Double-source CT scattering correction method and system based on deep learning model

Abstract

The invention relates to an artificial intelligence technology, and discloses a double-source CT scatter correction method and system based on a deep learning model, and the method comprises the steps: obtaining the head mode double-source CT projection data of a preset head mode, obtaining a comprehensive cross scattering signal of the head mode double-source CT projection data, and obtaining the comprehensive cross scattering signal of the head mode double-source CT projection data; obtaining a forward scattering signal of the head mode double-source CT projection data by using a preset adaptive scattering kernel superposition algorithm, training a neural network model by using the head mode double-source CT projection data, the comprehensive cross scattering signal and the forward scattering signal, and obtaining a double-source CT scattering analysis model, and performing scattering signal estimation on pre-acquired target double-source CT projection data by using the double-source CT scattering analysis model to obtain a scattering signal, and subtracting the scattering signal from the target double-source CT projection data to obtain target double-source CT scattering correction data. The invention further provides a double-source CT scatter correction system based on the deep learning model. According to the invention, the image quality of double-source CT imaging can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a dual-source CT scatter correction method and system based on a deep learning model. Background Art

[0002] CT scanning technology is an important imaging method in medical imaging, which can provide detailed images of the internal structure of the patient's body. As an advanced form of CT technology, the dual-source CT system can significantly improve the scanning speed and image quality through two independent X-ray sources and detectors, especially showing unique advantages in cardiac imaging and large-scale scanning.

[0003] However, the existing dual-source CT system cannot avoid the phenomenon of X-ray scattering in practical applications. The scattered signal significantly affects the image quality and increases the complexity and difficulty of image reconstruction. Summary of the invention

[0004] The present invention provides a dual-source CT scatter correction method and system based on a deep learning model, the main purpose of which is to solve the problem of X-ray scattering in actual imaging of the existing dual-source CT system.

[0005] To achieve the above object, the present invention provides a dual-source CT scatter correction method based on a deep learning model, comprising:

[0006] Acquire dual-source CT projection data of a preset head model;

[0007] Acquire a comprehensive cross-scatter signal of the head model dual-source CT projection data;

[0008] Acquire the forward scattering signal of the head model dual-source CT projection data by using a preset adaptive scattering kernel superposition algorithm;

[0009] Training a neural network model using the head model dual-source CT projection data, the integrated cross-scatter signal, and the forward-scatter signal to obtain a dual-source CT scatter analysis model;

[0010] Using the dual-source CT scattering analysis model to estimate scattering signals of pre-acquired target dual-source CT projection data, to obtain scattering signals;

[0011] The scatter signal is subtracted from the target dual-source CT projection data to obtain target dual-source CT scatter correction data.

[0012] Optionally, the step of acquiring a comprehensive cross-scatter signal of the head model dual-source CT projection data includes:

[0013] Turning on a first ray tube and a second detector of a preset dual-source CT system;

[0014] Acquiring a first cross-scatter signal of the first ray tube using the second detector;

[0015] Turn off the first ray tube and the second detector, and turn on the second ray tube and the first detector of the preset dual-source CT system;

[0016] Acquiring a second cross-scatter signal of the second ray tube using the first detector;

[0017] The first cross heat dissipation signal and the second cross scattering signal are aggregated to obtain a comprehensive cross scattering signal.

[0018] Optionally, the using a preset adaptive scattering kernel superposition algorithm to obtain a forward scattering signal of the head model dual-source CT projection data includes:

[0019] Through Monte Carlo simulation, the scattered signals of the ray beams of different energy levels emitted by the high-voltage bulb under different currents and voltages passing through the preset water models of different lengths are simulated to obtain the scattered signal set;

[0020] Fitting the distribution of each scattering signal in the scattering signal set to obtain the scattering signal distribution of the ray beam passing through water models of different lengths at different energy levels;

[0021] Traversing each pixel of the head model dual-source CT projection data according to the scattered signal distribution, to obtain the scattered distribution of each ray in the head model dual-source CT projection data;

[0022] The scattering distribution of each ray in the head model dual-source CT projection data is superimposed to obtain a forward scattering signal of the head model dual-source CT projection data.

[0023] Optionally, the using the head model dual-source CT projection data, the integrated cross-scatter signal and the forward-scatter signal to train a neural network model to obtain a dual-source CT scatter analysis model includes:

[0024] Acquire projection image signals in the head model dual-source CT projection data, and perform standardization processing on the projection image signals to obtain standardized projection image signals;

[0025] combining the integrated cross scatter signal with the forward scatter signal to obtain a standard scatter signal;

[0026] Using a neural network model to perform scattering signal analysis based on the standardized projection image signal to obtain a scattering signal estimate;

[0027] Calculating a loss value between the estimated scattered signal and the standard scattered signal;

[0028] Optimizing the loss value by updating the weight parameters of the neural network model using a stochastic gradient descent algorithm;

[0029] Using a preset step learning rate algorithm to update the learning rate of the neural network model to obtain an updated learning rate;

[0030] When the preset number of iterations is reached, the weight parameter corresponding to the minimum loss value among all iterations is selected as the final weight parameter;

[0031] The updated learning rate and the final weight parameter are input into a neural network model to obtain a dual-source CT scattering analysis model.

[0032] Optionally, the step of performing standardization processing on the projection image signal to obtain a standardized projection image signal includes:

[0033] The projection image signal is standardized using the following formula to obtain a standardized projection image signal:

[0034]

[0035] Wherein, p is the standardized projection image signal, I is the projection image signal, and I0 is the bright field projection image signal contained in the head model dual-source CT projection data.

[0036] Optionally, the calculating a loss value between the estimated scattered signal and the standard scattered signal includes:

[0037] The loss value is calculated using the following formula:

[0038]

[0039] Wherein, S is the loss value, N is the number of pixels of the first detector and the second detector, and I s is the standard scattering signal, I m is the scattered signal estimate, I p is the standardized projection image signal.

[0040] Optionally, the updating the learning rate of the neural network model by using a preset step learning rate algorithm to obtain an updated learning rate includes:

[0041] The learning rate is updated using the following formula:

[0042]

[0043] Among them, lr new is the updated learning rate, lr old is the learning rate, K is the current number of iterations, and M is the preset number of iterations.

[0044] Optionally, the using the dual-source CT scatter analysis model to estimate scatter signals of pre-acquired target dual-source CT projection data to obtain corrected scatter signals includes:

[0045] Using a convolution kernel preset in a convolution layer of the dual-source CT scattering analysis model to move on the target dual-source CT projection data according to a preset step length, calculating the dot product of the convolution kernel and the data of the area covered by the convolution kernel after each movement, summing all dot products, and obtaining a convolution feature;

[0046] A rectangular window preset in the pooling layer of the dual-source CT scattering analysis model is used to continuously move in the convolution feature, an average value of data in the area covered by the rectangular window after each movement is calculated, and the average value calculated in each area replaces the data in the area to obtain a pooling feature;

[0047] Using the activation function preset in the activation layer of the dual-source CT scattering analysis model to introduce a nonlinear factor into the pooling feature, so as to obtain an activation feature;

[0048] The fully connected layer of the dual-source CT scatter analysis model is used to obtain the output of each neuron according to the activation feature, and the output of each neuron is summed to obtain a corrected scatter signal.

[0049] Optionally, the step of subtracting the scatter signal from the target dual-source CT projection data to obtain target dual-source CT scatter correction data includes:

[0050] The target dual-source CT projection data is scatter corrected using the following formula:

[0051]

[0052] Among them, I corr is the target dual-source CT scatter correction data, I total is the target dual-source CT projection data, I scatter,model is the scattered signal.

[0053] In order to solve the above problems, the present invention also provides a dual-source CT scatter correction system based on a deep learning model, the system comprising:

[0054] A data acquisition module, used to acquire dual-source CT projection data of a preset head model, and to acquire a comprehensive cross-scatter signal of the dual-source CT projection data of the head model;

[0055] A scattering estimation module, used for acquiring a forward scattering signal of the head model dual-source CT projection data by using a preset adaptive scattering kernel superposition algorithm;

[0056] A model training module, used to train a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal and the forward scattering signal to obtain a dual-source CT scattering analysis model;

[0057] The scatter correction module is used to use the dual-source CT scatter analysis model to estimate the scatter signal of the pre-acquired target dual-source CT projection data to obtain the scatter signal, and subtract the scatter signal from the target dual-source CT projection data to obtain the target dual-source CT scatter correction data.

[0058] The embodiment of the present invention obtains the head model dual-source CT projection data of a preset head model, obtains the comprehensive cross-scattering signal of the head model dual-source CT projection data, and uses the preset adaptive scatter kernel superposition algorithm to obtain the forward scattering signal of the head model dual-source CT projection data, thereby improving the efficiency of training the dual-source CT scattering analysis model, using the head model dual-source CT projection data, the comprehensive cross-scattering signal and the forward scattering signal to train the neural network model to obtain the dual-source CT scattering analysis model, using the dual-source CT scattering analysis model to estimate the scattering signal of the pre-acquired target dual-source CT projection data to obtain the scattering signal, and using the target dual-source CT projection data to subtract the scattering signal to obtain the target dual-source CT scattering correction data, thereby improving the efficiency and accuracy of scattering correction of the dual-source CT projection data. Therefore, the dual-source CT scattering correction method and system based on the deep learning model proposed in the present invention can solve the problem of X-ray scattering in actual imaging of the existing dual-source CT system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A schematic diagram of a flow chart of a dual-source CT scatter correction method based on a deep learning model provided in one embodiment of the present invention;

[0060] Figure 2 A schematic diagram of a process for obtaining a comprehensive cross-scattering signal provided by an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of a process for obtaining a forward scattered signal according to an embodiment of the present invention;

[0062] Figure 4 A functional module diagram of a dual-source CT scatter correction system based on a deep learning model provided in one embodiment of the present invention.

[0063] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0064] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0065] The embodiment of the present application provides a dual-source CT scatter correction method based on a deep learning model. The execution subject of the dual-source CT scatter correction method based on a deep learning model includes but is not limited to at least one of the electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided in the embodiment of the present application. In other words, the dual-source CT scatter correction method based on a deep learning model can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be 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, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0066] Reference Figure 1 FIG. 1 is a flow chart of a dual-source CT scatter correction method based on a deep learning model provided in an embodiment of the present invention. In this embodiment, the dual-source CT scatter correction method based on a deep learning model includes:

[0067] S1. Acquire dual-source CT projection data of a preset head model.

[0068] In the embodiment of the present invention, the dual-source CT projection data is projection data obtained by using a dual-source CT system. The dual-source CT system uses two types of X-rays with different energies or energy spectra during the CT scanning process to obtain high- and low-energy projection attenuation information of the object at various angles, and then obtains a distribution map of the atomic number and electron density of the imaged object using a dual-source CT image reconstruction algorithm based on the relationship between the linear attenuation coefficient of the material with respect to the X-rays and the ray energy, the atomic number and the density of the material.

[0069] In detail, the dual-source CT system includes a first ray tube, a second ray tube, a first detector and a second detector.

[0070] In detail, the first detector is used to receive the signal of the first ray tube, and the second detector is used to receive the signal of the second ray tube.

[0071] In the embodiment of the present invention, the head mold is a preset head model used to simulate the head of a patient in an actual usage scenario.

[0072] In the embodiment of the present invention, the head phantom dual-source CT projection data of the preset head phantom is obtained by using the first detector to receive the ray signal of the ray emitted by the first ray tube passing through the head phantom, and using the second detector to receive the ray signal of the ray emitted by the second ray tube passing through the head phantom, and using the dual-source CT image reconstruction algorithm to construct the head phantom dual-source CT projection data according to the two groups of ray signals.

[0073] In the embodiment of the present invention, by acquiring the head phantom dual-source CT projection data of the preset head phantom, the efficiency of subsequently establishing the dual-source CT scattering analysis model is improved.

[0074] S2. Acquire the comprehensive cross-scatter signal of the dual-source CT projection data of the head model.

[0075] In the embodiment of the present invention, the comprehensive cross-scatter signal refers to the fact that in a dual-source CT system, since two ray tubes emit X-rays simultaneously, these rays may cross when passing through the human body, thereby generating a comprehensive cross-scatter signal. The comprehensive cross-scatter signal will increase the noise of the image and reduce the image clarity.

[0076] In the embodiment of the present invention, refer to Figure 2 As shown, a schematic diagram of a process of obtaining a comprehensive cross scatter signal is provided in an embodiment of the present invention, wherein the process of obtaining the comprehensive cross scatter signal of the head model dual-source CT projection data includes:

[0077] S21, turning on a first ray tube and a second detector of a preset dual-source CT system;

[0078] S22, using the second detector to acquire a first cross-scattering signal of the first ray tube;

[0079] S23, turning off the first ray tube and the second detector, and turning on the second ray tube and the first detector of the preset dual-source CT system;

[0080] S24, using the first detector to acquire a second cross-scattering signal of the second ray tube;

[0081] S25. Summarize the first cross heat dissipation signal and the second cross scattering signal to obtain a comprehensive cross scattering signal.

[0082] In detail, since the second detector is located at the side of the first ray tube and the first detector is located at the side of the second ray tube, the real scattering signal can be obtained by turning on the first ray tube and the second detector alone or turning on the second ray tube and the first detector alone.

[0083] In the embodiment of the present invention, by acquiring the comprehensive cross-scattering signal of the head model dual-source CT projection data, the efficiency of subsequent training of the dual-source CT scattering analysis model is improved.

[0084] S3. Acquire the forward scattering signal of the head model dual-source CT projection data using a preset adaptive scattering kernel superposition algorithm.

[0085] In the embodiment of the present invention, the forward scatter signal refers to the forward scatter signal generated when the X-rays are deviated due to different substances when passing through the human body tissue.

[0086] In the embodiment of the present invention, the adaptive scattering kernel superposition algorithm is a technology used in image processing and computer vision, especially in image segmentation, feature extraction and pattern recognition. This algorithm enhances the local features of the image and improves the accuracy of image analysis by decomposing the image into different scattering components.

[0087] In the embodiment of the present invention, refer to Figure 3 As shown, a schematic diagram of a process of obtaining a forward scattering signal is provided in an embodiment of the present invention. The forward scattering signal of the head model dual-source CT projection data is obtained by using a preset adaptive scattering kernel superposition algorithm, including:

[0088] S31, through Monte Carlo simulation, simulating the scattered signals of the ray beams of different energy levels emitted by the high-voltage bulb under different currents and voltages passing through the preset water models of different lengths, to obtain a scattered signal set;

[0089] S32, fitting the distribution of each scattering signal in the scattering signal set to obtain the scattering signal distribution of the ray beam passing through water models of different lengths at different energy levels;

[0090] S33, traversing each pixel of the head model dual-source CT projection data according to the scattering signal distribution to obtain the scattering distribution of each ray in the head model dual-source CT projection data;

[0091] S34, superimposing the scattering distribution of each ray in the head model dual-source CT projection data to obtain a forward scattering signal of the head model dual-source CT projection data.

[0092] In detail, the Monte Carlo simulation is a method based on random sampling or statistical simulation, which estimates the numerical solution or probability distribution of a complex problem by repeatedly generating random samples and calculating statistics.

[0093] In detail, the Monte Carlo simulation is used to simulate the scattering signals of ray beams of different energy levels emitted by a high-voltage bulb under different currents and voltages passing through preset water models of different lengths, which refers to generating random current values ​​and voltage values ​​through Monte Carlo simulation, generating ray beams of different energy levels according to the randomly generated current values ​​and voltage values, and obtaining scattering signals of ray beams of different energy levels passing through water models of different lengths.

[0094] In the embodiment of the present invention, the forward scattering signal of the head model dual-source CT projection data is obtained by using a preset adaptive scattering kernel superposition algorithm, so as to improve the accuracy of the obtained forward scattering signal and the accuracy of the subsequently established dual-source CT scattering analysis model.

[0095] S4. Training a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal, and the forward-scattering signal to obtain a dual-source CT scattering analysis model.

[0096] In the embodiment of the present invention, the head model dual-source CT projection data is a data set including continuous frame projection image information.

[0097] In the embodiment of the present invention, the use of the head model dual-source CT projection data, the integrated cross-scatter signal and the forward scatter signal to train the neural network model refers to using the head model dual-source CT projection data as training data and the integrated cross-scatter signal and the forward scatter signal as labels to train the neural network model.

[0098] In an embodiment of the present invention, the neural network model includes but is not limited to a convolutional neural network model and a recurrent neural network model.

[0099] In an embodiment of the present invention, the step of training a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal, and the forward-scattering signal to obtain a dual-source CT scattering analysis model includes:

[0100] Acquire projection image signals in the head model dual-source CT projection data, and perform standardization processing on the projection image signals to obtain standardized projection image signals;

[0101] combining the integrated cross scatter signal with the forward scatter signal to obtain a standard scatter signal;

[0102] Using a neural network model to perform scattering signal analysis based on the standardized projection image signal to obtain a scattering signal estimate;

[0103] Calculating a loss value between the estimated scattered signal and the standard scattered signal;

[0104] Optimizing the loss value by updating the weight parameters of the neural network model using a stochastic gradient descent algorithm;

[0105] Using a preset step learning rate algorithm to update the learning rate of the neural network model to obtain an updated learning rate;

[0106] When the preset number of iterations is reached, the weight parameter corresponding to the minimum loss value among all iterations is selected as the final weight parameter;

[0107] The updated learning rate and the final weight parameter are input into a neural network model to obtain a dual-source CT scattering analysis model.

[0108] In detail, the step of performing standardization processing on the projection image signal to obtain a standardized projection image signal includes:

[0109] The projection image signal is standardized using the following formula to obtain a standardized projection image signal:

[0110]

[0111] Wherein, p is the standardized projection image signal, I is the projection image signal, and I0 is the bright field projection image signal contained in the head model dual-source CT projection data.

[0112] In detail, the bright field projection image signal refers to an X-ray image without attenuation in CT imaging, that is, an image obtained by assuming that all X-rays are not absorbed or scattered;

[0113] In detail, the calculating the loss value between the estimated scattered signal and the standard scattered signal includes:

[0114] The loss value is calculated using the following formula:

[0115]

[0116] Wherein, S is the loss value, N is the number of pixels of the first detector and the second detector, and I s is the standard scattering signal, I m is the scattered signal estimate, I p is the standardized projection image signal.

[0117] In detail, the method of updating the weight parameters of the neural network model by using the stochastic gradient descent algorithm is to update the preset weight parameters by using a preset weight decay coefficient. The preset weight decay coefficient may be 0.0001, and the preset weight parameter may be 0.9.

[0118] In detail, the method of updating the learning rate of the neural network model using a preset step learning rate algorithm to obtain an updated learning rate includes:

[0119] The learning rate is updated using the following formula:

[0120]

[0121] Among them, lr new is the updated learning rate, lr old is the learning rate, K is the current number of iterations, and M is the preset number of iterations.

[0122] In detail, the preset number of iterations may be 300.

[0123] In the embodiment of the present invention, the weight parameters are variables connecting neurons in the neural network model, and they determine the strength of the interaction between neurons.

[0124] In the embodiment of the present invention, the learning rate is an important hyperparameter in the training process of the neural network model, which controls the step size adjusted each time the model weights are updated.

[0125] In an embodiment of the present invention, before the training of the dual-source CT scattering analysis model reaches a preset number of iterations, if the loss value is less than a preset loss value threshold, it is directly confirmed that the training of the dual-source CT scattering analysis model is completed, thereby preventing the dual-source CT scattering analysis model from overfitting.

[0126] In an embodiment of the present invention, a dual-source CT scattering analysis model is obtained by training a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal, and the forward scattering signal, thereby improving the efficiency of training the dual-source CT scattering analysis model.

[0127] S5. Use the dual-source CT scatter analysis model to estimate scatter signals of pre-acquired target dual-source CT projection data to obtain corrected scatter signals.

[0128] In the embodiment of the present invention, the pre-acquired target dual-source CT projection data may be real dual-source CT data of a patient acquired by using a dual-source CT system in actual medical work.

[0129] In an embodiment of the present invention, the step of using the dual-source CT scatter analysis model to estimate scatter signals of pre-acquired target dual-source CT projection data to obtain corrected scatter signals includes:

[0130] Using a convolution kernel preset in a convolution layer of the dual-source CT scattering analysis model to move on the target dual-source CT projection data according to a preset step length, calculating the dot product of the convolution kernel and the data of the area covered by the convolution kernel after each movement, summing all dot products, and obtaining a convolution feature;

[0131] A rectangular window preset in the pooling layer of the dual-source CT scattering analysis model is used to continuously move in the convolution feature, an average value of data in the area covered by the rectangular window after each movement is calculated, and the average value calculated in each area replaces the data in the area to obtain a pooling feature;

[0132] Using the activation function preset in the activation layer of the dual-source CT scattering analysis model to introduce a nonlinear factor into the pooling feature, so as to obtain an activation feature;

[0133] The output of each neuron is obtained according to the activation feature using the fully connected layer of the dual-source CT scatter analysis model, and the output of each neuron is summed to obtain a corrected scatter signal.

[0134] In detail, the convolution kernel is a preset square matrix, the scale of which is much smaller than the target dual-source CT projection data.

[0135] In detail, the preset step size determines the amplitude of each movement of the convolution kernel. If the step size is too long, the efficiency of extracting convolution features will be higher, but it will cause information loss. If the step size is too short, the amount of calculation will increase.

[0136] In detail, the activation function is a function added to the nodes between the layers of the neural network model to introduce nonlinear factors, so that the neural network model can learn and handle complex problems. Common activation functions include Sigmoid function, Softmax function and ReLU function.

[0137] In the embodiment of the present invention, the dual-source CT scatter analysis model is used to estimate the scatter signal of the pre-acquired target dual-source CT projection data to obtain a corrected scatter signal, thereby improving the efficiency and accuracy of dual-source CT signal correction.

[0138] S6. Subtract the corrected scatter signal from the target dual-source CT projection data to obtain target dual-source CT scatter correction data.

[0139] In the embodiment of the present invention, the corrected scattering signal includes a comprehensive cross scattering signal and a forward scattering signal.

[0140] In an embodiment of the present invention, the step of subtracting the scatter signal from the target dual-source CT projection data to obtain target dual-source CT scatter correction data includes:

[0141] The target dual-source CT projection data is scatter corrected using the following formula:

[0142]

[0143] Among them, I corr is the target dual-source CT scatter correction data, I total is the target dual-source CT projection data, I scatter,model is the scattered signal.

[0144] In the embodiment of the present invention, the target dual-source CT scatter correction data is obtained by subtracting the corrected scatter signal from the target dual-source CT projection data, thereby improving the correction efficiency of the dual-source CT projection data.

[0145] like Figure 4 , is a functional module diagram of a dual-source CT scatter correction system based on a deep learning model provided in one embodiment of the present invention.

[0146] The dual-source CT scatter correction system 100 based on a deep learning model of the present invention can be installed in an electronic device. According to the functions implemented, the dual-source CT scatter correction system 100 based on a deep learning model can include a data acquisition module 101, a scatter estimation module 102, a model training module 103, and a scatter correction module 104. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0147] In this embodiment, the functions of each module / unit are as follows:

[0148] The data acquisition module 101 is used to acquire the head model dual-source CT projection data of a preset head model, and acquire the comprehensive cross-scattering signal of the head model dual-source CT projection data;

[0149] The scatter estimation module 102 is used to obtain the forward scatter signal of the head model dual-source CT projection data by using a preset adaptive scatter kernel superposition algorithm;

[0150] The model training module 103 is used to train a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal and the forward scattering signal to obtain a dual-source CT scattering analysis model;

[0151] The scatter correction module 104 is used to estimate the scatter signal of the pre-acquired target dual-source CT projection data using the dual-source CT scatter analysis model to obtain a scatter signal, and to subtract the scatter signal from the target dual-source CT projection data to obtain target dual-source CT scatter correction data.

[0152] In detail, each module described in the dual-source CT scatter correction system 100 based on the deep learning model in the embodiment of the present invention is used in the same manner as described above. Figures 1 to 3 The dual-source CT scatter correction method based on deep learning model described in the present invention has the same technical means and can produce the same technical effects, so it will not be repeated here.

[0153] In the embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0154] The modules described as separate components may or may not be physically separated, and the components shown as modules 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 modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0155] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, 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 hardware plus software functional modules.

[0156] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0157] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0158] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0159] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A dual-source CT scatter correction method based on a deep learning model, characterized in that: The method comprises: Acquire dual-source CT projection data of a preset head model; Acquire a comprehensive cross-scatter signal of the head model dual-source CT projection data; Acquire the forward scattering signal of the head model dual-source CT projection data by using a preset adaptive scattering kernel superposition algorithm; Training a neural network model using the head model dual-source CT projection data, the integrated cross-scatter signal, and the forward-scatter signal to obtain a dual-source CT scatter analysis model; Using the dual-source CT scattering analysis model to estimate scattering signals of pre-acquired target dual-source CT projection data, to obtain scattering signals; The scatter signal is subtracted from the target dual-source CT projection data to obtain target dual-source CT scatter correction data.

2. The dual-source CT scatter correction method based on a deep learning model according to claim 1, characterized in that: The step of acquiring a comprehensive cross-scatter signal of the head model dual-source CT projection data comprises: Turning on a first ray tube and a second detector of a preset dual-source CT system; Acquiring a first cross-scatter signal of the first ray tube using the second detector; Turn off the first ray tube and the second detector, and turn on the second ray tube and the first detector of the preset dual-source CT system; Acquiring a second cross-scatter signal of the second ray tube using the first detector; The first cross heat dissipation signal and the second cross scattering signal are aggregated to obtain a comprehensive cross scattering signal.

3. The dual-source CT scatter correction method based on a deep learning model according to claim 2, characterized in that: The method of obtaining the forward scattered signal of the head model dual-source CT projection data by using a preset adaptive scattering kernel superposition algorithm includes: Through Monte Carlo simulation, the scattered signals of the ray beams of different energy levels emitted by the high-voltage bulb under different currents and voltages passing through the preset water models of different lengths are simulated to obtain the scattered signal set; Fitting the distribution of each scattering signal in the scattering signal set to obtain the scattering signal distribution of the ray beam passing through water models of different lengths at different energy levels; Traversing each pixel of the head model dual-source CT projection data according to the scattered signal distribution, to obtain the scattered distribution of each ray in the head model dual-source CT projection data; The scattering distribution of each ray in the head model dual-source CT projection data is superimposed to obtain a forward scattering signal of the head model dual-source CT projection data.

4. The dual-source CT scatter correction method based on a deep learning model according to claim 3, characterized in that: The step of training a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal, and the forward-scattering signal to obtain a dual-source CT scattering analysis model includes: Acquire projection image signals in the head model dual-source CT projection data, and perform standardization processing on the projection image signals to obtain standardized projection image signals; combining the integrated cross scatter signal with the forward scatter signal to obtain a standard scatter signal; Using a neural network model to perform scattering signal analysis based on the standardized projection image signal to obtain a scattering signal estimate; Calculating a loss value between the estimated scattered signal and the standard scattered signal; Optimizing the loss value by updating the weight parameters of the neural network model using a stochastic gradient descent algorithm; Using a preset step learning rate algorithm to update the learning rate of the neural network model to obtain an updated learning rate; When the preset number of iterations is reached, the weight parameter corresponding to the minimum loss value among all iterations is selected as the final weight parameter; The updated learning rate and the final weight parameter are input into a neural network model to obtain a dual-source CT scattering analysis model.

5. The dual-source CT scatter correction method based on a deep learning model according to claim 4, characterized in that: The step of performing standardization processing on the projection image signal to obtain a standardized projection image signal includes: The projection image signal is standardized using the following formula to obtain a standardized projection image signal: Wherein, p is the standardized projection image signal, I is the projection image signal, and I0 is the bright field projection image signal contained in the head model dual-source CT projection data.

6. The dual-source CT scatter correction method based on a deep learning model according to claim 4, characterized in that: The calculating the loss value between the estimated scattered signal and the standard scattered signal comprises: The loss value is calculated using the following formula: Wherein, S is the loss value, N is the number of pixels of the first detector and the second detector, and I s is the standard scattering signal, I m is the scattered signal estimate, I p is the standardized projection image signal.

7. The dual-source CT scatter correction method based on a deep learning model according to claim 4, characterized in that: The method of updating the learning rate of the neural network model by using a preset step learning rate algorithm to obtain an updated learning rate includes: The learning rate is updated using the following formula: Among them, lr new is the updated learning rate, lr old is the learning rate, K is the current number of iterations, and M is the preset number of iterations.

8. The dual-source CT scatter correction method based on a deep learning model according to claim 4, characterized in that: The step of using the dual-source CT scatter analysis model to estimate scatter signals of pre-acquired target dual-source CT projection data to obtain corrected scatter signals includes: Using a convolution kernel preset in a convolution layer of the dual-source CT scattering analysis model to move on the target dual-source CT projection data according to a preset step length, calculating the dot product of the convolution kernel and the data of the area covered by the convolution kernel after each movement, summing all dot products, and obtaining a convolution feature; A rectangular window preset in the pooling layer of the dual-source CT scattering analysis model is used to continuously move in the convolution feature, an average value of data in the area covered by the rectangular window after each movement is calculated, and the average value calculated in each area replaces the data in the area to obtain a pooling feature; Using the activation function preset in the activation layer of the dual-source CT scattering analysis model to introduce a nonlinear factor into the pooling feature, so as to obtain an activation feature; The output of each neuron is obtained according to the activation feature using the fully connected layer of the dual-source CT scatter analysis model, and the output of each neuron is summed to obtain a corrected scatter signal.

9. The dual-source CT scatter correction method based on a deep learning model according to claim 1, characterized in that: The step of subtracting the scatter signal from the target dual-source CT projection data to obtain target dual-source CT scatter correction data includes: The target dual-source CT projection data is scatter corrected using the following formula: Among them, I corr is the target dual-source CT scatter correction data, I total is the target dual-source CT projection data, I scatter,model is the scattered signal.

10. A dual-source CT scatter correction system based on a deep learning model, characterized in that: The system comprises: A data acquisition module, used to acquire dual-source CT projection data of a preset head model, and to acquire a comprehensive cross-scatter signal of the dual-source CT projection data of the head model; A scattering estimation module, used for acquiring a forward scattering signal of the head model dual-source CT projection data by using a preset adaptive scattering kernel superposition algorithm; A model training module, used to train a neural network model using the head model dual-source CT projection data, the integrated cross-scattering signal and the forward scattering signal to obtain a dual-source CT scattering analysis model; The scatter correction module is used to use the dual-source CT scatter analysis model to estimate the scatter signal of the pre-acquired target dual-source CT projection data to obtain the scatter signal, and subtract the scatter signal from the target dual-source CT projection data to obtain the target dual-source CT scatter correction data.

Images