WO2021031069A1 - Image reconstruction method and apparatus - Google Patents

Abstract

An image reconstruction method and apparatus, applied to the technical field of computer applications. The method comprises: obtaining a first image to be processed (S101); and performing super-resolution processing on the first image according to a preset mapping matrix to obtain a second image (S102). The mapping relation of mapping a low-resolution image to a high-resolution image is fit by means of a decision tree method of the obtained sample image and the projection image in advance, to map a low-resolution positron emission computed tomography PET into a high-resolution projection image, so that the quality of the PET image after each iterative reconstruction is improved, the reconstruction is converged in advance, and the quality of the PET reconstructed image is improved when the number of iterations is reduced.

Description

Image reconstruction method and device Technical field

This application belongs to the field of computer application technology, and in particular relates to an image reconstruction method and device.

Background technique

Positron Emission Computed Tomography (PET) is a relatively advanced clinical examination imaging technology in the field of nuclear medicine, and high-quality PET can improve the diagnosis accuracy of doctors, so improving the PET image reconstruction algorithm has always The subject of research. The existing PET image reconstruction algorithms are mainly divided into two categories: analytical reconstruction algorithms and iterative reconstruction algorithms. Analytical reconstruction algorithms mainly include back projection, filtered back projection and Fourier reconstruction. The most widely used algorithm is Filtered Back-projection (FBP). The iterative reconstruction algorithm also includes algebraic reconstruction and statistical reconstruction. At present, the maximum likelihood expectation maximization in statistical reconstruction is widely used in clinical and practice because of its better performance. However, when there is relatively serious statistical noise in the projected image, the image quality will produce checkerboard artifacts as the number of iterations increase. This method will amplify the noise accordingly, and the quality of the reconstructed image will be lower.

technical problem

The embodiments of the present application provide an image reconstruction method and device, which can solve the problem that image noise is amplified during image reconstruction in the prior art and the quality of the reconstructed image obtained is low.

Technical solutions

In the first aspect, an embodiment of the present application provides an image reconstruction method, including:

Acquiring the first image to be processed;

According to a preset mapping matrix, super-resolution processing is performed on the first image to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through the decision tree method, and is used to convert low The resolution image is mapped to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.

Wherein, before performing super-resolution processing on the first image according to a preset mapping matrix to obtain a second image, the method further includes:

Obtain sample images to be trained;

Performing affine transformation on the sample image to obtain the projection image;

Performing decision tree training on the sample image and the projection image to obtain a sample processing result;

Calculate the loss function between the sample processing result and the projection image, adjust the preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image .

Wherein, performing decision tree training on the sample image and the projection image to obtain a sample processing result includes:

According to a preset decision tree method, iteratively process the sample image to obtain the sample processing result;

The calculating the loss function between the sample processing result and the projection image, and adjusting a preset mapping relationship according to the loss function to obtain the mapping matrix includes:

Calculating a square loss function value between each of the sample processing results and the projected image, and identifying the sample processing result when the square loss function value is the smallest as the target result;

The relationship matrix between the target result and the projected image is determined as the mapping matrix.

Wherein, said performing iterative processing on said sample image to obtain said sample processing result according to a preset decision tree method includes:

According to a preset decision tree method, perform image smoothing iterative processing on the sample image to obtain a smooth image;

Performing pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image;

Performing iterative super-resolution processing on the fused image to obtain the sample processing result.

Wherein, the iterative processing of the sample image to obtain the sample processing result is:

Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the The pixel value of the sample image; y represents the pixel value of the projected image.

Wherein, the calculation of the square loss function value between each of the sample processing results and the projection image includes:

Calculate the square loss function value between each of the sample processing results and the projected image by the following formula:

Wherein, N represents the total number of sample images, x ⁿ denotes the n-th sample image, y ⁿ represents the n-th sample image corresponding to the projected image;

The mapping matrix is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ;

Wherein, X represents the sample image, Y represents the projected image, λ represents a preset regularization parameter, and I represents a unit matrix.

Wherein, the iterative super-resolution processing on the fused image to obtain the sample processing result includes:

Perform iterative super-resolution processing on the fused image according to the following formula to obtain the sample processing result:

Wherein, W represents the mapping matrix;

Represents the pixel value of the fused image in the n+1th iteration;

The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.

In a second aspect, an embodiment of the present application provides an image reconstruction device, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, and the processor executes the The following steps are implemented when computer-readable instructions:

Acquiring the first image to be processed;

According to a preset mapping matrix, super-resolution processing is performed on the first image to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through the decision tree method, and is used to convert low The resolution image is mapped to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.

In a third aspect, an embodiment of the present application provides an image reconstruction device, including:

An acquiring unit for acquiring the first image to be processed;

The reconstruction unit is configured to perform super-resolution processing on the first image according to a preset mapping matrix to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through a decision tree method , Used to map a low-resolution image to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium that stores computer-readable instructions, the computer-readable instructions include program instructions, and the program instructions when executed by a processor The processor is caused to execute the method of the first aspect described above.

In the fifth aspect, the embodiments of the present application provide a computer-readable instruction product, which when the computer-readable instruction product runs on a terminal device, causes the terminal device to execute the image reconstruction method described in any one of the above-mentioned first aspects.

It can be understood that, for the beneficial effects of the second aspect to the fifth aspect described above, reference may be made to the related description in the first aspect described above, and details are not repeated here.

Beneficial effect

Compared with the prior art, the embodiment of the present application has the following beneficial effects: obtaining the first image to be processed; performing super-resolution processing on the first image according to a preset mapping matrix to obtain the second image. In this embodiment, the obtained sample image and the projection image are passed through a decision tree method in advance to fit the mapping relationship from the low-resolution image to the high-resolution image, so as to convert the low-resolution positron emission computer The tomographic PET mapping is a high-resolution projection image, which improves the quality of the PET image reconstructed in each iteration, so that the reconstruction reaches convergence in advance, and improves the quality of the PET reconstructed image while reducing the number of iterations.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only of the present application. For some embodiments, for those of ordinary skill in the art, other drawings can be obtained from these drawings without creative labor.

FIG. 1 is a flowchart of an image reconstruction method provided in Embodiment 1 of the present application;

FIG. 2 is a flowchart of an image reconstruction method provided by Embodiment 2 of the present application;

Fig. 3 is an experimental result of image reconstruction provided in the second embodiment of the present application;

FIG. 4 is a schematic diagram of an image reconstruction device provided in Embodiment 3 of the present application;

FIG. 5 is a schematic diagram of an image reconstruction device provided in Embodiment 4 of the present application.

Embodiments of the invention

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are proposed for a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from obstructing the description of this application.

It should be understood that when used in the specification and appended claims of this application, the term "comprising" indicates the existence of the described features, wholes, steps, operations, elements and/or components, but does not exclude one or more other The existence or addition of features, wholes, steps, operations, elements, components, and/or collections thereof.

It should also be understood that the term "and/or" used in the specification and appended claims of this application refers to any combination of one or more of the items listed in the associated and all possible combinations, and includes these combinations.

As used in the description of this application and the appended claims, the term "if" can be construed as "when" or "once" or "in response to determination" or "in response to detecting ". Similarly, the phrase "if determined" or "if detected [described condition or event]" can be interpreted as meaning "once determined" or "response to determination" or "once detected [described condition or event]" depending on the context ]" or "in response to detection of [condition or event described]".

In addition, in the description of the specification of this application and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

The reference to "one embodiment" or "some embodiments" described in the specification of this application means that one or more embodiments of this application include a specific feature, structure, or characteristic described in combination with the embodiment. Therefore, the phrases "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. appearing in different places in this specification are not necessarily All refer to the same embodiment, but mean "one or more but not all embodiments" unless it is specifically emphasized otherwise. The terms "including", "including", "having" and their variations all mean "including but not limited to" unless otherwise specifically emphasized.

Refer to FIG. 1, which is a flowchart of an image reconstruction method provided in Embodiment 1 of the present application. The execution subject of the image reconstruction method in this embodiment is a device with an image reconstruction function, including but not limited to devices such as computers, servers, tablets, or terminals. The image reconstruction method shown in the figure may include the following steps:

S101: Acquire a first image to be processed.

This embodiment proposes a PET positron emission computer tomography image reconstruction algorithm based on a decision tree. The super-resolution technology based on the decision tree is added to the PET image reconstruction algorithm based on the stain, and the decision tree is used to fit The model that maps the resolution image to the high resolution image, and super-resolution is performed on the image reconstructed in each iteration. It is possible to reduce the number of iterations to achieve convergence earlier, and at the same time reduce the time to adjust parameters, achieve better reconstruction results under relatively poor parameter settings, and improve the quality of PET reconstructed images.

Positron emission computed tomography is a relatively advanced clinical examination imaging technology in the field of nuclear medicine, and high-quality PET can improve the diagnosis accuracy of doctors, so improving the PET image reconstruction algorithm has always been the subject of research. The existing PET image reconstruction algorithms are mainly divided into two categories: analytical reconstruction algorithms and iterative reconstruction algorithms. Analytical reconstruction algorithms mainly include back projection, filtered back projection and Fourier reconstruction. One of the most widely used algorithms is filtered back projection. Iterative reconstruction algorithms also include algebraic reconstruction and statistical reconstruction. At present, maximum likelihood-expectation maximization in statistical reconstruction is widely used in clinical and practice because of its better performance. However, when there is relatively serious statistical noise in the projected image, the image quality will produce checkerboard artifacts as the number of iterations increases, but worse. Therefore, there is a penalized likelihood PET image reconstruction algorithm based on color spots that introduces regularization terms.

The first image in this embodiment is used to represent a PET image with a lower resolution or a reconstructed image. The acquisition method can be directly acquired through PET scanning equipment.

S102: Perform super-resolution processing on the first image according to a preset mapping matrix to obtain a second image; the mapping matrix is obtained by training the acquired sample image and projection image through a decision tree method, and is used for The low-resolution image is mapped to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.

Existing PET image reconstruction algorithms, commonly used include filtering back projection in analytical reconstruction and maximum likelihood-expectation maximization (MLEM) algorithms in iterative reconstruction. Although the reconstruction algorithm is simple and fast, its reconstruction results have poor resolution and noise characteristics, and require a complete projection image and a large count value. The problem in the actual use of the MLEM algorithm is that when there is relatively serious statistical noise in the projected image, as the iteration proceeds, the quality of the reconstructed image is not completely better, and the noise will be amplified accordingly. On this basis, an improved regularization iterative reconstruction algorithm based on patch patches introduces regularization items in the image iterative update process, but it is sensitive to the value of the algorithm parameters, and it takes a lot of time to adjust the parameters to achieve the best reconstruction effect. This embodiment preserves the edges and details to achieve better reconstruction of the image, but also reduces the time to adjust the parameter settings. Under poor parameter settings, better reconstruction can be achieved without a higher count level. .

The first image in this embodiment is used to represent an acquired low-resolution PET image or a reconstructed image, and the second image is used to represent a high-resolution projection image obtained by performing PET image reconstruction on the first image. In this embodiment, in the process of reconstructing the PET image, a machine learning algorithm is added. In each iteration of the image reconstruction, a decision tree is used to train the low-resolution image block and the corresponding high-resolution image block to fit the low-resolution image The mapping relationship that is mapped to the high-resolution image, that is, the mapping matrix, is used to perform super-resolution processing on the first image through the mapping matrix to obtain the second image, which improves the quality of the PET image after each iteration and makes the reconstruction reach convergence in advance. It reduces the number of iterations while improving the quality of PET reconstructed images.

In the above solution, the first image to be processed is acquired; the super-resolution processing is performed on the first image according to the preset mapping matrix to obtain the second image. In this embodiment, the obtained sample image and the projection image are passed through a decision tree method in advance to fit the mapping relationship from the low-resolution image to the high-resolution image, so as to convert the low-resolution positron emission computer The tomographic PET mapping is a high-resolution projection image, which improves the quality of the PET image reconstructed in each iteration, so that the reconstruction reaches convergence in advance, and improves the quality of the PET reconstructed image while reducing the number of iterations.

Refer to FIG. 2, which is a flowchart of an image reconstruction method provided by an embodiment of the present application. The image reconstruction method as shown in the figure may include the following steps before step S101:

S201: Obtain a sample image to be trained.

The sample image in this embodiment may be a reconstructed image, where the reconstructed image x is a PET image of a hospital patient, and the projection image y is obtained by performing affine transformation on x.

S202: Perform affine transformation on the sample image to obtain the projection image.

The data obtained in this embodiment is not a projection image collected in real time, so the projection image y is obtained by projecting the reconstructed PET image x, and the super-resolution process is added during the reconstruction of the projection image y into a PET image, and finally the reconstructed PET image is obtained.

Projected image of this embodiment

The reconstructed image x is obtained by affine transformation:

Among them, P represents the system matrix, which represents the probability that the detector detects a coincidence event for the pixel i in the sample image; r represents a random background event, and s represents a scattering event.

S203: Perform decision tree training on the sample image and the projection image to obtain a sample processing result.

Step S203 includes: performing iterative processing on the sample image to obtain the sample processing result according to a preset decision tree method.

In this embodiment, the sample image is processed iteratively and the sample processing result is:

Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the sample image The pixel value of; y represents the pixel value of the projected image.

Specifically, since the decayed positron emission itself satisfies the Poisson distribution, we assume that the PET projection image y can be regarded as the distribution of independent Poisson random variables. The penalty likelihood reconstruction estimates the reconstructed image x by maximizing the penalty likelihood function:

Among them, Φ(x)=L(y|x)-βU(x); among them, the regularization parameter β can be set to β=2 ^-7 during initialization, we choose Q _L (x; x ⁿ ),

As the likelihood proxy function L(y|x) and the penalty proxy function U(x):

among them,

Among them, n _j represents the total number of pixels in the image, j and k respectively represent two image blocks; system matrix P = {p _ij }, p _ij represents the probability that the detector pair i detects a coincidence event at pixel j; y _i represents The i-th pair of projection data collected by the detector;

Represents the predicted projection data of the nth iteration; x ⁿ represents the image of the nth iteration;

Represents the jth image block of the nth iteration;

Represents the kth image block of the nth iteration; N _j represents the total pixel value of the jth image block;

Is the weight related to the neighborhood block, which is adaptively determined by the penalty function and the current estimated image of each iteration;

Represents the pixel-based weight for the nth iteration.

Further, the step of iteratively processing the sample image to obtain the sample processing result according to a preset decision tree method includes S2031 to S2033:

S2031: According to a preset decision tree method, perform image smoothing iterative processing on the sample image to obtain a smooth image.

This embodiment implements image smoothing according to the following formula:

Where x ⁿ represents the nth iteration image;

Represents the jth image block of the nth iteration;

Represents the kth image block of the nth iteration; N _j represents the total pixel value of the jth image block; w _jk represents the weight related to the neighborhood block, which is adaptively determined by the penalty function and the current estimated image of each iteration .

S2032: Perform pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image.

In this embodiment, the EM image is obtained by updating the _{sinogram {y i}}

Then get through image smoothing

Finally, it is fused pixel by pixel, and the iterative image of each penalty likelihood reconstruction is obtained through the KKT condition:

among them,

The regularization parameter β is a constant, used to control the weight of the prior and balance the log-likelihood term and the penalty term.

S2033: Perform iterative super-resolution processing on the fused image to obtain the sample processing result.

In this embodiment, iterative super-resolution processing is performed on the fused image according to the following formula to obtain the sample processing result:

Wherein, W represents the mapping matrix;

Represents the pixel value of the fused image in the n+1th iteration;

The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.

Specifically, after each iteration of the reconstructed image is obtained, the decision tree is used to perform super-resolution processing. The model from low-resolution image to high-resolution image fitted through multiple decision tree training is used to reconstruct the image in each iteration

on:

among them,

It is the mapping matrix fitted by the decision tree training. The image x and the high-resolution reference image y reconstructed in each iteration of the low resolution training sample are used to fit the closest mapping from x to y using the decision tree clustering Model, namely W.

It should be noted that W and W in this embodiment

The same, both are used to represent the mapping matrix, here is just for concise expression without the independent variables in parentheses.

S204: Calculate the loss function between the sample processing result and the projection image, and adjust a preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is between the sample processing result and the projection image Mapping relations.

Further, step S204 includes:

S2041: Calculate the square loss function value between each of the sample processing results and the projection image, and identify the sample processing result when the square loss function value is the smallest as the target result.

In this embodiment, the square loss function value between the processing result of each sample and the projected image is calculated by the following formula:

Where N represents the total number of sample images, x ⁿ represents the nth sample image, and y ⁿ represents the projection image corresponding to the nth sample image;

specific,

It is the mapping matrix fitted by the decision tree training. The training sample, that is, the image x and the high-resolution reference image y reconstructed in each iteration of the sample image, are mapped from x to y using the decision tree clustering The closest model is W. And W depends on the low-resolution image block x. Then this mapping relationship can be written as,

Through decision tree training, according to the principle of minimizing the square loss function:

Find the best-fitting model W, where N represents the size of the training sample, x ⁿ represents the low-resolution image block of the nth training sample, and y ⁿ represents the corresponding high-resolution image block.

S2042: Determine the relationship matrix between the target result and the projection image as the mapping matrix.

The mapping matrix of this embodiment is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ; where X represents the sample image, Y represents the projection image, and λ represents the preset regularization The parameter, I represents the identity matrix.

In this embodiment, in order to solve the problem of ridge regression, that is, the problem of regularized least squares regression, W ^T = (X ^T X + λI) ^-1 X ^T ·Y can be calculated; where X and Y respectively represent low resolution For images and corresponding high-resolution images, λI represents the regular term added; I represents the unit matrix; λ represents the regularization parameter, and λ can be set to 0.01.

Train T decision trees to obtain T models, and the final prediction result is the average of the predicted values of T trees. How to train a decision tree is to split and cluster each group of corresponding high- and low-resolution image blocks {x _H , x _L } to the left and right sub-nodes recursively and disjointly, like a binary tree. Until the node sample size is less than 2 and cannot be split or reaches the maximum depth, it stops splitting to form leaf nodes. The leaf node model is the model we need to train and fit. The principle of splitting is to calculate the response function according to the characteristics of the image block. The response function is: r _θ (x _L ) = x _L [θ]-θ _th ; where θ represents the feature of the _{image block x L and x L} [·] represents the image A one-dimensional vector of the data matrix of the block x _{L , θ th} represents the threshold. When r _θ (x _L )<0, the image block pair {x _H , x _L } is split to the left child node; otherwise, it is split to the right child node.

As for calculating the features of the response function, each split will traverse all the features of the image block to select the optimal feature. The reference quantity of the optimal feature is defined as:

Q(σ, θ, X _H , X _L )=∑ _{c∈{Le, Ri}} |X ^c |·E(X _H ^c , X _L ^c );

Among them, (X _H ^c , X _L ^c ) represents the corresponding high and low resolution image blocks split into the left child node and the right child node.

Represents the mean value of the sample x _L ⁿ _{, W(x L} ⁿ )x _L ⁿ represents the predicted value of the sample x _L ⁿ , |N| represents the number of samples split to the child node, that is, the image block, k represents the set hyperparameter; σ represents When the value defined by the left and right child nodes is split to the left child node, σ=0, otherwise σ=1.

The pseudo code diagram of this embodiment is shown in Algorithm 1:

The decision tree model is not only applied to the image reconstructed in each iteration, but also to the projection image {y _i } at the beginning of the loop. It only replaces the training set with the projection image calculated by the iterative reconstruction of each back projection and the corresponding original Reference projection. The decision tree of this embodiment is to cluster the samples into the last few leaf nodes, that is, several categories through the above-described splitting process, and then find the mapping matrix W through the above-described principle of minimizing the square loss function, that is, to find the mapping model.

Please refer to Figure 3 and Table 1 for the experimental results of this embodiment:

Table 1. PET image resolution evaluation parameter table

Evaluation parameters	Peak signal to noise ratio PNSR	Structural similarity SSIM
Patch-based reconstruction image	29.67	0.87
Reconstructed image of this embodiment	34.51	0.88

Figure a in Figure 3 is a reference PET image, Figure b is a PET image reconstructed based on the patch regularization iterative reconstruction algorithm, and Figure c is a PET image reconstructed by the algorithm of this embodiment. It can be seen from the figure that the resolution of the PET image reconstructed in this embodiment is better than the PET image reconstructed based on the patch regularization iterative reconstruction algorithm, and it is also closer to the reference PET image, which fully illustrates the feasibility of the algorithm. It can also be seen from the evaluation parameter table that the image reconstructed by this algorithm is better than patch-based reconstruction on PNSR and SSIM, which further proves the effectiveness of this algorithm.

The above solution is to obtain the sample image to be trained; perform affine transformation on the sample image to obtain the projection image; perform decision tree training on the sample image and the projection image to obtain the sample processing result; calculate the sample For the loss function between the processing result and the projection image, the preset mapping relationship is adjusted according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image. By using a decision tree to fit a model from low-resolution images to high-resolution images, super-resolution is performed on the reconstructed images in each iteration. It is possible to reduce the number of iterations to achieve convergence earlier, and at the same time reduce the time to adjust parameters, achieve better reconstruction results under relatively poor parameter settings, and improve the quality of PET reconstructed images.

Refer to FIG. 4, which is a schematic diagram of an image reconstruction device provided in Embodiment 3 of the present application. The image reconstruction device 400 may be a mobile terminal such as a smart phone or a tablet computer. The units included in the image reconstruction apparatus 400 of this embodiment are used to execute the steps in the embodiment corresponding to FIG. 1. For details, please refer to the related descriptions in the embodiment corresponding to FIG. 1 and FIG. 1, which will not be repeated here. The image reconstruction device 400 of this embodiment includes:

The acquiring unit 401 is configured to acquire the first image to be processed;

The reconstruction unit 402 is configured to perform super-resolution processing on the first image according to a preset mapping matrix to obtain a second image; the mapping matrix is used to train the acquired sample images and projection images through a decision tree method Obtained, used to map a low-resolution image to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.

Further, the image reconstruction device 400 further includes:

The first acquiring unit is used to acquire the sample image to be trained;

A transformation unit, configured to perform affine transformation on the sample image to obtain the projection image;

A training unit, configured to train the sample image and the projection image on a decision tree to obtain a sample processing result;

The calculation unit is configured to calculate a loss function between the sample processing result and the projection image, and adjust a preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the sample processing result and the projection The mapping relationship between images.

Further, the training unit includes:

An iterative processing unit, configured to iteratively process the sample image according to a preset decision tree method to obtain the sample processing result;

Further, the calculation unit includes:

The first calculation unit is configured to calculate the square loss function value between each of the sample processing results and the projection image, and identify the sample processing result when the square loss function value is the smallest as the target result;

The matrix determination unit is used to determine the relationship matrix between the target result and the projection image as the mapping matrix.

Further, the iterative processing unit includes:

A smoothing unit, configured to perform image smoothing iterative processing on the sample image according to a preset decision tree method to obtain a smooth image;

A fusion unit, configured to perform pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image;

The super-resolution processing unit is configured to perform iterative super-resolution processing on the fused image to obtain the sample processing result.

Wherein, the iterative processing of the sample image to obtain the sample processing result is:

Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the The pixel value of the sample image; y represents the pixel value of the projected image.

The calculating the square loss function value between each of the sample processing results and the projection image includes:

Calculate the square loss function value between each of the sample processing results and the projected image by the following formula:

Wherein, N represents the total number of sample images, x ⁿ denotes the n-th sample image, y ⁿ represents the n-th sample image corresponding to the projected image;

The mapping matrix is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ;

Wherein, X represents the sample image, Y represents the projected image, λ represents a preset regularization parameter, and I represents a unit matrix.

Further, the super-resolution processing unit is used for:

Perform iterative super-resolution processing on the fused image according to the following formula to obtain the sample processing result:

Wherein, W represents the mapping matrix;

Represents the pixel value of the fused image in the n+1th iteration;

The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.

In the above solution, the first image to be processed is acquired; the super-resolution processing is performed on the first image according to the preset mapping matrix to obtain the second image. In this embodiment, the obtained sample image and the projection image are passed through a decision tree method in advance to fit the mapping relationship from the low-resolution image to the high-resolution image, so as to convert the low-resolution positron emission computer The tomographic PET mapping is a high-resolution projection image, which improves the quality of the PET image after each iteration, so that the reconstruction reaches convergence in advance, and improves the quality of the PET reconstruction image while reducing the number of iterations.

It should be understood that the size of the sequence number of each step in the foregoing embodiment does not mean the order of execution. The execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiment of the present application.

Referring to FIG. 5, FIG. 5 is a schematic diagram of an image reconstruction apparatus provided in Embodiment 4 of the present application. The image reconstruction apparatus 500 in this embodiment as shown in FIG. 5 may include: a processor 501, a memory 502, and computer-readable instructions 503 stored in the memory 502 and running on the processor 501. When the processor 501 executes the computer-readable instruction 503, the steps in the foregoing image reconstruction method embodiments are implemented. The memory 502 is configured to store computer-readable instructions, and the computer-readable instructions include program instructions. The processor 501 is configured to execute program instructions stored in the memory 502. Wherein, the processor 501 is configured to call the program instructions to perform the following operations:

The processor 501 is used for:

Obtain sample images to be trained;

Performing affine transformation on the sample image to obtain the projection image;

Performing decision tree training on the sample image and the projection image to obtain a sample processing result;

Calculate the loss function between the sample processing result and the projection image, adjust the preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image .

Further, the processor 501 is specifically configured to:

According to a preset decision tree method, iteratively process the sample image to obtain the sample processing result;

The calculating the loss function between the sample processing result and the projection image, and adjusting a preset mapping relationship according to the loss function to obtain the mapping matrix includes:

Calculating a square loss function value between each of the sample processing results and the projected image, and identifying the sample processing result when the square loss function value is the smallest as the target result;

The relationship matrix between the target result and the projected image is determined as the mapping matrix.

Further, the processor 501 is specifically configured to:

According to a preset decision tree method, perform image smoothing iterative processing on the sample image to obtain a smooth image;

Performing pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image;

Performing iterative super-resolution processing on the fused image to obtain the sample processing result.

Wherein, the iterative processing of the sample image to obtain the sample processing result is:

Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the The pixel value of the sample image; y represents the pixel value of the projected image.

The calculating the square loss function value between each of the sample processing results and the projection image includes:

Calculate the square loss function value between each of the sample processing results and the projected image by the following formula:

Wherein, N represents the total number of sample images, x ⁿ denotes the n-th sample image, y ⁿ represents the n-th sample image corresponding to the projected image;

The mapping matrix is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ;

Wherein, X represents the sample image, Y represents the projected image, λ represents a preset regularization parameter, and I represents a unit matrix.

Further, the processor 501 is specifically configured to:

Perform iterative super-resolution processing on the fused image according to the following formula to obtain the sample processing result:

Wherein, W represents the mapping matrix;

Represents the pixel value of the fused image in the n+1th iteration;

The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.

In the above solution, the first image to be processed is acquired; the super-resolution processing is performed on the first image according to the preset mapping matrix to obtain the second image. In this embodiment, the obtained sample image and the projection image are passed through a decision tree method in advance to fit the mapping relationship from the low-resolution image to the high-resolution image, so as to convert the low-resolution positron emission computer The tomographic PET mapping is a high-resolution projection image, which improves the quality of the PET image after each iteration, so that the reconstruction reaches convergence in advance, and improves the quality of the PET reconstructed image while reducing the number of iterations.

It should be understood that in the embodiment of the present application, the processor 501 may be a central processing unit (Central Processing Unit, CPU), and the processor may also be other general-purpose processors or digital signal processors (DSP). , Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.

The memory 502 may include a read-only memory and a random access memory, and provides instructions and data to the processor 501. A part of the memory 502 may also include a non-volatile random access memory. For example, the memory 502 may also store device type information.

In specific implementation, the processor 501, memory 502, and computer-readable instructions 503 described in the embodiments of this application can execute the implementations described in the first embodiment and the second embodiment of the image reconstruction method provided in the embodiments of this application. The way, the implementation way of the terminal described in the embodiment of this application can also be implemented, which will not be repeated here.

In another embodiment of the present application, a computer-readable storage medium is provided, the computer-readable storage medium stores computer-readable instructions, the computer-readable instructions include program instructions, and the program instructions are executed by a processor When realized:

Obtain sample images to be trained;

Performing affine transformation on the sample image to obtain the projection image;

Performing decision tree training on the sample image and the projection image to obtain a sample processing result;

Calculate the loss function between the sample processing result and the projection image, adjust the preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image .

Further, when the computer-readable instruction is executed by the processor, it also implements:

According to a preset decision tree method, iteratively process the sample image to obtain the sample processing result;

The calculating the loss function between the sample processing result and the projection image, and adjusting a preset mapping relationship according to the loss function to obtain the mapping matrix includes:

Calculating a square loss function value between each of the sample processing results and the projected image, and identifying the sample processing result when the square loss function value is the smallest as the target result;

The relationship matrix between the target result and the projected image is determined as the mapping matrix.

Further, when the computer-readable instruction is executed by the processor, it also implements:

According to a preset decision tree method, perform image smoothing iterative processing on the sample image to obtain a smooth image;

Performing pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image;

Performing iterative super-resolution processing on the fused image to obtain the sample processing result.

Wherein, the iterative processing of the sample image to obtain the sample processing result is:

Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the The pixel value of the sample image; y represents the pixel value of the projected image.

Further, when the computer-readable instruction is executed by the processor, it also implements:

Calculate the square loss function value between each of the sample processing results and the projected image by the following formula:

Wherein, N represents the total number of sample images, x ⁿ denotes the n-th sample image, y ⁿ represents the n-th sample image corresponding to the projected image;

The mapping matrix is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ;

Wherein, X represents the sample image, Y represents the projected image, λ represents a preset regularization parameter, and I represents a unit matrix.

Further, when the computer-readable instruction is executed by the processor, it also implements:

Perform iterative super-resolution processing on the fused image according to the following formula to obtain the sample processing result:

Wherein, W represents the mapping matrix;

Represents the pixel value of the fused image in the n+1th iteration;

The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.

In the above solution, the first image to be processed is acquired; the super-resolution processing is performed on the first image according to the preset mapping matrix to obtain the second image. In this embodiment, the obtained sample image and the projection image are passed through a decision tree method in advance to fit the mapping relationship from the low-resolution image to the high-resolution image, so as to convert the low-resolution positron emission computer The tomographic PET mapping is a high-resolution projection image, which improves the quality of the PET image reconstructed in each iteration, so that the reconstruction reaches convergence in advance, and improves the quality of the PET reconstructed image while reducing the number of iterations.

The computer-readable storage medium may be the internal storage unit of the terminal described in any of the foregoing embodiments, such as the hard disk or memory of the terminal. The computer-readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk equipped on the terminal, a smart memory card (Smart Media Card, SMC), or a Secure Digital (SD) card , Flash Card, etc. Further, the computer-readable storage medium may also include both an internal storage unit of the terminal and an external storage device. The computer-readable storage medium is used to store the computer-readable instructions and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

A person of ordinary skill in the art may realize that the units and algorithm steps of the examples described in the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two, in order to clearly illustrate the hardware and software Interchangeability. In the above description, the composition and steps of each example have been generally described in terms of function. Whether these functions are executed by hardware or software depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.

Those skilled in the art can clearly understand that, for the convenience and conciseness of description, the specific working process of the terminal and unit described above can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed terminal and method can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components can be combined or It can be integrated into another system, or some features can be ignored or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may also be electrical, mechanical or other forms of connection.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments of the present application. In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solution of this application is essentially or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium It includes several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program code .

The above are only specific implementations of this application, but the protection scope of this application is not limited to this. Anyone familiar with the technical field can easily think of various equivalents within the technical scope disclosed in this application. Modifications or replacements, these modifications or replacements shall be covered within the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the claims.

Claims (20)

1. An image reconstruction method, characterized in that it comprises:

   Acquiring the first image to be processed;

   According to a preset mapping matrix, super-resolution processing is performed on the first image to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through the decision tree method, and is used to convert low The resolution image is mapped to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.
2. 8. The image reconstruction method according to claim 1, wherein said performing super-resolution processing on said first image according to a preset mapping matrix to obtain a second image, further comprising:

   Obtain sample images to be trained;

   Performing affine transformation on the sample image to obtain the projection image;

   Performing decision tree training on the sample image and the projection image to obtain a sample processing result;

   Calculate the loss function between the sample processing result and the projection image, adjust the preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image .
3. 3. The image reconstruction method according to claim 2, wherein the training of the decision tree on the sample image and the projection image to obtain a sample processing result comprises:

   According to a preset decision tree method, iteratively process the sample image to obtain the sample processing result;

   The calculating the loss function between the sample processing result and the projection image, and adjusting a preset mapping relationship according to the loss function to obtain the mapping matrix includes:

   Calculating a square loss function value between each of the sample processing results and the projected image, and identifying the sample processing result when the square loss function value is the smallest as the target result;

   The relationship matrix between the target result and the projected image is determined as the mapping matrix.
4. 8. The image reconstruction method according to claim 3, wherein the iterative processing of the sample image to obtain the sample processing result according to a preset decision tree method comprises:

   According to a preset decision tree method, perform image smoothing iterative processing on the sample image to obtain a smooth image;

   Performing pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image;

   Performing iterative super-resolution processing on the fused image to obtain the sample processing result.
5. 8. The image reconstruction method according to claim 3, wherein the iterative processing of the sample image to obtain the sample processing result is:

   

   Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the The pixel value of the sample image; y represents the pixel value of the projected image.
6. 5. The image reconstruction method according to claim 3, wherein the calculating the square loss function value between each of the sample processing results and the projected image comprises:

   Calculate the square loss function value between each of the sample processing results and the projected image by the following formula:

   

   Wherein, N represents the total number of sample images, x ⁿ denotes the n-th sample image, y ⁿ represents the n-th sample image corresponding to the projected image;

   The mapping matrix is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ;

   Wherein, X represents the sample image, Y represents the projected image, λ represents a preset regularization parameter, and I represents a unit matrix.
7. 5. The image reconstruction method according to claim 4, wherein the iterative super-resolution processing on the fused image to obtain the sample processing result comprises:

   Perform iterative super-resolution processing on the fused image according to the following formula to obtain the sample processing result:

   

   Wherein, W represents the mapping matrix;

   

   Represents the pixel value of the fused image in the n+1th iteration;

   

   The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.
8. An image reconstruction device, characterized in that it comprises:

   An acquiring unit for acquiring the first image to be processed;

   The reconstruction unit is configured to perform super-resolution processing on the first image according to a preset mapping matrix to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through a decision tree method , Used to map a low-resolution image to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.
9. 8. The image reconstruction device of claim 8, wherein the image reconstruction device further comprises:

   The first acquiring unit is used to acquire the sample image to be trained;

   A transformation unit, configured to perform affine transformation on the sample image to obtain the projection image;

   A training unit, configured to train the sample image and the projection image on a decision tree to obtain a sample processing result;

   The calculation unit is configured to calculate a loss function between the sample processing result and the projection image, and adjust a preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the sample processing result and the projection The mapping relationship between images.
10. 9. The image reconstruction device according to claim 9, wherein the training unit comprises:

    An iterative processing unit, configured to iteratively process the sample image according to a preset decision tree method to obtain the sample processing result;

    The calculation unit includes:

    The first calculation unit is configured to calculate the square loss function value between each of the sample processing results and the projection image, and identify the sample processing result when the square loss function value is the smallest as the target result;

    The matrix determination unit is used to determine the relationship matrix between the target result and the projection image as the mapping matrix.
11. An image reconstruction device, including a memory, a processor, and computer-readable instructions stored in the memory and running on the processor, wherein the processor executes the computer-readable instructions to implement The following steps:

    Acquiring the first image to be processed;

    According to a preset mapping matrix, super-resolution processing is performed on the first image to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through the decision tree method, and is used to convert low The resolution image is mapped to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.
12. 11. The image reconstruction device according to claim 11, wherein said performing super-resolution processing on said first image according to a preset mapping matrix to obtain a second image, further comprising:

    Obtain sample images to be trained;

    Performing affine transformation on the sample image to obtain the projection image;

    Performing decision tree training on the sample image and the projection image to obtain a sample processing result;

    Calculate the loss function between the sample processing result and the projection image, adjust the preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image .
13. 11. The image reconstruction device according to claim 12, wherein the training of a decision tree on the sample image and the projection image to obtain a sample processing result comprises:

    According to a preset decision tree method, iteratively process the sample image to obtain the sample processing result;

    The calculating the loss function between the sample processing result and the projection image, and adjusting a preset mapping relationship according to the loss function to obtain the mapping matrix includes:

    Calculating a square loss function value between each of the sample processing results and the projected image, and identifying the sample processing result when the square loss function value is the smallest as the target result;

    The relationship matrix between the target result and the projected image is determined as the mapping matrix.
14. 15. The image reconstruction device according to claim 13, wherein the iterative processing of the sample image to obtain the sample processing result according to a preset decision tree method comprises:

    According to a preset decision tree method, perform image smoothing iterative processing on the sample image to obtain a smooth image;

    Performing pixel image fusion iterative processing on the pixels in the smooth image to obtain a fused image;

    Performing iterative super-resolution processing on the fused image to obtain the sample processing result.
15. The image reconstruction device according to claim 13, wherein the iterative processing of the sample image to obtain the sample processing result is:

    

    Among them, Φ(x)=L(y|x)-βU(x); L(y|x) represents the likelihood proxy function; U(x) represents the penalty proxy function; β represents the regularization parameter, and x represents the The pixel value of the sample image; y represents the pixel value of the projected image.
16. The image reconstruction device according to claim 13, wherein said calculating the square loss function value between each of said sample processing results and said projection image comprises:

    Calculate the square loss function value between each of the sample processing results and the projected image by the following formula:

    

    Wherein, N represents the total number of sample images, x ⁿ denotes the n-th sample image, y ⁿ represents the n-th sample image corresponding to the projected image;

    The mapping matrix is: W=[(X ^T X+λI) ^-1 X ^T ·Y] ^T ;

    Wherein, X represents the sample image, Y represents the projected image, λ represents a preset regularization parameter, and I represents a unit matrix.
17. The image reconstruction device according to claim 14, wherein the iterative super-resolution processing on the fused image to obtain the sample processing result comprises:

    Perform iterative super-resolution processing on the fused image according to the following formula to obtain the sample processing result:

    

    Wherein, W represents the mapping matrix;

    

    Represents the pixel value of the fused image in the n+1th iteration;

    

    The sample processing result corresponding to the pixel value of the fused image in the n+1th iteration.
18. A computer-readable storage medium that stores computer-readable instructions, wherein the computer-readable instructions implement the following steps when executed by a processor:

    Acquiring the first image to be processed;

    According to a preset mapping matrix, super-resolution processing is performed on the first image to obtain a second image; the mapping matrix is obtained by training the acquired sample images and projection images through the decision tree method, and is used to convert low The resolution image is mapped to a high-resolution image; the projection image is obtained by performing affine transformation on the sample image.
19. 18. The computer-readable storage medium according to claim 18, wherein said performing super-resolution processing on said first image according to a preset mapping matrix to obtain a second image, further comprising:

    Obtain sample images to be trained;

    Performing affine transformation on the sample image to obtain the projection image;

    Performing decision tree training on the sample image and the projection image to obtain a sample processing result;

    Calculate the loss function between the sample processing result and the projection image, adjust the preset mapping relationship according to the loss function to obtain the mapping matrix; the mapping relationship is the mapping relationship between the sample processing result and the projection image .
20. 19. The computer-readable storage medium according to claim 19, wherein the training of the decision tree on the sample image and the projection image to obtain a sample processing result comprises:

    According to a preset decision tree method, iteratively process the sample image to obtain the sample processing result;

    The calculating a loss function between the sample processing result and the projection image, and adjusting a preset mapping relationship according to the loss function to obtain the mapping matrix includes:

    Calculating a square loss function value between each of the sample processing results and the projection image, and identifying the sample processing result when the square loss function value is the smallest as the target result;

    The relationship matrix between the target result and the projected image is determined as the mapping matrix.

Images