A method and system for scatter correction of a pet image

Abstract

The application relates to a PET image scattering correction method and system, which comprises the following steps: modeling a PET acquisition process; alternately solving radioactive activity distribution and detector space attenuation coefficients based on an ADMM algorithm; realizing mapping of PET detection data and detector space attenuation coefficients to accurate scattering distribution estimation based on a deep learning network G; keeping the radioactive activity distribution and the detector space attenuation coefficients constant, and utilizing an unfitted scattering distribution estimation to maximize a log-likelihood function to obtain a scattering correction factor; and utilizing the detector space attenuation coefficients and the accurate scattering distribution after fitting of the scattering correction factor to obtain a PET image after scattering correction. The method has the beneficial effects of avoiding image quality decline caused by inaccurate scattering evaluation, greatly reducing scattering correction time consumption, and greatly improving stability for high-noise data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical imaging, and particularly relates to a PET image scatter correction method and system. BACKGROUND

[0002] Positron Emission Tomography (PET) is a cutting-edge nuclear medical imaging technology. In actual diagnosis and treatment, this technology uses radioisotopes such as 18F and 11C to label metabolic substances, and injects these labeled compounds into the human body. Subsequently, functional metabolic imaging is performed on the patient by the PET system, vividly showing the metabolic activity state in the living body, so as to achieve the goal of disease diagnosis. The commercial PET devices currently circulating in the market are generally combined with other imaging modes, such as Computed Tomography (CT) or Magnetic Resonance Imaging (MRI), which not only can simultaneously obtain the anatomical structure information of the patient, but also can significantly improve the accuracy of lesion positioning. Finally, the functional and anatomical images are fused on the same device, which integrates the advantages of the two modalities, so that doctors can more intuitively and comprehensively understand the patient's physical condition, which is helpful for early detection and accurate diagnosis of diseases, especially in the diagnosis and treatment of tumors, heart and brain diseases, which has shown excellent performance.

[0003] However, in the process of PET data acquisition, photons may experience Compton scattering in the human body, causing the direction of travel to change. Due to the limitations of the energy resolution of the detector, such scattered events are sometimes recorded as real coincidence events, which interferes with the position information of the original signal and induces artifacts in the generated image, which seriously affects the image quality. Especially in three-dimensional acquisition mode, the proportion of scattered coincidence events can be as high as 30% to 60% of the total count, therefore, effective scatter correction has become an indispensable part of the PET reconstruction process.

[0004] The currently widely used scatter correction technique is Single Scatter Simulation (SSS) algorithm. This technique is based on the radioactivity distribution and linear attenuation coefficient distribution of the scanned object, by calculating the probability of single scattering of a single gamma ray particle before detection, to estimate the overall scatter distribution. However, due to the mixed components of multiple scattering in the actual measurement data, the SSS method is difficult to accurately estimate the proportional relationship between the scatter component and the total coincidence rate. Generally, the scatter contribution is determined by analyzing the radial tail characteristics of the sinogram, or by means of Monte Carlo simulation. In actual clinical collection, the accuracy of SSS is usually affected by the following factors:

[0005] First, before scatter correction, the accurate radioactivity initial distribution is usually unknown, and the scatter-free PET reconstruction image is often used as an approximation. When the patient's weight is large, the deviation between the scatter-free PET initial image and the accurate radioactivity initial distribution is large, so this approximation usually leads to a large error in the estimation of the scatter distribution. Even if the scatter parameters are iteratively solved in the SSS process, global convergence cannot be guaranteed, and the above error still exists. The SSS calculation process is complex, and even in the case of a small number of iterations, it still takes a lot of time, which greatly limits the efficiency of scatter correction and increases the time cost of image reconstruction.

[0006] Secondly, for multi-modal acquisition methods, although different imaging techniques can provide us with the necessary attenuation information for scatter estimation, we still face some challenges in actual operation. Taking the PET / CT system as an example, the coverage of PET often exceeds the working area of CT or other auxiliary imaging techniques. Especially for subjects with a larger body size, other imaging methods may be insufficient in field width, resulting in the truncation of the linear attenuation distribution in some areas, thereby affecting the quality of the PET image. In addition, CT imaging is fast and can be completed almost instantaneously, while PET takes a longer time to complete a scan of a part. This leads to the difficulty of keeping the patient completely still during the entire examination process, and factors such as respiratory motion can cause spatial misregistration between PET and CT images. In addition, it is worth noting that if the subject has metal objects in the body, such as a heart pacemaker, dental implants, etc., there may be obvious "metal artifacts" on the CT image, which also brings additional challenges to subsequent processing work. In summary, in clinical practice, due to various reasons, the problem of incomplete matching of the radioactivity distribution image and other modal images will directly affect the effect of scatter correction, and even may introduce serious artifacts in the final results. Finally, when applying the tail fitting algorithm of the SSS algorithm, only relying on the data outside the object as a reference, if the amount of available data is limited, especially for large target objects, this method may become unreliable and prone to errors in estimating the scatter distribution. SUMMARY

[0007] TECHNICAL PROBLEM

[0008] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present application provides a technical problem of how to further improve the image quality and reduce the reconstruction time of PET image scatter correction.

[0009] TECHNICAL SCHEME

[0010] In order to achieve the above-mentioned purpose, the main technical scheme adopted by the present application includes:

[0011] In a first aspect, the present application provides a PET image scatter correction method, comprising:

[0012] Modeling the PET acquisition process;

[0013] Alternating iterative solution of radioactivity distribution and detector space attenuation coefficient based on ADMM algorithm;

[0014] Based on the deep learning network G, the mapping of PET detection data and detector space attenuation coefficient to accurate scatter distribution estimation is realized;

[0015] The log-likelihood function is maximized by using the un-fitted scatter distribution estimation and keeping the radioactivity distribution and the detector spatial attenuation coefficient constant to obtain a scatter correction factor;

[0016] The scatter-corrected PET image is obtained by using the accurate scatter distribution fitted by the detector spatial attenuation coefficient and the scatter correction factor.

[0017] Optionally, the PET acquisition process is modeled according to the following formula:

[0018]

[0019] In the formula, y = [y 1t , y 2t , …, y it , …, y NT ] represents the detected data, represents the mean value of the detection data, N represents the size of the detection data sinogram, T represents the size of the time-of-flight (TOF) discrete space, i represents the variable index of the detection data sinogram response line, and t represents the variable index of the time-of-flight (TOF) discrete space; the prime superscript represents the matrix transposition operation; x = [x1, x2, …, x j , …, x M ] represents an unknown radioactivity distribution image, M represents the size of the radioactivity distribution image space, and j represents the variable index of the radioactivity distribution image space; a = [a1, a2, …, a j , …, a N] ] represents the attenuation coefficient of the detector space; A = [A ijt ] is a system matrix; s = [s 1t , s 2t , …, s NT ] represents the estimated value of the scatter, and α = [α 1t , α 2t , …, α NT ] represents the correction factor of the scatter estimation at each position.

[0020] Optionally, the radioactivity distribution and the attenuation coefficient of the detector space are alternately iterated and solved based on the ADMM algorithm, including:

[0021] The radioactivity distribution x and the attenuation coefficient a of the detector space are alternately iterated and solved according to the formula .

[0022] Optionally, the mapping of the PET detection data and the attenuation coefficient of the detector space to the accurate scatter distribution estimation is realized based on a deep learning network G, including:

[0023] PET detection data x and detector spatial attenuation coefficient a as input, accurate scatter distribution s label As a label to train the network, optimize the network parameters θ to minimize the loss function L, and finally get the accurate scatter distribution s, as follows:

[0024]

[0025] Optionally, keeping the radioactivity distribution and the detector spatial attenuation coefficient constant, using the un-fitted scatter distribution estimate, maximizing the log-likelihood function, obtaining the scatter correction factor, including:

[0026] Keeping the radioactivity distribution and the detector spatial attenuation coefficient constant, according to the formula Get the new scatter correction factor;

[0027] Wherein,

[0028] Optionally, the trained network G includes: a CNN network, a Unet network or a GAN network.

[0029] Optionally, the log-likelihood function is:

[0030]

[0031] Secondly, the present application provides a PET image scatter correction system, comprising:

[0032] The modeling module models the PET acquisition process;

[0033] The iteration module alternately iterates to solve the radioactivity distribution and the detector spatial attenuation coefficient based on the ADMM algorithm;

[0034] The mapping module realizes the mapping of the PET detection data and the detector spatial attenuation coefficient to the accurate scatter distribution estimate based on the deep learning network G;

[0035] The scatter correction factor determination module keeps the radioactivity distribution and the detector spatial attenuation coefficient constant, uses the un-fitted scatter distribution estimate, maximizes the log-likelihood function, and obtains the scatter correction factor;

[0036] The correction module uses the accurate scatter distribution fitted by the detector spatial attenuation coefficient and the scatter correction factor to obtain the scatter corrected PET image.

[0037] Thirdly, the present application provides a computer readable storage medium, which stores a computer program, and the program is executed to realize the PET image scatter correction method of any one of the first aspect.

[0038] In a fourth aspect, the present application provides a storage device comprising a storage medium and a processor, wherein the storage medium stores a computer program, and the program is executed by the processor to implement the PET image scatter correction method of any one of the first aspect.

[0039] Advantages

[0040] The present application has the following advantages: the PET image scatter correction method of the present application uses an alternating iterative algorithm to obtain the attenuation coefficient of the detector space, then uses a deep learning network to obtain an accurate scatter estimation distribution, and finally combines the estimated scatter distribution to iteratively obtain the optimal solution for the scatter correction factor by maximum likelihood estimation, thereby avoiding the image quality degradation caused by inaccurate scatter estimation, greatly reducing the time consumption of scatter correction, and greatly improving the stability of high-noise data. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 A PET image scatter correction method flowchart is provided for the embodiments of the present application.

[0042] Figure 2 A PET reconstruction image obtained by calculating scatter using a traditional algorithm is provided for the embodiments of the present application.

[0043] Figure 3 A PET reconstruction image obtained by calculating scatter distribution using the present application is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0044] In order to better explain the present application and facilitate understanding, the present application is described in detail below with reference to the accompanying drawings and through specific embodiments.

[0045] The present application directly obtains the attenuation coefficient of the detector space from the PET data, then uses a deep learning network to obtain an accurate scatter distribution, and finally iteratively solves the optimal scatter correction factor from the PET detection data to fit the numerically accurate scatter distribution, thereby greatly reducing the time cost of scatter correction and greatly improving the accuracy of scatter correction.

[0046] Firstly, since the SSS estimation algorithm needs accurate radioactivity distribution and linear attenuation coefficient distribution for model estimation, under the initial condition, accurate radioactivity distribution information is missing, and only the PET image without scatter correction can be used to obtain the final scatter estimation through multiple iterations, so the error is large and the time consumption is relatively high. The present application directly obtains the accurate scatter distribution through a deep learning network, which reduces the dependence on the initial radioactivity and linear attenuation coefficient and greatly shortens the calculation time of scatter correction.

[0047] Secondly, the attenuation information is extracted directly from the PET acquisition data without relying on other modal imaging, which can effectively eliminate the attenuation artifacts caused by the mismatch of different modal data, thereby expanding the application range of PET imaging. By applying the PET acquisition data to obtain the detector spatial attenuation coefficient, the PET distribution and the attenuation distribution are strictly matched, the motion artifacts are effectively eliminated, and the image quality is improved.

[0048] Finally, in the process of fitting the scatter correction factor, the method fully utilizes all the data for evaluation through iterative calculation and accurately models the scatter process. Compared with the traditional tail fitting method, the method has higher tolerance and redundancy processing capability for noise in the acquisition data. The present patent only uses PET data to more accurately estimate the scatter component, thereby optimizing the PET reconstruction process. This method not only enhances the applicability and accuracy for patients with large body weight, but also realizes independence from other modal images without relying on them. At the same time, the method also improves the robustness of scatter correction to data noise, ensuring that more accurate PET images can be obtained in clinical acquisition, effectively eliminating scatter correction artifacts.

[0049] In order to better understand the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer, more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0050] In a first aspect, with reference to Figure 1 The present embodiment provides a scatter correction method for PET images, comprising:

[0051] S1, modeling the PET acquisition process.

[0052] S2, solving the radioactivity distribution and the detector spatial attenuation coefficient based on the ADMM algorithm alternating iteration.

[0053] S3, realizing the mapping of the PET detection data and the detector spatial attenuation coefficient to the accurate scatter distribution estimation based on the deep learning network G.

[0054] S4, keeping the radioactivity distribution and the detector spatial attenuation coefficient constant, and using the unfitted scatter distribution estimation to maximize the log-likelihood function to obtain the scatter correction factor.

[0055] S5, using the detector spatial attenuation coefficient and the accurate scatter distribution after fitting to obtain the scatter corrected PET image.

[0056] Alternatively, the PET acquisition process can be modeled according to the following formula:

[0057]

[0058] In the formula y=[y 1t y 2t , ..., y it , ..., y NT ]' represents the detected data. Let represent the average value of the probe data, N represent the size of the probe data sine curve, T represent the size of the Time-of-Flight (TOF) discrete space, i represent the index of the variable in the LOR (line of response) of the probe data sine curve, and t represent the index of the variable in the TOF discrete space. A single quote superscript indicates a matrix transpose operation. x = [x1, x2, ..., x...] j , ..., x M ]' represents an image of unknown radioactivity distribution, M represents the size of the radioactivity distribution image space, j represents the variable index of the radioactivity distribution image space, and represents the point source at the corresponding spatial location. a=[a1,a2,…,a j …, a N] ' represents the attenuation coefficient of the detector space. A = [A ijt [S] is the system matrix, which mathematically expresses the probability that a spatial point source j in the PET system is detected by the response line LOR i with a time-of-flight (TOF) of t, reflecting the physical characteristics of the system. 1t s 2t , ..., s NT ]' represents the estimated value of scattering, α=[α 1t α 2t , …, α NT ]' represents the correction factor for the scattering estimate at each location, used to ensure that the estimated scattering components are consistent with the scattering components of the actual acquired data.

[0059] PET detection data follows a Poisson distribution, with the unknowns being the radioactivity distribution x, the detector space attenuation coefficient a, and the scattering correction factor α. The log-likelihood function of the detection data is then expressed as:

[0060]

[0061] Substituting formula (1) into formula (2), and ignoring terms irrelevant to the unknowns, the log-likelihood function can be written as:

[0062]

[0063] The attenuation coefficient a of the detector space is estimated using PET detection data, and the influence of the scattering component is ignored, so formula (3) becomes:

[0064]

[0065] Alternatively, the radioactive activity distribution and the attenuation coefficient of the detector space are alternately iterated to solve based on the ADMM algorithm, including:

[0066] According to formula The radioactive activity distribution x and the attenuation coefficient a of the detector space are alternately iterated to solve.

[0067] The radioactive activity distribution x and the attenuation coefficient a of the detector space are alternately iterated to solve using the Alternating Direction Method of Multipliers (ADMM).

[0068] First, ensure that the attenuation coefficient a of the detector space is constant, and maximize the log-likelihood function, i.e. the traditional MLEM algorithm:

[0069]

[0070] In the formula, n represents the current iteration number.

[0071] In formula (5), the initial value of x can be set to a constant or an image obtained by other reconstruction methods or a PET image without correction. In this application, it is set to a constant distribution within the imaging field of view, and the constant is selected as 1000.

[0072] Second, ensure that the radioactive activity distribution x is constant, maximize the log-likelihood function for the unknown attenuation coefficient a of the detector space, and derive the attenuation coefficient a of the detector space, i.e. formula (6):

[0073]

[0074] Keep all variables non-negative and monotonic:

[0075]

[0076] Bring formula (6) into formula (7), i.e.

[0077]

[0078] In the formula, n represents the current iteration number.

[0079] The initial value of a in formula (8) is usually set as a constant. Since the evaluation algorithm in the formula is only related to the radioactivity distribution, the detector spatial attenuation coefficient obtained by the method is strictly matched with the radioactivity distribution.

[0080] In order to reduce the calculation time of scatter correction and increase the accuracy of scatter correction, the application applies a deep learning network to obtain an accurate scatter distribution estimation, and uses the network G to realize the mapping of the PET detection data x and the detector spatial attenuation coefficient a to the accurate scatter distribution s.

[0081] Optionally, based on the deep learning network G, the mapping of the PET detection data and the detector spatial attenuation coefficient to the accurate scatter distribution estimation is realized, and the mapping includes:

[0082] The PET detection data x and the detector spatial attenuation coefficient a are taken as inputs, and the accurate scatter distribution s label The network is trained as a label, and the network parameters θ are optimized to minimize the loss function L, and finally the accurate scatter distribution s is obtained, and the formula is as follows:

[0083]

[0084] s label The scatter distribution that meets the condition for training the network is represented, and the distribution is from the screened training data set. The training data set can come from simulation simulation or actual acquisition. The actual acquisition training data set needs to be preprocessed, and is screened to ensure that the scatter distribution is accurate. Especially for the data of a large volume target object, the situation that the scatter correction is inaccurate due to too few iterations of scatter calculation should be avoided, and the situation that the training data scatter is inaccurate due to mismatching of attenuation correction or inaccurate initial activity distribution of scatter should also be avoided. Since the tail fitting of the scatter component is easily affected by noise, the network parameters are over-fitted, and therefore the network label s label is not fitted and is directly obtained by the SSS algorithm, which improves the robustness of the network training process.

[0085] The training network G can select a CNN network, a Unet network, a GAN network or other networks.

[0086] Optionally, the radioactivity distribution and the detector spatial attenuation coefficient are kept constant, the un-fitted scatter distribution estimation is used to maximize the log-likelihood function, and the scatter correction factor is obtained, and the method includes:

[0087] The radioactivity distribution and the detector spatial attenuation coefficient are kept constant, and the new scatter correction factor is obtained according to the formula

[0088] Wherein, ​

[0089] To determine the proportional relationship between the un-fitted scattering component and the actual scattering component, it needs to be fitted, that is, to solve a suitable alpha.

[0090] Formula (1) is written in matrix form as shown in the following formula (10):

[0091] Y = A x x + S x alpha (10)

[0092] A is a system matrix, and S is a diagonal matrix As can be seen from formula (10), x and alpha are symmetrical in the formula, therefore, the unknown alpha can be solved by referring to the iterative algorithm of the unknown x, and thus formula (10) can be simplified into the following formula (11) form, that is, keeping the radioactivity distribution x constant, the attenuation coefficient a of the detector space is constant, the log-likelihood function is maximized for the unknown scattering correction factor alpha, and the new scattering correction factor alpha is calculated directly using the PET data, and the corresponding algorithm is:

[0093]

[0094] Wherein

[0095] The initial value of alpha in formula (11) can be set as a constant or a scattering correction factor distribution obtained by other fitting methods, and in the embodiment, it is set as a constant 1.

[0096] In the implementation process of the present application, firstly, to obtain the attenuation coefficient a of the detector space, the objective function is maximized for the radioactivity distribution x, that is, the traditional MLEM iterative reconstruction algorithm, and then keeping the radioactivity distribution x constant, the objective function is maximized for the unknown attenuation coefficient a of the detector space, and the attenuation coefficient a of the detector space is obtained; secondly, the deep learning network is applied, and the un-fitted scattering distribution s is obtained in combination with the radioactivity distribution x and the attenuation coefficient a of the detector space; finally, keeping the radioactivity distribution x and the attenuation coefficient a of the detector space constant, the objective function is maximized for the scattering correction factor alpha, and the un-fitted scattering distribution is iteratively calculated, the scattering correction factor is continuously corrected to approximate the real situation, the scattering correction factor meeting the requirement of the maximized objective function is obtained, and finally the matched detector space attenuation coefficient and the fitted accurate scattering estimation are applied in the reconstruction, which not only improves the accuracy of the scattering correction, but also greatly reduces the reconstruction time.

[0097] Figure 2 For the PET reconstruction image of the uniform barrel source with a radius close to the edge of the field of view calculated by using the traditional algorithm to calculate the scattering, the barrel source center is obviously high in uptake, and the scattering evaluation is inaccurate; Figure 3For calculating the scatter distribution by using the method of the patent, the PET image obtained under the same reconstruction method and parameters is improved in the non-uniformity of the barrel source, the image is more uniform, the scatter distribution is more accurate, and the image quality is better.

[0098] The method of the embodiment is a more accurate scatter correction method for PET images, which extracts a detector spatial attenuation coefficient from PET acquisition data and derives an accurate scatter distribution by means of a deep learning network. The primary advantage is that the method not only effectively overcomes the scatter estimation deviation caused by inaccurate initial activity distribution, but also avoids the significant time consumption caused by complex iterative calculation of SSS. Secondly, the method realizes independence from other modal images, so that the accurate scatter distribution can be obtained even without the support of other modal images or in the case of poor quality of these images. Furthermore, the scatter correction factor is calculated by an iterative algorithm, which solves the problem of limited stability caused by limited data volume and large patient weight. In addition, the scheme significantly reduces the radiation dose of the patient while improving the accuracy and stability of the scatter distribution estimation, thereby enhancing its practical value. Compared with the traditional scatter correction algorithm, the present application not only has higher correction accuracy, but also consumes less time, which is of great benefit to improving image quality and shortening reconstruction time.

[0099] The following is a specific implementation case of a clinical acquisition data, which further illustrates the method of the present application for scatter correction of PET images:

[0100] The patient is scanned by using a PET device, wherein the PET image is reconstructed by using an ordered subsets expectation-maximization (OSEM) method, 17 subsets, 2 iterations, an image matrix of 192x192, and a Gaussian filter half-width of 4mm.

[0101] The deep learning network model parameters: the network model structure of a feature pyramid, and a hybrid loss function of MS-SSIM loss and perceptual difference loss. The training rounds are 500 rounds, the batch size is 128, the Adam optimization operator is used to train the network, the learning rate decay value is set to 0.0001, the remaining parameters use the default value, the first five rounds use the Warm up strategy, the learning rate is linearly increased from 0.0001 to 0.0005, and in the remaining training process, the cosine decay learning rate scheduler is used, so that the learning rate gradually decays to 0.

[0102] The default initial value of x is set to 1000, the initial value of a is set to 1, the initial value of a is set to 1, the iteration number of the detector spatial attenuation coefficient is 3, the iteration number of the scatter correction factor is 3, and the reconstruction result is shown in Figure 2 and Figure 3 . Figure 2The PET reconstruction image of the uniform barrel source with a radius close to the edge of the field of view is not uniform, and the center of the barrel source has obvious high uptake, and the scatter evaluation is inaccurate; Figure 3 To calculate the scatter distribution by using the method of the present application, the PET image obtained under the same reconstruction method and parameters is obviously improved in the non-uniformity of the image of the barrel source with a large volume.

[0103] In a second aspect, the embodiment provides a PET image scatter correction system, comprising: a modeling module, modeling a PET acquisition process; an iteration module, iteratively solving a radioactivity distribution and a detector space attenuation coefficient based on an ADMM algorithm; a mapping module, realizing mapping of PET detection data and the detector space attenuation coefficient to an accurate scatter distribution estimation based on a deep learning network G; a scatter correction factor determination module, keeping the radioactivity distribution and the detector space attenuation coefficient constant, and obtaining a scatter correction factor by maximizing a log-likelihood function using an unfitted scatter distribution estimation; and a correction module, obtaining a scatter-corrected PET image by using the detector space attenuation coefficient and the accurate scatter distribution fitted after the scatter correction factor.

[0104] The PET image scatter correction system provided by the embodiment has all the technical effects of the PET image scatter correction method, and details are not repeated here.

[0105] In a third aspect, the embodiment provides a computer readable storage medium having a computer program stored thereon, and the program is executed to implement the PET image scatter correction method of any one of the first aspect.

[0106] In a fourth aspect, the embodiment provides a storage device comprising a storage medium and a processor, and the storage medium stores a computer program, and the program is executed by the processor to implement the PET image scatter correction method of any one of the first aspect.

[0107] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0108] It is apparent that those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, the present application should be construed to encompass all such modifications and variations as fall within the scope of the present claims and their equivalents.

[0109] Although the embodiments of the present application have been shown and described above, it is to be understood that the above-described embodiments are merely exemplary and are not to be taken in a limiting sense, but the scope of the present application is not to be understood to be limited to the above-described embodiments but can be variously changed, modified, replaced and varied by those skilled in the art within the scope of the present application.

Claims

1. A scattering correction method for PET images, characterized in that, include: Modeling the PET collection process; The ADMM algorithm is used to iteratively solve for the radioactivity distribution and the attenuation coefficient in the detector space. Based on the deep learning network G, a mapping is achieved from PET detection data and detector spatial attenuation coefficient to accurate scattering distribution estimation; By keeping the radioactivity distribution and detector spatial attenuation coefficient constant, the scattering correction factor is obtained by maximizing the log-likelihood function using the unfitted scattering distribution estimate. By fitting the accurate scattering distribution using the detector spatial attenuation coefficient and scattering correction factor, a scattering-corrected PET image is obtained.

2. The scattering correction method for PET images according to claim 1, characterized in that, The PET collection process is modeled based on the following formula: In the formula y=[y 1t y 2t , ..., y it , ..., y NT ]' represents the detected data. The value represents the average value of the probe data; N represents the size of the probe data sine curve; T represents the size of the time-of-flight (TOF) discrete space; i represents the variable index of the response line of the probe data sine curve; t represents the variable index of the time-of-flight (TOF) discrete space; a single quote superscript indicates a matrix transpose operation; x = [x1, x2, ..., x...]. j , ..., x M ]' represents an image of unknown radioactivity distribution, M represents the size of the radioactivity distribution image space, and j represents the variable index of the radioactivity distribution image space; a = [a1, a2, ..., a j …, a N] ' represents the attenuation coefficient of the detector space; A = [A ijt [S] is the system matrix; S = [s] 1t s 2t , ..., s NT ]' represents the estimated value of scattering, α=[α 1t α 2t , …, α NT ]' represents the correction factor for the scattering estimate at each location.

3. The scattering correction method for PET images according to claim 2, characterized in that, The ADMM algorithm is used to iteratively solve for the radioactivity distribution and the attenuation coefficient in the detector space, including: According to the formula The radioactivity distribution x and the detector spatial attenuation coefficient a are solved by alternating iterations.

4. The scattering correction method for PET images according to claim 3, characterized in that, Based on a deep learning network G, a mapping is achieved from PET detection data and detector spatial attenuation coefficients to accurate scattering distribution estimation, including: Using PET detection data x and detector spatial attenuation coefficient a as inputs, accurate scattering distribution s label The network is trained using labels, and the network parameters θ are optimized to minimize the loss function L, ultimately yielding the accurate scattering distribution s, as shown in the following formula:

5. The scattering correction method for PET images according to claim 4, characterized in that, Keeping the radioactivity distribution and detector spatial attenuation coefficient constant, the scattering correction factor is obtained by maximizing the log-likelihood function using the unfitted scattering distribution estimate, including: To maintain the radioactivity distribution and the attenuation coefficient of the detector space as constants, according to the formula... A new scattering correction factor is obtained; in, 6. The scattering correction method for PET images according to claim 5, characterized in that, The training network G includes: CNN network, Unet network, or GAN network.

7. The scattering correction method for PET images according to claim 6, characterized in that, The log-likelihood function is:

8. A scattering correction system for PET images, characterized in that, include: The modeling module models the PET acquisition process; The iterative module uses the ADMM algorithm to iteratively solve for the radioactivity distribution and the attenuation coefficient in the detector space. The mapping module, based on the deep learning network G, realizes the mapping of PET detection data and detector spatial attenuation coefficient to accurate scattering distribution estimation; The scattering correction factor determination module keeps the radioactivity distribution and detector spatial attenuation coefficient constant, and uses the unfitted scattering distribution estimate to maximize the log-likelihood function to obtain the scattering correction factor; The correction module uses the detector spatial attenuation coefficient and the scattering correction factor to fit the accurate scattering distribution, and obtains a scattering-corrected PET image.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the scattering correction method for a PET image as described in any one of claims 1 to 7.

10. A storage device comprising a storage medium and a processor, the storage medium storing a computer program, characterized in that, When the processor executes the computer program, it implements the scattering correction method for a PET image as described in any one of claims 1 to 7.

Images