CT scattering correction method and device, electronic equipment and storage medium

Abstract

The invention discloses a CT scatter correction method and device, electronic equipment and a storage medium. The method comprises the steps that CBCT projections of multiple part cases are acquired; performing Monte Carlo simulation on CBCT projection based on interaction of photon energy to obtain scattering distribution corresponding to the CBCT projection; the CBCT projection is used as model input, scattering distribution is used as a training label, optimization training is carried out on a pre-constructed residual convolutional neural network, and a target network is obtained after training is completed; and performing iterative scatter correction on the to-be-processed CBCT projection by using the target network to obtain a corrected CBCT image. According to the embodiment of the invention, the high-quality training label is obtained through the Monte Carlo method, the residual convolutional neural network is embedded into the CBCT iterative scatter correction framework, so that the time cost of iteration is reduced, the CBCT scatter removal image meeting the clinical requirements is obtained, CT scatter correction can be accurately carried out, and the CBCT scatter correction method can be widely applied to the technical field of image processing.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a CT scatter correction method, device, electronic equipment and storage medium. Background Art

[0002] Cone beam computed tomography (CBCT) is the imaging technology foundation for online adaptive radiation therapy (ART). CBCT images provide information about changes in the patient's physiological or anatomical structure before treatment. Re-optimizing the dose distribution through fusion of CBCT and CT can make real-time dynamic adjustments to the treatment plan. Although CBCT has its unique imaging advantages, its physical imaging characteristics will lead to a particularly prominent problem of photon scattering during imaging. Scattered rays will lead to a decrease in the contrast of soft tissues in the image, inaccurate CT values, and the introduction of cupping and streak artifacts. Due to the poor image quality, the current application of CBCT in radiotherapy is limited to correcting positioning errors. The presence of scattering contamination poses great challenges to CBCT-based target delineation, image registration, and dose calculation during radiotherapy. Therefore, how to correct the scattering artifacts of CBCT images has become the key to further promoting the in-depth application of CBCT in ART. Summary of the invention

[0003] The main purpose of the embodiments of the present invention is to provide a CT scatter correction method, device, electronic device and storage medium, in order to solve at least one problem of the prior art. The present invention can accurately perform CT scatter correction.

[0004] To achieve the above object, an embodiment of the present invention provides a CT scatter correction method, the method comprising:

[0005] Obtain CBCT projections of multiple site cases;

[0006] Monte Carlo simulation of CBCT projections is performed based on the interaction of photon energy to obtain the scattering distribution corresponding to the CBCT projections;

[0007] The CBCT projection is used as the model input, and the scatter distribution is used as the training label. The pre-built residual convolutional neural network is optimized and trained. After the training is completed, the target network is obtained.

[0008] The target network is used to perform iterative scatter correction on the processed CBCT projections to obtain the corrected CBCT images.

[0009] In some embodiments, the types of interaction include photoelectric effect, Compton effect and Rayleigh scattering; performing Monte Carlo simulation on the CBCT projection based on the interaction of photon energy to obtain the scattering distribution corresponding to the CBCT projection includes the following steps:

[0010] The random number is obtained by uniformly sampling the sampling interval within the preset range;

[0011] The types of interaction of CBCT projections are sampled based on random numbers to obtain sampling results;

[0012] Determine whether the sampling result is a Compton effect; when the sampling result is a Compton effect, sample the scattering direction and the photon residual energy based on the differential scattering cross section, and when the photon residual energy is less than a preset energy threshold, obtain the scattering distribution corresponding to the CBCT projection according to the pre-loaded detector response curve, otherwise, return to the step of uniformly sampling a sampling interval within a preset range to obtain a random number until the photon residual energy is less than the preset energy threshold or the sampling result is not a Compton effect; wherein the differential scattering cross section is physical data associated with the CBCT projection;

[0013] When the sampling result is not Compton effect, determine whether the sampling result is Rayleigh scattering; when the sampling result is Rayleigh scattering, sample the scattering direction based on the differential scattering cross section, and then return to execute the step of uniformly sampling the sampling interval within the preset range to obtain a random number; when the sampling result is not Rayleigh scattering, determine that the sampling result is photoelectric effect, absorb and deposit photon energy based on the differential scattering cross section, and then obtain the scattering distribution corresponding to the CBCT projection according to the pre-loaded detector response curve.

[0014] In some embodiments, before the step of inputting the CBCT projections as a model, the method further comprises the following steps:

[0015] Perform data enhancement processing on CBCT projections;

[0016] The result of data enhancement processing is used as model input, and the data enhancement processing includes rotation processing and flipping processing.

[0017] In some embodiments, the method further comprises the following steps:

[0018] A coding path is constructed based on a first preset number of coding modules; wherein a downsampling operation and a maximum pooling operation are provided at the output of each coding module in the coding path;

[0019] A decoding path is constructed based on a second preset number of decoding modules; wherein an upsampling operation is provided at the input of each decoding module in the decoding path;

[0020] Among them, both the encoding module and the decoding module are constructed based on the residual block, which includes a convolutional layer, batch normalization processing and ReLU activation operation.

[0021] In some embodiments, the CBCT projection is used as a model input, the scatter distribution is used as a training label, and the pre-built residual convolutional neural network is optimized and trained. After the training is completed, a target network is obtained, including the following steps:

[0022] The training set and the test set are obtained according to the CBCT projection; the residual convolutional neural network is used as the first network;

[0023] Inputting the training data in the training set into the first network for processing to obtain a first scattering distribution;

[0024] Performing loss calculation based on the first scattering distribution and the training labels of the scattering distribution corresponding to the training data, and adjusting the network parameters of the first network according to the result of the loss calculation to obtain the second network;

[0025] The test set is used to perform error evaluation on the second network. When the result of the error evaluation is less than a preset error threshold, the second network is used as the target network after training. Otherwise, the second network is used as the first network, and then the step of inputting the training data in the training set into the first network to obtain the first scattering distribution is returned to the execution until the target network is obtained.

[0026] In some embodiments, iterative scatter correction is performed on the processed CBCT projections using the target network to obtain a corrected CBCT image, including the following steps:

[0027] The target network is used to process the CBCT projection to be processed to obtain a second scattering distribution, and the CBCT projection to be processed is iteratively corrected based on the second scattering distribution until a preset condition is met to obtain a corrected CBCT image.

[0028] In some embodiments, a target network is used to process a CBCT projection to be processed to obtain a second scattering distribution, and the CBCT projection to be processed is iteratively corrected based on the second scattering distribution until a preset condition is met to obtain a corrected CBCT image, including the following steps:

[0029] Taking the CBCT projection to be processed as the input projection;

[0030] The input projection is input into the target network for processing to obtain a second scattering distribution;

[0031] Correcting the scattering signal deviation of the input projection based on the second scattering distribution to obtain a corrected projection;

[0032] The corrected projection is used as the input projection, and then the step of inputting the input projection into the target network for processing to obtain the second scattering distribution is returned to execute until the difference between the corrected projections of two adjacent iterative processes is less than a preset difference threshold; image reconstruction is performed based on the corrected projection finally obtained to obtain a corrected CBCT image.

[0033] To achieve the above object, another aspect of an embodiment of the present invention provides a CT scatter correction device, the device comprising:

[0034] The first module is used to obtain CBCT projections of cases with multiple parts;

[0035] The second module is used to perform Monte Carlo simulation on the CBCT projection based on the interaction of photon energy to obtain the scattering distribution corresponding to the CBCT projection;

[0036] The third module is used to use the CBCT projection as the model input and the scattering distribution as the training label to optimize the pre-built residual convolutional neural network and obtain the target network after the training is completed;

[0037] The fourth module is used to perform iterative scatter correction on the CBCT projection to be processed using the target network to obtain a corrected CBCT image.

[0038] In some embodiments, the apparatus further comprises:

[0039] The fifth module is used to perform data enhancement processing on CBCT projections;

[0040] Among them, the fifth module is executed before the third module, and the result of data enhancement processing is used as model input, and the data enhancement processing includes rotation processing and flipping processing.

[0041] In some embodiments, the apparatus further comprises:

[0042] A sixth module is used to construct a coding path based on the first preset number of coding modules; wherein a downsampling operation and a maximum pooling operation are provided at the output of each coding module in the coding path;

[0043] A seventh module is used to construct a decoding path based on a second preset number of decoding modules; wherein an upsampling operation is provided at the input of each decoding module in the decoding path;

[0044] Among them, both the encoding module and the decoding module are constructed based on the residual block, which includes a convolutional layer, batch normalization processing and ReLU activation operation.

[0045] To achieve the above objective, another aspect of an embodiment of the present invention provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above method when executing the computer program.

[0046] To achieve the above objective, another aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above method is implemented.

[0047] The embodiments of the present invention include at least the following beneficial effects: The present invention provides a CT scatter correction method, device, electronic device and storage medium. The scheme obtains CBCT projections of cases in multiple parts; performs Monte Carlo simulation on the CBCT projections based on the interaction of photon energy to obtain the scatter distribution corresponding to the CBCT projections; uses the CBCT projections as model inputs and the scatter distribution as training labels to optimize and train the pre-constructed residual convolutional neural network, and obtains the target network after the training is completed; uses the target network to perform iterative scatter correction on the CBCT projections to be processed to obtain a corrected CBCT image. The embodiments of the present invention collect CBCT projections of cases in various parts, combine the Monte Carlo method to obtain high-quality training labels, and then realize accurate scatter distribution estimation based on the residual network. Specifically, the residual convolutional neural network is embedded in the CBCT iterative scatter correction framework to reduce the time cost of iteration and obtain a CBCT de-scattered image that meets clinical requirements. The embodiments of the present invention can accurately perform CT scatter correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a flow chart of a CT scatter correction method provided by an embodiment of the present invention;

[0049] Figure 2 is a schematic diagram of a Monte Carlo scattering simulation process provided by an embodiment of the present invention;

[0050] Figure 3 is a schematic diagram of a cone-beam CT imaging geometric model and its related data curves provided by an embodiment of the present invention;

[0051] Figure 4 A schematic diagram of the ResDCNN model structure provided by an embodiment of the present invention;

[0052] Figure 5 is a schematic diagram of a CBCT iterative scatter correction framework provided by an embodiment of the present invention;

[0053] Figure 6 A schematic diagram of the structure of a CT scatter correction device provided in an embodiment of the present invention;

[0054] Figure 7 It is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present invention. They are only examples of devices and methods consistent with some aspects of the embodiments of the present invention as detailed in the attached claims.

[0056] It is understood that the terms "first", "second", etc. used in the present invention may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0057] The terms "at least one", "multiple", "each", "any", etc. used in the present invention, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0059] To facilitate the understanding of the technical solution of the present invention, the technical terms in the proprietary field that may be applied in the embodiments of the present invention are explained:

[0060] Cone-beam Computed Tomography (CBCT);

[0061] Adaptive Radiotherapy (ART);

[0062] Signal-to-Noise Ratio (SNR);

[0063] Monte Carlo: Monte Carlo (MC);

[0064] Deep Convolutional Neural Network (DCNN);

[0065] Residual Deep Convolutional Neural Network (ResDCNN);

[0066] The CT scatter correction method provided by the embodiment of the present invention relates to the field of image processing technology. The CT scatter correction method provided by the embodiment of the present invention can be applied to a terminal, can be applied to a server, or can be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a vehicle-mounted terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements the CT scatter correction method, etc., but is not limited to the above forms.

[0067] The present invention can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0068] Figure 1 is an optional flow chart of the CT scatter correction method provided in an embodiment of the present invention. Figure 1 The method may include but is not limited to steps S100 to S400.

[0069] S100, obtaining CBCT projections of cases at multiple locations;

[0070] For example, in some specific implementations, CBCT projections and images of cases of various parts of the human body can be acquired.

[0071] S200, performing a Monte Carlo simulation on the CBCT projection based on the interaction of photon energy to obtain a scattering distribution corresponding to the CBCT projection;

[0072] It should be noted that the types of interaction include photoelectric effect, Compton effect and Rayleigh scattering; in some embodiments, step S200 may include: obtaining a random number by uniformly sampling a sampling interval within a preset range; sampling the type of interaction of the CBCT projection based on the random number to obtain a sampling result; determining whether the sampling result is a Compton effect; when the sampling result is a Compton effect, sampling the scattering direction and the photon residual energy based on the differential scattering cross section, and when the photon residual energy is less than a preset energy threshold, obtaining the scattering distribution corresponding to the CBCT projection according to the pre-loaded detector response curve, otherwise, returning to execute uniform sampling through the sampling interval within the preset range. The step of obtaining a random number is performed until the residual energy of the photon is less than a preset energy threshold or the sampling result is not a Compton effect; wherein the differential scattering cross section is a physical data associated with the CBCT projection; when the sampling result is not a Compton effect, determining whether the sampling result is Rayleigh scattering; when the sampling result is Rayleigh scattering, sampling the scattering direction based on the differential scattering cross section, and then returning to execute the step of obtaining a random number by uniformly sampling a sampling interval within a preset range; when the sampling result is not Rayleigh scattering, determining that the sampling result is a photoelectric effect, absorbing and depositing photon energy based on the differential scattering cross section, and then obtaining the scattering distribution corresponding to the CBCT projection according to the pre-loaded detector response curve.

[0073] For example, in some specific embodiments, constructing an accurate imaging physics model is crucial to obtaining accurate scattering estimates. The geometric model can be described by an xyz-ISO world coordinate system with the center of the imaging object as the origin and a coordinate system with the center of the detector as the origin. When the projection angle is 0 degrees, the x-axis and y-axis of the world coordinate system are parallel to the v-axis and u-axis of the detector, respectively. The z-axis of the world coordinate system points to the midpoint O of the detector and points in the opposite direction to the focus of the ray source. In the MC simulation, the source-to-Axis Distance (SAD) is set to 100cm, and the source-to-detector distance (SID) is set to 150cm. The size of the detector is 40cm×30cm, and the pixel size is 512×384. In actual scanning, the photon energy emitted by the X-ray source is not single, but varies continuously within a certain energy range. The multi-energy spectrum curve describes the distribution of different photon energies in the form of probability. For example Figure 2As shown in the figure, different energy segments can be divided into groups for simulation during the MC simulation process, and the number of sampled photons is proportional to the probability of the energy segment. The number of simulated photons is 1×10 9 , the simulation angle range is 360°, and the angle interval is 1°. The whole process of photon transport is simulated after the X-ray source emits along the direction. Move to next position It is necessary to determine whether the location of the photon is inside the object. If the photon is inside the object, it is necessary to sample the type of interaction between the photon and the material. Since the photon energy of CBCT is generally in the kV range, only the photoelectric effect, Compton effect and Rayleigh scattering are considered in the simulation process. Assume that the probability of a photon undergoing a certain reaction is P i =σ i / σt(i=1,2,3). Among them, σt is the total microscopic cross section, and σi is the microscopic cross section where a certain interaction occurs. When sampling the action type, first uniformly sample the random number ξ1 from the interval [0,1], and then use ξ1 to determine which probability interval it falls in to determine the type of reaction. If a photoelectric effect occurs, all the energy of the photon will be deposited at the point of action, so it will not be recorded by the detector. If a Compton or Rayleigh reaction occurs, it is necessary to further sample the differential cross section (DCS) to determine the scattering angle. Keep tracking the photon until it leaves the Morti, and finally calculate the size of the scattering signal at each position of the detector based on the loaded detector response curve to obtain the final scattering distribution.

[0074] Specifically, CBCT images are the objects of MC simulation to calculate the scattering distribution to obtain high-quality training labels. In order to obtain a scattering distribution that is closer to the actual situation, a series of real physical data needs to be loaded in the initial stage, including differential scattering cross-section data, linear attenuation coefficient, material properties, etc. In addition, geometric modeling, detector modeling, and energy spectrum modeling are also required to build a CBCT imaging model that is consistent with the actual situation, such as Figure 3 As shown (where Figure 3 In the figure, the top represents the cone beam CT imaging geometry model; the middle left represents the X-ray source energy spectrum curve; the middle right represents the photon fluence diagram; the lower left represents the CT value-density curve; the lower right represents the detector response curve). When the photon is emitted from the X-ray source, its initial state needs to be determined, so the energy and emission direction of the X-ray source need to be sampled. Since the photon energy of CBCT is generally in the kV range, only the photoelectric effect, Compton effect and Rayleigh scattering are considered in the simulation process. Assume that the probability of a photon undergoing a certain reaction is P i =σ i / σ t(i=1, 2, 3), where σ t is the total microscopic cross section, σ i is the microscopic cross section where a certain interaction occurs. When sampling the type of action, first uniformly sample the random number ξ1 from the interval [0,1], and then use ξ1 to determine which probability interval it falls in to determine the type of reaction. If a Compton or Rayleigh reaction occurs, it is necessary to further sample the differential scattering cross section to determine the scattering angle (scattering direction) and calculate the residual energy of the photon. If the kinetic energy of the photon during transportation is lower than the preset energy threshold, it is considered that the energy of the photon is insufficient to penetrate the material and its tracking is terminated. Finally, the scattering signal size of each position point of the detector will be calculated based on the loaded detector response curve to obtain the final scattering distribution.

[0075] S300, using the CBCT projection as a model input and the scattering distribution as a training label, optimizing and training the pre-built residual convolutional neural network, and obtaining a target network after the training is completed;

[0076] In some embodiments, the method may further include the following steps: performing data enhancement processing on the CBCT projection; wherein the result of the data enhancement processing is used as a model input, and the data enhancement processing includes rotation processing and flipping processing.

[0077] In some embodiments, the method may further include the following steps: constructing an encoding path based on a first preset number of encoding modules; wherein a downsampling operation and a maximum pooling operation are provided at the output of each encoding module in the encoding path; constructing a decoding path based on a second preset number of decoding modules; wherein an upsampling operation is provided at the input of each decoding module in the decoding path; wherein both the encoding module and the decoding module are constructed based on a residual block, and the residual block includes a convolutional layer, a batch normalization process, and a ReLU activation operation.

[0078] It should be noted that, in some embodiments, step S300 may include: obtaining a training set and a test set according to the CBCT projection arrangement; using the residual convolutional neural network as the first network; inputting the training data in the training set into the first network for processing to obtain a first scattering distribution; performing loss calculation based on the first scattering distribution and the training labels of the scattering distribution corresponding to the training data, and adjusting the network parameters of the first network according to the result of the loss calculation to obtain a second network; performing error evaluation on the second network using the test set, and when the result of the error evaluation is less than a preset error threshold, using the second network as the target network after training, otherwise, using the second network as the first network, and then returning to execute the step of inputting the training data in the training set into the first network for processing to obtain the first scattering distribution, until the target network is obtained.

[0079] For example, in some specific implementations, the construction and training of the ResDCNN network model (i.e., residual convolutional neural network) can be implemented as follows:

[0080] Collect CBCT projections of patients undergoing radiotherapy at various parts of the body and rotate and flip the data. Randomly rotate a specific angle around the center point in the range of 0 to 360 degrees, and flip it horizontally and vertically with a certain probability to expand the data set to improve the generalization ability of the model. The above data is divided into training set, validation set and test set in a ratio of 6:2:2. The present invention will design a ResDCNN network that combines DCNN and residual blocks, such as Figure 4 As shown. The encoding path of the left half of the network consists of 4 modules, each of which includes two 3×3 convolution layers, batch normalization, and ReLU activation operations. The features output by each module are downsampled, and a 2×2 maximum pooling operation is used to further extract important features and reduce the image resolution. The decoding path of the network also includes four modules, each of which includes two 3×3 convolution layers, batch normalization, ReLU activation, and upsampling operations. The problem of gradient disappearance is solved by adding residual blocks to the encoding path and decoding path to improve the mapping accuracy. In order to speed up the convergence of the network, the output of the convolution layer is batch normalized. The mean square error damage function L is used. MSE and the structural similarity index loss function L SSIM Combined as a hybrid loss function L hybrid , the calculation formula is as follows:

[0081] K hybrid =λ1K MSE +λ2K SSIM ;

[0082]

[0083] Where λ1 and λ2 are equilibrium constants, N is the number of pixels in the scattering image; μ I and They are the scatter distribution image I and the network output scatter estimation The mean of σ I , are their covariance and variance respectively; ε1 and ε2 are minimum constants, set to 10 -3 and 10 -4 .

[0084] S400, performing iterative scatter correction on the CBCT projection to be processed using the target network to obtain a corrected CBCT image;

[0085] It should be noted that, in some embodiments, step S400 may include: using the target network to process the CBCT projection to be processed to obtain a second scattering distribution, and iteratively correcting the CBCT projection to be processed based on the second scattering distribution until a preset condition is met to obtain a corrected CBCT image.

[0086] In some embodiments, the target network is used to process the CBCT projection to be processed to obtain a second scattering distribution, and the CBCT projection to be processed is iteratively corrected based on the second scattering distribution until a preset condition is met to obtain a corrected CBCT image, which may include: using the CBCT projection to be processed as an input projection; inputting the input projection into the target network for processing to obtain a second scattering distribution; correcting the scattering signal deviation of the input projection based on the second scattering distribution to obtain a corrected projection; using the corrected projection as the input projection, and then returning to execute the step of inputting the input projection into the target network for processing to obtain a second scattering distribution, until the difference between the corrected projections of two adjacent iterative processes is less than a preset difference threshold; and reconstructing an image based on the corrected projection finally obtained to obtain a corrected CBCT image.

[0087] For example, in some specific embodiments, Figure 5 As shown in the figure, the original CBCT projection can be used as the input of ResDCNN to obtain the initial scattering distribution. The initial scattering distribution is subtracted from the original projection to obtain the corrected projection, and then the reconstruction is performed to obtain the first corrected CBCT image. Since the initial scattering distribution is inaccurate, the deviation of the scattering signal is continuously corrected through iteration. During the iteration process, the updated projection is used as the input of ResDCNN. If the mean absolute error between the current scattering estimate and the previous scattering estimate is less than the threshold of 0.001, the scattering signal can be considered to converge and the iteration process is terminated. The corrected CBCT image can be obtained by performing FDK reconstruction on the final corrected projection. Among them, the FDK (Feldkamp-Davis-Kress) algorithm is an algorithm widely used in computed tomography (CT) image reconstruction.

[0088] In order to explain the principle of the technical solution of the present invention in detail, the overall process of the present invention is described below in combination with some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present invention and cannot be regarded as a limitation of the present invention.

[0089] First of all, it should be noted that in order to solve the scattering problem of CBCT, various scattering correction methods have been proposed one after another. Hardware methods include collimator method, anti-scattering grid method, beam stop array technology, etc. However, in order to keep the signal-to-noise ratio (SNR) of the image unchanged, these methods will increase the patient's exposure dose. The more representative algorithmic methods include scattering kernel deconvolution method and Monte Carlo (MC) simulation method. The effect of CBCT scattering correction is largely dependent on the accuracy of scattering estimation. However, the scattering kernel design is too ideal, and the calculation accuracy is limited when the tissue differences are large. The MC method is the gold standard for calculating particle transport. It obtains accurate scattering distribution by simulating the interaction between a large number of photons and matter, but the MC simulation efficiency is low. In addition, due to the inaccurate CT value of CBCT images containing scattering, it is necessary to perform iterative MC simulation to obtain accurate scattering distribution for scattering correction without relying on planned CT, which further increases the time cost based on the MC method. In recent years, deep learning methods have achieved remarkable success in the field of computer vision, providing new ideas for realizing scattering correction of CBCT. For example, by using methods such as Deep Convolutional Neural Network (DCNN) and Generative Adversarial Network (GAN), the scattering artifacts of CBCT images can be removed by learning and training the mapping relationship between the projection domain or the image domain. Once the model is trained, the target projection or image can be quickly obtained, which is expected to be applied in real time in clinical practice. However, most of the current deep learning-based methods rely on paired CBCT / CT images to indirectly generate labels. Due to the inconsistent acquisition time of CT and CBCT, it is difficult to ensure the accuracy of the registration of the two when the target area or surrounding organs undergo large deformation during treatment, and the model performance has certain limitations.

[0090] Based on the above analysis, the present invention will solve the quality and efficiency problems of CBCT scatter correction from the following two aspects: 1) Collect CBCT projections of cases in various parts, combine the MC method to obtain high-quality training labels, and design a residual CNN network ResDCNN to achieve accurate scatter distribution estimation; 2) Embed ResDCNN into the CBCT iterative scatter correction framework to reduce the time cost of iteration and obtain CBCT de-scattered images that meet clinical requirements.

[0091] In view of this, in some specific implementations, the present invention can be implemented as follows:

[0092] 1. Acquisition of MC high-quality training labels:

[0093] Building an accurate imaging physical model is crucial to obtain accurate scattering estimates. The geometric model can be described by an xyz-ISO world coordinate system with the center of the imaging object as the origin and a coordinate system with the center of the detector as the origin. When the projection angle is 0 degrees, the x-axis and y-axis of the world coordinate system are parallel to the v-axis and u-axis of the detector, respectively. The z-axis of the world coordinate system points to the midpoint O of the detector and points in the opposite direction to the focus of the ray source. In the MC simulation, the source-to-axis distance (SAD) is set to 100 cm, and the source-to-imager distance (SID) is set to 150 cm. The size of the detector is 40 cm × 30 cm, and the pixel size is 512 × 384. In actual scanning, the photon energy emitted by the X-ray source is not single, but varies continuously within a certain energy range. The multi-energy spectrum curve describes the distribution of different photon energies in the form of probability. In the MC simulation process, different energy segments can be divided for group simulation, and the number of sampled photons is proportional to the probability of the energy segment. The number of simulated photons is 1×10 9 , the simulation angle range is 360°, and the angle interval is 1°. The whole process of photon transport is simulated after the X-ray source emits along the direction. Move to next position It is necessary to determine whether the location of the photon is inside the object. If the photon is inside the object, it is necessary to sample the type of interaction between the photon and the material. Since the photon energy of CBCT is generally in the kV range, only the photoelectric effect, Compton effect and Rayleigh scattering are considered in the simulation process. Assume that the probability of a photon undergoing a certain reaction is P i =σ i / σ t (i=1,2,3). Where σt is the total microscopic cross section, and σi is the microscopic cross section where a certain interaction occurs. When sampling the action type, first uniformly sample the random number ξ1 from the interval [0,1], and then use ξ1 to determine which probability interval it falls in to determine the type of reaction. If a photoelectric effect occurs, all the energy of the photon will be deposited at the point of action, so it will not be recorded by the detector. If a Compton or Rayleigh reaction occurs, it is necessary to further sample the differential scattering section (DCS) to determine the scattering angle. Keep tracking the photon until it leaves the Morte, and finally calculate the size of the scattering signal at each position of the detector based on the loaded detector response curve to obtain the final scattering distribution.

[0094] 2. ResDCNN network realizes scattering distribution estimation:

[0095] CBCT projections of patients undergoing radiotherapy in various parts of the body were collected, including the head and neck, chest and abdomen, and pelvic area, and the corresponding models were trained for the data of these parts. First, the data was rotated and flipped, randomly rotated around the center point at a specific angle in the range of 0 to 360 degrees, and flipped in the horizontal and vertical directions to expand the data set to improve the generalization ability of the model. The above data were divided into training set, validation set, and test set in a ratio of 6:2:2. The problem of gradient disappearance was solved by adding residual blocks to the encoding path and decoding path to improve the mapping accuracy. In order to speed up the convergence of the network, the output of the convolutional layer was batch normalized. The mean square error MSE was used as the loss function, and the original CBCT scatter projection was used as the input of the network. During the training process of the neural network, the Adam optimizer was used for optimization, the batch-size was set to 32, and the initial learning rate was set to 10 -5 .

[0096] 3. Iterative scatter correction:

[0097] The original CBCT projection is used as the input of the trained ResDCNN network to obtain the initial scattering distribution. The initial scattering distribution is subtracted from the original projection to obtain the corrected projection. The first corrected CBCT image can be obtained by performing FDK reconstruction. Since the initial scattering distribution is inaccurate, the deviation of the scattering signal is continuously corrected through iteration. During the iteration process, the updated projection is used as the input of ResDCNN. If the difference between the current scattering estimate and the previous scattering estimate is less than the threshold of 0.001, the scattering signal can be considered to have converged and the iteration process is terminated. The corrected CBCT image can be obtained by performing FDK reconstruction on the final corrected projection.

[0098] In summary, compared with the prior art, the present invention has at least the following beneficial effects:

[0099] (1) Most of the deep models currently studied rely on paired CBCT / CT images to indirectly generate labels. Due to the inconsistent acquisition time of CT and CBCT, it is difficult to ensure the accuracy of the registration of the two when the target area or surrounding organs undergo large deformation during the treatment process, and the model performance has certain limitations. The present invention combines the MC method to obtain high-quality training labels and designs a DCNN residual network to realize scattering distribution estimation. (2) Without relying on the planned CT, the corrected cone-beam CT reconstructed image will cause inaccurate initial MC simulation results due to the scattering artifacts it contains. Previous studies have shown that iterative correction is required to effectively remove scattering, which undoubtedly increases the time cost. To this end, the present invention embeds the DCNN residual network into the CBCT iterative scattering correction framework to achieve rapid calculation of scattering signals to reduce the time cost of iteration and obtain CBCT de-scattering images that meet clinical requirements.

[0100] See also Figure 6 The embodiment of the present invention further provides a CT scatter correction device 900, which can implement the above CT scatter correction method, and the device includes:

[0101] The first module 910 is used to obtain CBCT projections of multiple parts of the case;

[0102] The second module 920 is used to perform Monte Carlo simulation on the CBCT projection based on the interaction of photon energy to obtain the scattering distribution corresponding to the CBCT projection;

[0103] The third module 930 is used to use the CBCT projection as a model input and the scatter distribution as a training label to optimize the pre-built residual convolutional neural network and obtain the target network after the training is completed;

[0104] The fourth module 940 is used to perform iterative scatter correction on the CBCT projection to be processed using the target network to obtain a corrected CBCT image.

[0105] In some embodiments, the apparatus may further include:

[0106] The fifth module is used to perform data enhancement processing on CBCT projections;

[0107] Among them, the fifth module is executed before the third module, and the result of data enhancement processing is used as model input, and the data enhancement processing includes rotation processing and flipping processing.

[0108] In some embodiments, the apparatus may further include:

[0109] A sixth module is used to construct a coding path based on the first preset number of coding modules; wherein a downsampling operation and a maximum pooling operation are provided at the output of each coding module in the coding path;

[0110] A seventh module is used to construct a decoding path based on a second preset number of decoding modules; wherein an upsampling operation is provided at the input of each decoding module in the decoding path;

[0111] Among them, both the encoding module and the decoding module are constructed based on the residual block, which includes a convolutional layer, batch normalization processing and ReLU activation operation.

[0112] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0113] The embodiment of the present invention further provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above CT scatter correction method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0114] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0115] See also Figure 7 , Figure 7 The hardware structure of an electronic device 1000 of another embodiment is illustrated. The electronic device includes:

[0116] The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention;

[0117] The memory 1002 may be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 may store an operating system and other application programs. When the technical solution provided in the embodiments of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the CT scatter correction method of the embodiment of the present invention.

[0118] Input / output interface 1003, used to implement information input and output;

[0119] The communication interface 1004 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0120] A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 );

[0121] The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0122] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned CT scatter correction method is implemented.

[0123] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0124] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0125] The CT scatter correction method, CT scatter correction device, electronic device and storage medium provided by the embodiment of the present invention obtain CBCT projections of cases in multiple parts; perform Monte Carlo simulation on the CBCT projections based on the interaction of photon energy to obtain the scatter distribution corresponding to the CBCT projections; use the CBCT projections as model inputs and the scatter distributions as training labels to optimize and train the pre-constructed residual convolutional neural network, and obtain the target network after the training is completed; use the target network to perform iterative scatter correction on the CBCT projections to be processed to obtain a corrected CBCT image. The embodiment of the present invention collects CBCT projections of cases in various parts, combines the Monte Carlo method to obtain high-quality training labels, and then realizes accurate scatter distribution estimation based on the residual network, specifically embeds the residual convolutional neural network into the CBCT iterative scatter correction framework to reduce the time cost of iteration and obtain a CBCT de-scattered image that meets clinical requirements. The embodiment of the present invention can accurately perform CT scatter correction.

[0126] The embodiments described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. Those skilled in the art can appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0127] Those skilled in the art will appreciate that the technical solutions shown in the figures do not limit the embodiments of the present invention and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0128] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0129] Those skilled in the art will appreciate that all or some of the steps, devices, and functional modules / units in the above disclosed methods may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0130] The terms "first", "second", "third", "fourth", etc. (if any) in the description of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0131] It should be understood that in the present invention, "at least one (item)" refers to one or more, and "plurality" refers to two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can represent: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0132] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0133] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0134] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0135] 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 the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including multiple instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store programs.

[0136] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the embodiments of the present invention is not limited thereby. Any modification, equivalent substitution and improvement made by those skilled in the art without departing from the scope and essence of the embodiments of the present invention shall be within the scope of the rights of the embodiments of the present invention.

Claims

1. A CT scatter correction method, It is characterized in that The method comprises the following steps: Obtain CBCT projections of multiple site cases; performing a Monte Carlo simulation on the CBCT projection based on the interaction of photon energy to obtain a scattering distribution corresponding to the CBCT projection; Using the CBCT projection as a model input and the scattering distribution as a training label, optimizing and training the pre-constructed residual convolutional neural network, and obtaining a target network after the training is completed; The target network is used to perform iterative scatter correction on the CBCT projection to be processed to obtain a corrected CBCT image.

2. The method according to claim 1, It is characterized in that The types of interaction include photoelectric effect, Compton effect and Rayleigh scattering; the Monte Carlo simulation of the CBCT projection based on the interaction of photon energy to obtain the scattering distribution corresponding to the CBCT projection includes the following steps: The random number is obtained by uniformly sampling the sampling interval within the preset range; Sampling the types of the interactions of the CBCT projections based on the random numbers to obtain sampling results; Determine whether the sampling result is the Compton effect; when the sampling result is the Compton effect, sample the scattering direction and the photon residual energy based on the differential scattering cross section, and when the photon residual energy is less than a preset energy threshold, obtain the scattering distribution corresponding to the CBCT projection according to the pre-loaded detector response curve, otherwise, return to the step of uniformly sampling the sampling interval within the preset range to obtain a random number until the photon residual energy is less than the preset energy threshold or the sampling result is not the Compton effect; wherein the differential scattering cross section is physical data associated with the CBCT projection; When the sampling result is not the Compton effect, determine whether the sampling result is the Rayleigh scattering; when the sampling result is the Rayleigh scattering, sample the scattering direction based on the differential scattering cross section, and then return to execute the step of uniformly sampling the sampling interval within the preset range to obtain a random number; when the sampling result is not the Rayleigh scattering, determine that the sampling result is the photoelectric effect, absorb and deposit photon energy based on the differential scattering cross section, and then obtain the scattering distribution corresponding to the CBCT projection according to the pre-loaded detector response curve.

3. The method according to claim 1, It is characterized in that Prior to the step of inputting the CBCT projection as a model, the method further comprises the following steps: performing data enhancement processing on the CBCT projection; The result of the data enhancement processing is used as the model input, and the data enhancement processing includes rotation processing and flipping processing.

4. The method according to claim 1, It is characterized in that The method further comprises the following steps: A coding path is constructed based on a first preset number of coding modules; wherein a downsampling operation and a maximum pooling operation are provided at the output of each coding module in the coding path; A decoding path is constructed based on a second preset number of decoding modules; wherein an upsampling operation is provided at the input of each decoding module in the decoding path; The encoding module and the decoding module are both constructed based on a residual block, and the residual block includes a convolutional layer, a batch normalization process, and a ReLU activation operation.

5. The method according to claim 1 or 4, It is characterized in that The CBCT projection is used as a model input, the scattering distribution is used as a training label, and the pre-constructed residual convolutional neural network is optimized and trained. After the training is completed, a target network is obtained, which includes the following steps: Arrange the training set and the test set according to the CBCT projection; use the residual convolutional neural network as the first network; Inputting the training data in the training set into the first network for processing to obtain a first scattering distribution; Performing loss calculation based on the first scattering distribution and the training label of the scattering distribution corresponding to the training data, and adjusting the network parameters of the first network according to the result of the loss calculation to obtain a second network; The second network is evaluated for error using the test set. When the result of the error evaluation is less than a preset error threshold, the second network is used as the target network after training. Otherwise, the second network is used as the first network. Then, the step of inputting the training data in the training set into the first network to obtain the first scattering distribution is returned to the execution until the target network is obtained.

6. The method according to claim 1, It is characterized in that The method of using the target network to perform iterative scatter correction on the processed CBCT projection to obtain a corrected CBCT image comprises the following steps: The target network is used to process the CBCT projection to be processed to obtain a second scattering distribution, and the CBCT projection to be processed is iteratively corrected based on the second scattering distribution until a preset condition is met to obtain the corrected CBCT image.

7. The method according to claim 6, It is characterized in that The method of obtaining a second scattering distribution by processing the CBCT projection to be processed using the target network, and iteratively correcting the CBCT projection to be processed based on the second scattering distribution until a preset condition is met to obtain the corrected CBCT image comprises the following steps: Using the CBCT projection to be processed as an input projection; Inputting the input projection into the target network for processing to obtain the second scattering distribution; Correcting the scattering signal deviation of the input projection based on the second scattering distribution to obtain a corrected projection; The corrected projection is used as the input projection, and then the step of inputting the input projection into the target network for processing to obtain the second scattering distribution is returned to be executed until the difference between the corrected projections of two adjacent iterative processes is less than a preset difference threshold; image reconstruction is performed based on the corrected projection finally obtained to obtain the corrected CBCT image.

8. A CT scatter correction device, It is characterized in that The device comprises: The first module is used to obtain CBCT projections of cases with multiple parts; A second module is used to perform a Monte Carlo simulation on the CBCT projection based on the interaction of photon energy to obtain a scattering distribution corresponding to the CBCT projection; The third module is used to use the CBCT projection as a model input and the scattering distribution as a training label to optimize the pre-constructed residual convolutional neural network and obtain a target network after the training is completed; The fourth module is used to perform iterative scatter correction on the CBCT projection to be processed using the target network to obtain a corrected CBCT image.

9. An electronic device, It is characterized in that The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program. It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images