computer 12, including both volatile and non-volatile media, removable and non-removable media.
  • System memory 28 may include computer system readable media in the form of volatile memory.
  • Computer device 12 may also communicate with one or more external devices 14 (eg, keyboards, pointing devices, displays, etc.), may also communicate with one or more devices that enable a user to interact with computer device 12, and/or communicate with The computer device 12 is capable of communicating with any device that communicates with one or more other computing devices.
  • external devices 14 eg, keyboards, pointing devices, displays, etc.
  • the computer device 12 is capable of communicating with any device that communicates with one or more other computing devices.
  • the processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, for example, to implement a low-dose PET image restoration method provided in Embodiment 1 of the present invention, and the method includes:
  • S2 specifically includes: S21, taking the second patch as a sample to obtain an initialization dictionary including a low-dose PET dictionary, an MR dictionary and a standard-dose PET dictionary; S22, constructing an initialization joint dictionary according to the initialization dictionary, and constructing a target according to the second patch Matrix; S23. Obtain the sparse code according to the target matrix, and iteratively update the sparse code and the initialized joint dictionary to meet the iterative stop condition, and obtain the dictionary including the first low-dose PET dictionary, the first MR dictionary and the first standard dose PET dictionary.
  • the first union dictionary of wherein, each iteration includes first fixing the dictionary to update the sparse coding, and then fixing the sparse coding to update the dictionary, and each iteration randomly selects some samples for sparse coding and dictionary update.
  • S3 specifically includes: S31, performing block processing on the low-dose PET image and the MR image to obtain a third patch, and performing second preprocessing on the third patch to obtain a fourth patch; S32, processing the first low-dose obtained in S2.
  • the PET dictionary and the first MR dictionary are combined into a second joint dictionary, and a second sparse code is obtained according to the fourth patch and the second joint dictionary; S33, a prediction patch is obtained according to the first standard dose PET dictionary and the second sparse code obtained in S2, The predicted patch is restored to a two-dimensional lattice to obtain a restored image of standard dose PET.
  • a low-dose PET image restoration method is applied to a specific computer device, and the method is stored in the memory.
  • the actuator executes the memory, the method will be run to restore the low-dose PET image, which is fast and convenient to use. ,Wide range of applications.
  • processor can also implement the technical solution of the PET image restoration method provided by any embodiment of the present invention.
  • Embodiment 4 provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the steps of the low-dose PET image restoration method provided by any embodiment of the present invention, and the method includes:
  • S2 specifically includes: S21, taking the second patch as a sample to obtain an initialization dictionary including a low-dose PET dictionary, an MR dictionary and a standard-dose PET dictionary; S22, constructing an initialization joint dictionary according to the initialization dictionary, and constructing a target according to the second patch Matrix; S23. Obtain the sparse code according to the target matrix, and iteratively update the sparse code and the initialized joint dictionary to meet the iterative stop condition, and obtain the dictionary including the first low-dose PET dictionary, the first MR dictionary and the first standard dose PET dictionary.
  • the first union dictionary of wherein, each iteration includes first fixing the dictionary to update the sparse coding, and then fixing the sparse coding to update the dictionary, and each iteration randomly selects some samples for sparse coding and dictionary update.
  • S3 specifically includes: S31, performing block processing on the low-dose PET image and the MR image to obtain a third patch, and performing second preprocessing on the third patch to obtain a fourth patch; S32, processing the first low-dose obtained in S2.
  • the PET dictionary and the first MR dictionary are combined into a second joint dictionary, and a second sparse code is obtained according to the fourth patch and the second joint dictionary; S33, a prediction patch is obtained according to the first standard dose PET dictionary and the second sparse code obtained in S2, The predicted patch is restored to a two-dimensional lattice to obtain a restored image of standard dose PET.
  • the computer storage medium of this embodiment may adopt any combination of one or more computer-readable media.
  • 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 any combination 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.
  • 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.
  • This embodiment applies a low-dose PET image restoration method to a computer-readable storage medium, which stores a computer program, and when the program is executed by a processor, implements the steps of the low-dose PET image restoration method provided by the present invention, Simple and fast, easy to store, not easy to lose.
  • low-dose PET images generally have problems such as poor restoration effect, complex restoration process, and low restoration accuracy, resulting in poor restoration image quality, lengthy restoration time, and lack of restoration image content. Therefore, there is a need for a restoration method for low-dose PET images, which can solve the problems of serious noise and loss of details, and at the same time can improve the efficiency of image restoration and the accuracy of restored images.
  • the present invention proposes a low-dose PET image restoration method, system, equipment and medium, which solves the common defects in the prior art such as poor image restoration effect, complex image restoration process, and low image restoration accuracy, and can solve the problem of low-dose PET image restoration.
  • the problem of serious noise and loss of details in PET images can improve the efficiency of image restoration and the accuracy of restored images.
  • dictionary learning and sparse matrix to restore standard dose PET images from low-dose PET images, it overcomes the shortcomings of traditional denoising methods that cannot preserve details; online learning related concepts are adopted, and small training samples are randomly obtained during the learning process. Compared with the traditional technology, the convergence speed is accelerated while the accuracy is guaranteed.
  • the application of the low-dose PET reduction method to specific systems, computer equipment and computer storage media, and the realization of the method is of great significance to the development of the medical impact field.
  • modules or steps of the present invention can be implemented by a general-purpose computing device, and they can be centralized on a single computing device, or distributed on a network composed of multiple computing devices.
  • they may be implemented in program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be fabricated separately into individual integrated circuit modules, or a plurality of modules of them Or the steps are made into a single integrated circuit module to realize.
  • the present invention is not limited to any specific combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Image Processing (AREA)
  • Nuclear Medicine (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

本发明提供一种低剂量PET图像还原方法、系统、设备和介质。其中，方法包括：S1、对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对第一补丁进行第一预处理获取第二补丁；S2、根据第二补丁，利用稀疏编码和字典更新获取第一联合字典；S3、根据第一联合字典将低剂量PET图像还原成标准剂量PET的还原图像。本发明解决了现有技术中低剂量PET图像还原效果差、图像还原过程复杂，图像还原准确度低等缺陷，能够解决低剂量PET图像噪声严重和细节丢失的问题，同时提高图像还原效率和图像还原准确度。

Description

一种低剂量PET图像还原方法、系统、设备和介质 技术领域
本发明涉及PET图像处理领域，具体而言，涉及一种低剂量PET图像还原方法、系统、设备和介质。
背景技术
正电子发射计算机断层显像(positronemissiontomography，PET)与核磁共振成像(magneticresonanceimaging，MRI)相结合的PET/MR大型功能代谢与分子影像诊断设备，同时具有PET和MR的检查功能，具有灵敏度搞、准确性好、辐射量少等优势，在诸多疾病的诊断尤其是肿瘤和心脑血管等疾病的诊断上效果显著。PET显影剂量范围为4.43-7.35mSv，平均5.89mSv，标准剂量的PET显影剂仍然对人体存在一定量的辐射，在累计效应下会增加各种疾病发生的可能性，影响人体健康。因此，合理降低PET的剂量，使其低于PET显影剂量范围，可减少对人体辐射的影响。实现从低剂量PET图像到标准剂量PET图像的还原，对PET/MR技术的应用和医疗诊断具有重要的意义。
低剂量PET图像由于在成像时降低了显影剂剂量，导致图像成像后存在大量噪声，同时会丢失图像细节，严重影响PET/MR设备对患者病灶的影像诊断效果。此外，PET图像由于分辨率较高，导致图像还原过程计算量大，还原过程复杂，还原时间长，严重影响诊断效率。
现有技术针对低剂量PET图像的还原普遍存在还原效果不好、还原过程复杂，还原准确度不高等问题，导致还原图像质量差、还原时间冗长、还原图像内容缺失等问题。因此，需要一种针对低剂量PET图像的还原方法，能够解决其噪声严重和细节丢失的问题，同时能提高图像还原的效率， 提高还原图像的准确度。
发明内容
基于现有技术存在的问题，本发明提供了一种低剂量PET图像还原方法，具体方案如下：
一种低剂量PET图像还原方法，包括如下步骤：S1、对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对所述第一补丁进行第一预处理获取第二补丁；S2、根据所述第二补丁，利用稀疏编码和字典更新获取第一联合字典；S3、根据所述第一联合字典将所述低剂量PET图像还原成标准剂量PET的还原图像。
进一步，在所述S2中还包括如下步骤：S21、以所述第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典；S22、根据所述初始化字典构建初始化联合字典，根据所述第二补丁构建目标矩阵；S23、根据所述目标矩阵获取稀疏编码，对所述稀疏编码和所述初始化联合字典进行迭代更新至满足迭代停止条件后，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的第一联合字典。
更进一步，在所述S23中，每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典。
更进一步，在所述S23中，每次迭代通过随机选取部分样本用于稀疏编码和字典更新。
进一步，在所述S3中，还包括以下步骤：S31、对低剂量PET图像和MR图像进行分块处理获取第三补丁，并对所述第三补丁进行第二预处理获取第四补丁；S32、将所述S2获取的所述第一低剂量PET字典和所述第一MR字典合并成第二联合字典，根据所述第四补丁和所述第二联合字典获取第二稀疏编码；S33、根据所述S2获取的所述第一标准剂量PET字典和所 述第二稀疏编码获取预测补丁，将所述预测补丁还原成二维点阵，得到标准剂量PET的还原图像。
特别地，所述第一补丁为从多帧图像中随机选取图像块并延展成的一维向量，包括所述低剂量PET图像的第一补丁、所述MR图像的第一补丁和所述标准剂量PET图像的第一补丁；所述第一补丁在多帧图像中处于同一位置。
特别地，所述第三补丁在选取位置时，按照多帧图像的顺序覆盖整个一帧图像。
特别地，在所述S1中，所述第一预处理包括通过预设矩阵将所述低剂量PET图像的第一补丁和所述MR图像的第一补丁映射到所述标准剂量PET图像的成像空间得到所述低剂量PET图像的第二补丁和所述MR图像的第二补丁。
特别地，在所述S31中，所述第二预处理包括通过预设矩阵将所述低剂量PET图像的第三补丁和所述MR图像的第三补丁映射到所述标准剂量PET图像的成像空间得到所述低剂量PET图像的第四补丁和所述MR图像的第四补丁。
特别地，在所述S21中包括采用K-means聚类算法，以所述第二补丁为样本，获取K个聚类中心作为初始化字典，并对所述初始化字典做归一化处理。
更进一步，在所述S22中，所述初始化联合字典和所述目标矩阵的表达式分别为：
Figure PCTCN2020134621-appb-000001
其中，D表示初始化联合字典，Y表示目标矩阵，Dl表示低剂量PET字典，Dr表示MR字典，Ds表示标准剂量PET字典，Yl表示低剂量PET图像的第二补丁，Yr表示MR图像的第二补丁，Ys表示标准剂量PET图 像的第二补丁。
更进一步，在所述S23中，所述稀疏编码表达式包括：
Figure PCTCN2020134621-appb-000002
其中，X表示稀疏编码，Λ是一个对角矩阵，其对角元素为Λ q＝d q-y i，这里d q为字典D的第q个原子，y i为Y的第i个元素，λ表示稀疏约束系数。
更进一步，在所述S23中，所述字典更新表达式包括：
Figure PCTCN2020134621-appb-000003
其中，y i为Y的第i个元素，
Figure PCTCN2020134621-appb-000004
为样本x各元素的绝对值，ψ i为以
Figure PCTCN2020134621-appb-000005
为对角元素的对角矩阵，y i为Y的第i个元素，μ表示稀疏约束系数。
特别地，通过梯度下降法求解所述字典更新表达式，所述梯度下降法表达式为：
Figure PCTCN2020134621-appb-000006
其中，d q为字典D的第q个原子，k为迭代次数，a q为
Figure PCTCN2020134621-appb-000007
的第q列元素，b q为
Figure PCTCN2020134621-appb-000008
的第q列元素，a qq为
Figure PCTCN2020134621-appb-000009
第q行q列元素。
一种低剂量PET图像还原系统，包括：样本获取单元，用于对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对所述第一补丁进行第一预处理获取第二补丁；联合字典获取单元，用于根据所述第二补丁，通过稀疏编码和字典更新获取第一联合字典；图像还原单元，用于根据所述第一联合字典将所述低剂量PET图像还原成标准剂量PET的还原图像。
进一步，所述联合字典获取单元还包括：初始化单元：用于以所述第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在 内的初始化字典；构建单元：用于根据所述初始化字典构建初始化联合字典，根据所述第二补丁构建目标矩阵；迭代单元：用于根据所述目标矩阵获取稀疏编码，对所述稀疏编码和所述初始化联合字典进行迭代更新至满足迭代停止条件后，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的所述第一联合字典。
特别地，所述迭代单元还包括：迭代更新单元：用于在每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典；样本选取单元：用于在每次迭代时随机选取部分样本进行稀疏编码和字典更新。
一种计算机设备，所述计算机设备包括：一个或多个处理器；存储器，用于存储一个或多个程序；当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如上述所述的低剂量PET图像还原方法。
一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现如上述所述的低剂量PET图像还原方法。
本发明具有如下有益效果：
针对低剂量PET图像噪声严重和细节丢失的问题，本发明提出一种低剂量PET图像还原方法、系统、设备和介质，解决了现有技术普遍存在的图像还原效果差、图像还原过程复杂，图像还原准确度低等缺陷，能够有效解决低剂量PET图像噪声严重和细节丢失的问题，同时能提高图像还原的效率，提高还原图像的准确度。将低剂量PET还原方法应用到具体的系统、计算机设备和计算机存储介质中，将该方法具体化，对医学影响领域的发展具有重要意义。
为使本发明的上述目的、特征和优点能更明显易懂，下文特举较佳实施例，并配合所附附图，作详细说明如下。
附图说明
为了更清楚地说明本发明实施例的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，应当理解，以下附图仅示出了本发明的某些实施例，因此不应被看作是对范围的限定，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他相关的附图。
图1是本发明的低剂量PET图像还原方法流程图；
图2是本发明的低剂量PET图像还原方法的S2具体流程图；
图3是本发明的低剂量PET图像还原系统框图；
图4是本发明的低剂量PET图像还原系统的具体框图；
图5是本发明的低剂量PET图像还原方法应用到一种计算机设备的示意图。
具体实施方式
实施例1
针对低剂量PET图像噪声严重和细节丢失的问题，本实施例提出了一种低剂量PET图像还原方法，具体的方法步骤如说明书附图1所示，具体方案如下：
S1、图像预处理：针对包括低剂量PET图像、MR图像和标准剂量PET图像的训练图像进行分块处理获取第一补丁，并对第一补丁进行第一预处理获取第二补丁；
S2、获取第一联合字典：根据第二补丁，利用稀疏编码和字典更新获取第一联合字典；
S3、获取还原图像：根据第一联合字典将低剂量PET图像还原成标准剂量PET的还原图像。
具体地，S1是从训练样本中获取第一补丁，并对第一补丁进行第一预处理获取第二补丁。本实施例针对的训练图像包括低剂量PET图像、MR 图像和标准剂量PET图像，其中，低剂量PET图像为待还原图像。PET图像为一组连续的人体断层图像，多个单帧图像构成多帧图像。本实施例是从训练图像中先获取第一补丁，第一补丁是从多帧图像中随机选取图像块并延展成的一维向量。由于第一补丁是在多帧图像中随机选取的，会存在重复，因此需要将重复的部分去除，再作为样本参与训练。特别地，低剂量PET图像的第一补丁、MR图像的第一补丁和标准剂量PET图像的第一补丁是在多帧图像上的同一个位置取得的。由于低剂量PET图像、MR图像的成像空间与标准剂量PET图像的成像空间不同，不能直接用于后续的字典学习，需要对第一补丁进行第一预处理，第一预处理包括获取映射矩阵M将低剂量PET图像的第一补丁和MR图像的第一补丁映射到标准剂量PET图像的成像空间，得到映射后的第一补丁，即第二补丁。映射矩阵在可以按照点对点、边对边等方式使得映射兼具准确性和一般性。具体的映射矩阵表达为：
f(Y)＝M*Y
其中，Y是第一补丁向量矩阵，M是映射矩阵。
具体地，S2主要获取第一联合字典。S2包含两个主要模块：稀疏编码和字典更新，通过稀疏编码和字典更新交替进行的方法实现字典的获取。S2具体包括如下步骤：S21、获取初始化字典：以第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典；S22、构建初始化字典和目标矩阵：根据初始化字典构建初始化联合字典，根据第二补丁构建目标矩阵；S23、获取第一联合字典：根据目标矩阵获取稀疏编码，根据K-SVD思想对稀疏编码和初始化联合字典进行迭代更新，直到满足迭代停止条件，每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典。获取的第一联合字典包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典。S2具体步骤如说明书附图2所示。优选地，本实施例针对字典学习采用Local Coordinate Coding(局部坐标编码)，具 体表达式为：
Figure PCTCN2020134621-appb-000010
其中，X表示稀疏编码，D表示特征矩阵(字典)，x i表示第i个样本的稀疏系数，d表示特征字典的特征维度，q表示字典的第q列，字典学习的主要目的是获取求取D。
S21以第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典。具体包括，通过K-means聚类算法，以第二补丁作为样本，获取K个聚类中心作为初始化字典，初始化字典包括低剂量PET字典、MR字典和标准剂量PET字典。此外，需对初始化字典进行normalize处理，即归一化处理。
S22根据初始化字典构建初始化联合字典，根据第二补丁构建目标矩阵。初始化联合字典D和目标矩阵Y的表达式为：
Figure PCTCN2020134621-appb-000011
其中，Dl表示低剂量PET字典，Dr表示MR字典，Ds表示标准剂量PET字典，Yl表示低剂量PET图像的第二补丁，Yr表示MR图像的第二补丁，Ys表示标准剂量PET图像的第二补丁。
S23根据K-SVD思想对稀疏编码和字典进行迭代更新，直到满足迭代停止条件，每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典。
针对稀疏编码部分，关于字典D和目标矩阵Y的稀疏编码表达式为：
Figure PCTCN2020134621-appb-000012
其中，X表示稀疏编码，Λ是一个对角矩阵，其对角元素为Λ q＝d q-y i，这里d q为D的第q个原子，y i为Y的第i个元素，λ表示稀疏约束系数。针对上式，本实施例选择用MP(Matching Pursuits)算法、OMP(Orthogonal  Matching Pursuit)算法或LASSO(Least absolute shrinkage and selection operator)算法进行求解。
针对字典更新部分，关于字典D和目标矩阵Y的字典更新表达式为：
Figure PCTCN2020134621-appb-000013
其中，y i为Y的第i个元素，
Figure PCTCN2020134621-appb-000014
为样本x各元素的绝对值，ψ i为以
Figure PCTCN2020134621-appb-000015
为对角元素的对角矩阵，y i为Y的第i个元素，μ表示稀疏约束系数。针对上式本实施例采用梯度下降法进行求解，具体表达式如下：
Figure PCTCN2020134621-appb-000016
其中，k为迭代次数，a q为
Figure PCTCN2020134621-appb-000017
的第q列，b q为
Figure PCTCN2020134621-appb-000018
Figure PCTCN2020134621-appb-000019
的第q列元素，a qq为
Figure PCTCN2020134621-appb-000020
第q行q列元素。
本实施例基于K-SVD思想对稀疏编码和字典进行迭代更新，直到满足迭代停止条件，每次迭代包括先固定字典D更新稀疏编码X，再固定稀疏编码X更新字典D。K-SVD思想是一种经典的字典训练算法，依据误差最小原则，对误差项进行SVD分解，选择使误差最小的分解项作为更新的字典原子和对应的原子系数，经过不断的迭代从而得到优化的解。
特别地，本实施例选用在线学习的方式，即每次迭代时随机选取N个样本进行训练。与现有技术中将所有训练样本同时进行迭代的方式不同，本实施例通过随机选取部分样本进行迭代，在保证训练精度的基础上极大的提高了训练速度，进而缩短了整个低剂量图像还原的时间。随机选取样本进行训练，保证了训练精度不会因为偶然性而产生差异，选取部分样本进行训练，避免了样本的重复训练，大大缩短了迭代的时间，提高了训练效率。
具体地，S3根据联合字典将低剂量PET图像还原成标准剂量PET的 还原图像。S3的具体步骤包括：S31对低剂量PET图像和MR图像进行分块处理获取第三补丁，并对第三补丁进行第二预处理获取第四补丁；S32将S2获取的第一低剂量PET字典和第一MR字典合并成第二联合字典，根据第四补丁和第二联合字典获取第二稀疏编码；S33根据S2获取的第一标准剂量PET字典和第二稀疏编码获取预测补丁，将预测补丁还原成二维点阵，得到标准剂量PET的还原图像。
其中，S31对低剂量PET图像和MR图像进行分块处理获取第三补丁，并对第三补丁进行第二预处理获取第四补丁。分块方法与S1相同，从多帧图像中随机选取图像块并延展成一维向量，得到第三补丁。第三补丁包括低剂量PET图像的第三补丁和MR图像的第三补丁。特别地，第三补丁的选取位置需要按照多帧图像的排列顺序，使其能够覆盖整个一帧图像。特别地，第三补丁之间可以有重叠部分，可减小结果的块效应。同样地，由于低剂量PET图像、MR图像的成像空间与标准剂量PET图像的成像空间不同，不能直接用于后续的字典学习，需要对第三补丁进行第二预处理，第二预处理包括获取映射矩阵M将低剂量PET图像的第三补丁和MR图像的第三补丁映射到标准剂量PET图像的成像空间，得到映射后的第三补丁，即第四补丁。映射矩阵在可以按照点对点、边对边等方式使得映射兼具准确性和一般性。
其中，S32将S2中获取的第一低剂量PET字典和第一MR字典合并成第二联合字典，根据第二联合字典D和第四补丁获取相应的第二稀疏编码，具体表达式如下：
Figure PCTCN2020134621-appb-000021
Dl表示第一低剂量PET字典，Dr表示第一MR字典，Yl表示低剂量PET图像的第四补丁，Yr表示MR图像的第四补丁。
最终，S33根据第一标准剂量PET字典和第二稀疏编码获取预测补丁， 将预测补丁还原成二维点阵，得到标准剂量PET的还原图像。通过S2获得的第一标准剂量PET字典和得到的第二稀疏编码预测出相应的预测补丁，将预测补丁重新还原成二维点阵，得到最终的标准剂量PET的还原图像。最终的标准剂量PET的还原图像即为低剂量PET图像的还原图像。
本实施例提出的方法具有兼容性还可应用于其它类型的医学图像重建领域，如CT图像等，结合深度学习相关方法能提高图像的重建效果。此外，本实施例提出的方法在降噪方面具有极强的优势，还可应用于图像去噪相关领域。
本实施例提供了一种低剂量PET图像还原方法。通过字典学习和稀疏矩阵对低剂量PET图像进行还原。在构建字典之后，采用相应算法对字典进行更新，得到更适用于还原标准剂量图像的联合字典，同时结合MR图像提高还原的准确度。同时采用在线学习相关方法应用在稀疏字典更新以加快收敛速度，使得整个图像还原过程所需的时间大大缩短，提高了图像还原的效率。
实施例2
在实施例1的基础上，本实施例提供了一种低剂量PET图像还原系统，将实施例1的低剂量PET图像还原方法模块化。具体方案如下：
一种低剂量PET图像还原系统，包括：样本获取单元、联合字典获取单元和图像还原单元，联合字典获取单元分别连接样本获取单元和图像还原单元，系统如说明书附图3所示。
具体地，样本获取单元用于对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对第一补丁进行第一预处理获取第二补丁。用户通过样本获取单元输入低剂量PET图像作为样本，样本获取单元根据训练图像获取第一补丁，并对第一补丁进行映射处理，通过映射矩阵M将低剂量PET图像的第一补丁和MR图像的 第一补丁映射到标准剂量PET图像的成像空间，得到映射后的第一补丁，即第二补丁。并将第二补丁传递到联合字典获取单元。
具体地，联合字典获取单元用于根据稀疏编码和字典更新获取第一联合字典。联合字典获取单元主要包括初始化单元、构建单元和迭代单元。构建单元分别连接初始化单元和迭代单元，具体如说明书附图4所示。
具体地，初始化单元用于以第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典。初始化单元接收来自样本获取单元传递的第二补丁，采用K-means聚类算法，以第二补丁为样本，获取K个聚类中心作为初始化字典，并对初始化字典做归一化处理。构建单元用于根据初始化字典构建初始化联合字典，根据第二补丁构建目标矩阵，其中初始化联合字典包括低剂量PET字典、MR字典和标准剂量PET字典，目标矩阵包括低剂量PET图像的第二补丁、MR图像的第二补丁和标准剂量PET图像的第二补丁；迭代单元用于根据目标矩阵获取稀疏编码，根据K-SVD思想对稀疏编码和初始化联合字典进行迭代更新，直到满足迭代停止条件，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的第一联合字典。迭代单元还包括迭代更新单元和样本选取单元。迭代更新单元用于在每次迭代时先固定字典更新稀疏编码，再固定稀疏编码更新字典。样本选取单元用于在每次迭代时随机选取部分样本进行稀疏编码和字典更新。迭代单元是核心的处理单元，采用K-SVD思想对稀疏编码和字典进行迭代更新，保证图像还原的准确性。此外，迭代单元还设置有样本选取单元，用于在每次迭代时随机选取部分样本进行稀疏编码和字典更新，在保障训练精度的同时提高训练效率。
具体地，图像还原单元用于根据联合字典获取单元获取的第一联合字典将低剂量PET图像还原成标准剂量PET的还原图像。图像还原单元是最后的还原模块。首先，图像还原单元对低剂量PET图像和MR图像进行分块处理，获取第三补丁，再通过映射矩阵M将低剂量PET图像的第三补丁 和MR图像的第三补丁映射到标准剂量PET图像的成像空间，得到映射后的第三补丁，即第四补丁。其次，用联合字典获取单元中获取的低剂量PET字典和MR字典合并成联合字典，结合第四补丁，获取第二稀疏编码。最后，通过标准剂量PET字典和得到的第二稀疏编码预测出相应的预测补丁，将预测补丁重新还原成二维点阵，得到最终的标准剂量PET的还原图像。最终的标准剂量PET的还原图像即为低剂量PET图像的还原图像。
本实施例根据实施例1提出的一种低剂量PET图像还原方法，提出了一种低剂量PET图像还原系统，采用实施例1的方法，能有效解决低剂量PET图像噪声严重和细节丢失的问题，同时能提高图像还原的效率，提高还原图像的准确度。
实施例3
图5为本发明实施例3提供的一种计算机设备的结构示意图。图5显示的计算机设备12仅仅是一个示例，不应对本发明实施例的功能和使用范围带来任何限制。
如图5所示，计算机设备12以通用计算设备的形式表现。计算机设备12的组件可以包括但不限于：一个或者多个处理器或者处理单元16，系统存储器28，连接不同系统组件(包括系统存储器28和处理单元16)的总线18。计算机设备12典型地包括多种计算机系统可读介质。这些介质可以是任何能够被设备计算机12访问的可用介质，包括易失性和非易失性介质，可移动的和不可移动的介质。系统存储器28可以包括易失性存储器形式的计算机系统可读介质。
计算机设备12也可以与一个或多个外部设备14(例如键盘、指向设备、显示器等)通信，还可与一个或者多个使得用户能与该计算机设备12交互的设备通信，和/或与使得该计算机设备12能与一个或多个其它计算设备进行通信的任何设备通信。
处理单元16通过运行存储在系统存储器28中的程序，从而执行各种功能应用以及数据处理，例如实现本发明实施例1所提供的一种低剂量PET图像还原方法，该方法包括：
S1、针对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对第一补丁进行第一预处理获取第二补丁；S2、根据第二补丁，利用稀疏编码和字典更新获取第一联合字典；S3、根据第一联合字典将低剂量PET图像还原成标准剂量PET的还原图像。
其中，S2具体包括：S21、以第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典；S22、根据初始化字典构建初始化联合字典，根据第二补丁构建目标矩阵；S23、根据目标矩阵获取稀疏编码，对稀疏编码和初始化联合字典进行迭代更新至满足迭代停止条件后，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的第一联合字典。其中，每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典，每次迭代通过随机选取部分样本用于稀疏编码和字典更新。
其中，S3具体包括：S31、对低剂量PET图像和MR图像进行分块处理获取第三补丁，并对第三补丁进行第二预处理获取第四补丁；S32、将S2获取的第一低剂量PET字典和第一MR字典合并成第二联合字典，根据第四补丁和第二联合字典获取第二稀疏编码；S33、根据S2获取的第一标准剂量PET字典和第二稀疏编码获取预测补丁，将预测补丁还原成二维点阵，得到标准剂量PET的还原图像。
本实施例将一种低剂量PET图像还原方法应用到具体的计算机设备中，将该方法存储到存储器中，当执行器执行该存储器时，会运行该方法进行低剂量PET图像还原，使用快捷方便，适用范围广。
当然，本领域技术人员可以理解，处理器还可以实现本发明任意实施 例所提供的PET图像还原方法的技术方案。
实施例4
本实施例4提供了一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现如本发明任意实施例所提供的低剂量PET图像还原方法步骤，该方法包括：
S1、针对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对第一补丁进行第一预处理获取第二补丁；S2、根据第二补丁，利用稀疏编码和字典更新获取第一联合字典；S3、根据第一联合字典将低剂量PET图像还原成标准剂量PET的还原图像。
其中，S2具体包括：S21、以第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典；S22、根据初始化字典构建初始化联合字典，根据第二补丁构建目标矩阵；S23、根据目标矩阵获取稀疏编码，对稀疏编码和初始化联合字典进行迭代更新至满足迭代停止条件后，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的第一联合字典。其中，每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典，每次迭代通过随机选取部分样本用于稀疏编码和字典更新。
其中，S3具体包括：S31、对低剂量PET图像和MR图像进行分块处理获取第三补丁，并对第三补丁进行第二预处理获取第四补丁；S32、将S2获取的第一低剂量PET字典和第一MR字典合并成第二联合字典，根据第四补丁和第二联合字典获取第二稀疏编码；S33、根据S2获取的第一标准剂量PET字典和第二稀疏编码获取预测补丁，将预测补丁还原成二维点阵，得到标准剂量PET的还原图像。
本实施例的计算机存储介质，可以采用一个或多个计算机可读的介质 的任意组合。计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质。计算机可读存储介质例如可以是但不限于：电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本文件中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。
本实施例将一种低剂量PET图像还原方法应用到一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现本发明提供的低剂量PET图像还原方法的步骤，简便快捷，易于存储，不易丢失。
现有技术针对低剂量PET图像普遍存在还原效果不好、还原过程复杂，还原准确度不高等问题，导致还原图像质量差、还原时间冗长、还原图像内容缺失等问题。因此，需要一种针对低剂量PET图像的还原方法，能够解决其噪声严重和细节丢失的问题，同时能提高图像还原的效率，提高还原图像的准确度。
综上，本发明提出一种低剂量PET图像还原方法、系统、设备和介质，解决了现有技术普遍存在图像还原效果差、图像还原过程复杂，图像还原准确度低等缺陷，能够解决低剂量PET图像噪声严重和细节丢失的问题，同时能提高图像还原的效率，提高还原图像的准确度。通过于字典学习和稀疏矩阵实现从低剂量PET图像还原标准剂量PET图像，克服了传统方法去噪无法保留细节的缺点；采用在线学习相关概念，在学习过程中随机获取较小的训练样本，相较于传统技术，在保证准确率的情况下加快了收敛 速度。将低剂量PET还原方法应用到具体的系统、计算机设备和计算机存储介质中，将该方法具体化，对医学影响领域的发展具有重要意义。
本领域普通技术人员应该明白，上述的本发明的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个计算装置上，或者分布在多个计算装置所组成的网络上，可选地，他们可以用计算机装置可执行的程序代码来实现，从而可以将它们存储在存储装置中由计算装置来执行，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本发明不限制于任何特定的硬件和软件的结合。
注意，上述仅为本发明的较佳实施例及所运用技术原理。本领域技术人员会理解，本发明不限于这里的特定实施例，对本领域技术人员来说能够进行各种明显的变化、重新调整和替代而不会脱离本发明的保护范围。因此，虽然通过以上实施例对本发明进行了较为详细的说明，但是本发明不仅仅限于以上实施例，在不脱离本发明构思的情况下，还可以包括更多其他等效实施例，而本发明的范围由所附的权利要求范围决定。
以上公开的仅为本发明的几个具体实施场景，但是，本发明并非局限于此，任何本领域的技术人员能思之的变化都应落入本发明的保护范围。

Claims (19)

  1. 一种低剂量PET图像还原方法，其特征在于，包括如下步骤：
    S1、对包括低剂量PET图像、MR图像和标准剂量PET图像在内的训练图像进行分块处理获取第一补丁，并对所述第一补丁进行第一预处理获取第二补丁；
    S2、根据所述第二补丁，利用稀疏编码和字典更新获取第一联合字典；
    S3、根据所述第一联合字典将所述低剂量PET图像还原成标准剂量PET的还原图像。
  2. 根据权利要求1所述的方法，其特征在于，在所述S2中还包括如下步骤：
    S21、以所述第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典；
    S22、根据所述初始化字典构建初始化联合字典，根据所述第二补丁构建目标矩阵；
    S23、根据所述目标矩阵获取稀疏编码，对所述稀疏编码和所述初始化联合字典进行迭代更新至满足迭代停止条件后，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的第一联合字典。
  3. 根据权利要求2所述的方法，其特征在于，在所述S23中，每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典。
  4. 根据权利要求2或3所述的方法，其特征在于，在所述S23中，每次迭代通过随机选取部分样本用于稀疏编码和字典更新。
  5. 根据权利要求2所述的方法，其特征在于，在所述S3中，还包括以下步骤：
    S31、对低剂量PET图像和MR图像进行分块处理获取第三补丁，并对所述第三补丁进行第二预处理获取第四补丁；
    S32、将所述S2获取的所述第一低剂量PET字典和所述第一MR字典合并成第二联合字典，根据所述第四补丁和所述第二联合字典获取第二稀疏编码；
    S33、根据所述S2获取的所述第一标准剂量PET字典和所述第二稀疏编码获取预测补丁，将所述预测补丁还原成二维点阵，得到标准剂量PET的还原图像。
  6. 根据权利要求1所述的方法，其特征在于，所述第一补丁为从多帧图像中随机选取图像块并延展成的一维向量，包括所述低剂量PET图像的第一补丁、所述MR图像的第一补丁和所述标准剂量PET图像的第一补丁；
    所述第一补丁在多帧图像中处于同一位置。
  7. 根据权利要求5所述的方法，其特征在于，所述第三补丁在选取位置时，按照多帧图像的顺序覆盖整个一帧图像。
  8. 根据权利要求1所述的方法，其特征在于，在所述S1中，所述第一预处理包括通过预设矩阵将所述低剂量PET图像的第一补丁和所述MR图像的第一补丁映射到所述标准剂量PET图像的成像空间得到所述低剂量PET图像的第二补丁和所述MR图像的第二补丁。
  9. 根据权利要求5所述的方法，其特征在于，在所述S31中，所述第二预处理包括通过预设矩阵将所述低剂量PET图像的第三补丁和所述MR图像的第三补丁映射到所述标准剂量PET图像的成像空间得到所述低剂量PET图像的第四补丁和所述MR图像的第四补丁。
  10. 根据权利要求2所述的方法，其特征在于，在所述S21中包括采用K-means聚类算法，以所述第二补丁为样本，获取K个聚类中心作为初始化字典，并对所述初始化字典做归一化处理。
  11. 根据权利要求2所述的方法，其特征在于，在所述S22中，所述初始化联合字典和所述目标矩阵的表达式分别为：
    Figure PCTCN2020134621-appb-100001
    其中，D表示初始化联合字典，Y表示目标矩阵，Dl表示低剂量PET字典，Dr表示MR字典，Ds表示标准剂量PET字典，Yl表示低剂量PET图像的第二补丁，Yr表示MR图像的第二补丁，Ys表示标准剂量PET图像的第二补丁。
  12. 根据权利要求2所述的方法，其特征在于，在所述S23中，所述稀疏编码表达式包括：
    Figure PCTCN2020134621-appb-100002
    其中，X表示稀疏编码，Λ是一个对角矩阵，其对角元素为Λ q＝d q-y i，这里d q为字典D的第q个原子，y i为Y的第i个元素，λ表示稀疏约束系数。
  13. 根据权利要求2所述的方法，其特征在于，在所述S23中，所述字典更新表达式包括：
    Figure PCTCN2020134621-appb-100003
    其中，y i为Y的第i个元素，
    Figure PCTCN2020134621-appb-100004
    为样本x各元素的绝对值，ψ i为以
    Figure PCTCN2020134621-appb-100005
    为对角元素的对角矩阵，y i为Y的第i个元素，μ表示稀疏约束系数。
  14. 根据权利要求13所述的方法，其特征在于，通过梯度下降法求解所述字典更新表达式，所述梯度下降法表达式为：
    Figure PCTCN2020134621-appb-100006
    其中，d q为字典D的第q个原子，k为迭代次数，a q为
    Figure PCTCN2020134621-appb-100007
    的第q列元素，b q为
    Figure PCTCN2020134621-appb-100008
    的第q列元素，a qq为
    Figure PCTCN2020134621-appb-100009
    第q行q列元素。
  15. 一种低剂量PET图像还原系统，其特征在于，包括：
    样本获取单元，用于对包括低剂量PET图像、MR图像和标准剂量PET 图像在内的训练图像进行分块处理获取第一补丁，并对所述第一补丁进行第一预处理获取第二补丁；
    联合字典获取单元，用于根据所述第二补丁，通过稀疏编码和字典更新获取第一联合字典；
    图像还原单元，用于根据所述第一联合字典将所述低剂量PET图像还原成标准剂量PET的还原图像。
  16. 根据权利要求15所述的系统，其特征在于，所述联合字典获取单元还包括：
    初始化单元：用于以所述第二补丁为样本获取包括低剂量PET字典、MR字典和标准剂量PET字典在内的初始化字典；
    构建单元：用于根据所述初始化字典构建初始化联合字典，根据所述第二补丁构建目标矩阵；
    迭代单元：用于根据所述目标矩阵获取稀疏编码，对所述稀疏编码和所述初始化联合字典进行迭代更新至满足迭代停止条件后，获取包括第一低剂量PET字典、第一MR字典和第一标准剂量PET字典在内的所述第一联合字典。
  17. 根据权利要求16所述的系统，其特征在于，所述迭代单元还包括：
    迭代更新单元：用于在每次迭代包括先固定字典更新稀疏编码，再固定稀疏编码更新字典；
    样本选取单元：用于在每次迭代时随机选取部分样本进行稀疏编码和字典更新。
  18. 一种计算机设备，其特征在于，所述计算机设备包括：
    一个或多个处理器；
    存储器，用于存储一个或多个程序；
    当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如权利要求1-14中任一项所述的低剂量PET图像还原方 法。
  19. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，该程序被处理器执行时实现如权利要求1-14中任一项所述的低剂量PET图像还原方法。
PCT/CN2020/134621 2020-12-08 2020-12-08 一种低剂量pet图像还原方法、系统、设备和介质 WO2022120588A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US18/265,685 US20240046459A1 (en) 2020-12-08 2020-12-08 Low-dose pet image restoration method and system, device, and medium
PCT/CN2020/134621 WO2022120588A1 (zh) 2020-12-08 2020-12-08 一种低剂量pet图像还原方法、系统、设备和介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/134621 WO2022120588A1 (zh) 2020-12-08 2020-12-08 一种低剂量pet图像还原方法、系统、设备和介质

Publications (1)

Publication Number Publication Date
WO2022120588A1 true WO2022120588A1 (zh) 2022-06-16

Family

ID=81974066

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/134621 WO2022120588A1 (zh) 2020-12-08 2020-12-08 一种低剂量pet图像还原方法、系统、设备和介质

Country Status (2)

Country Link
US (1) US20240046459A1 (zh)
WO (1) WO2022120588A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016033458A1 (en) * 2014-08-29 2016-03-03 The University Of North Carolina At Chapel Hill Restoring image quality of reduced radiotracer dose positron emission tomography (pet) images using combined pet and magnetic resonance (mr)
CN106204674A (zh) * 2016-06-29 2016-12-07 浙江大学 基于结构字典和动力学参数字典联合稀疏约束的动态pet图像重建方法
CN109035356A (zh) * 2018-07-05 2018-12-18 四川大学 一种基于pet图形成像的系统及方法
CN109242865A (zh) * 2018-09-26 2019-01-18 上海联影智能医疗科技有限公司 基于多图谱的医学图像自动分区系统、方法、装置及存储介质
WO2019134879A1 (en) * 2018-01-03 2019-07-11 Koninklijke Philips N.V. Full dose pet image estimation from low-dose pet imaging using deep learning

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016033458A1 (en) * 2014-08-29 2016-03-03 The University Of North Carolina At Chapel Hill Restoring image quality of reduced radiotracer dose positron emission tomography (pet) images using combined pet and magnetic resonance (mr)
CN106204674A (zh) * 2016-06-29 2016-12-07 浙江大学 基于结构字典和动力学参数字典联合稀疏约束的动态pet图像重建方法
WO2019134879A1 (en) * 2018-01-03 2019-07-11 Koninklijke Philips N.V. Full dose pet image estimation from low-dose pet imaging using deep learning
CN109035356A (zh) * 2018-07-05 2018-12-18 四川大学 一种基于pet图形成像的系统及方法
CN109242865A (zh) * 2018-09-26 2019-01-18 上海联影智能医疗科技有限公司 基于多图谱的医学图像自动分区系统、方法、装置及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG YAN, SHEN DINGGANG, MA GUANGKAI, AN LE, SHI FENG, ZHANG PEI, LALUSH DAVID S., WU XI, PU YIFEI, ZHOU JILIU: "Semisupervised Tripled Dictionary Learning for Standard-Dose PET Image Prediction Using Low-Dose PET and Multimodal MRI", IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, vol. 64, no. 3, 1 March 2017 (2017-03-01), USA, pages 569 - 579, XP055940777, ISSN: 0018-9294, DOI: 10.1109/TBME.2016.2564440 *

Also Published As

Publication number Publication date
US20240046459A1 (en) 2024-02-08

Similar Documents

Publication Publication Date Title
JP6878509B2 (ja) データ及びモデルを用いた画像再構成ならびに補正のためのシステム及び方法
CN107133996B (zh) 产生用于pet数据重建的衰减图的方法及pet/ct系统
CN111462020B (zh) 心脏图像的运动伪影校正方法、系统、存储介质和设备
WO2020114329A1 (zh) 磁共振快速参数成像方法及装置
Liu et al. Deep learning with noise‐to‐noise training for denoising in SPECT myocardial perfusion imaging
CN112365560B (zh) 基于多级网络的图像重建方法、系统、可读存储介质和设备
WO2020118829A1 (zh) 基于决策树的pet图像超分辨重建方法、装置、设备及介质
JP2023133132A (ja) トランスフォーマ特徴共有に基づくｐｅｔ画像再構成方法、装置、デバイス及び媒体
CN111968110A (zh) Ct成像方法、装置、存储介质及计算机设备
CN112488949A (zh) 一种低剂量pet图像还原方法、系统、设备和介质
Kläser et al. Deep boosted regression for MR to CT synthesis
Teng et al. Respiratory deformation registration in 4D-CT/cone beam CT using deep learning
WO2022193379A1 (zh) 图像重建模型生成及装置、图像重建方法及装置、设备、介质
Fu et al. A two-branch neural network for short-axis PET image quality enhancement
CN113989110A (zh) 肺部图像配准方法、装置、计算机设备和存储介质
WO2022120588A1 (zh) 一种低剂量pet图像还原方法、系统、设备和介质
CN111862255A (zh) 正则化图像重建方法、系统、可读存储介质和设备
CN112365479A (zh) Pet参数图像处理方法、装置、计算机设备及存储介质
JP7459243B2 (ja) １以上のニューラルネットワークとしての画像形成のモデル化による画像再構成
WO2024109762A1 (zh) Pet参数确定方法、装置、设备和存储介质
CN112991482B (zh) 基于gpu的快速重建成像方法、设备及可读存储介质
Luo et al. AMCNet: attention-based multiscale convolutional network for DCM MRI segmentation
US20240029324A1 (en) Method for image reconstruction, computer device and storage medium
CN113450427B (zh) 基于联合字典学习和深度网络的pet图像重建方法
US20230154067A1 (en) Output Validation of an Image Reconstruction Algorithm

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: 20964533

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18265685

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20964533

Country of ref document: EP

Kind code of ref document: A1