General imaging physically-driven CT data simulation and CT image noise artifact suppression method

Abstract

The invention discloses a universal imaging physically-driven CT data simulation and CT image noise artifact suppression method. According to the method, pixel values of bones, fat, muscles and air in a real CT image and a real CBCT image are calculated respectively, a mapping function is constructed, pixel values of tissues in the CT image are mapped into pixel values of corresponding parts in the real CBCT image, and simulated CBCT data are obtained. The method comprises the following steps: firstly, simulating CBCT data, then obtaining the projection of the simulated CBCT data by using an orthographic projection function, simulating the projection data of low-dose CBCT through a mode of sparse sampling angle and addition of projection domain noise, and then reconstructing the projection domain data through an FDK function, so as to obtain paired simulated CBCT data capable of supporting supervised learning in such a universal imaging physical driving mode. Test results show that the method has good results in the aspects of reducing artifacts, suppressing noise and recovering image details in the aspects of simulating data and real data.

Description

Technical Field

[0001] The invention relates to a universal imaging physics-driven CT data simulation and CT image noise artifact suppression method, and belongs to the field of computer image processing. Background Art

[0002] CT (Computed Tomography) is an X-ray imaging technology. As a core technology of medical imaging, it is widely used in the diagnosis and evaluation of various diseases and conditions. As an important development direction of CT technology, cone beam CT has the advantages of direct reconstruction of three-dimensional images, low X-ray radiation dose, and fast imaging speed, and has been vigorously promoted. However, due to the cone beam geometry causing artifacts and blurring at the edge of the image and the lower radiation dose causing noise, the image quality of CBCT is lower than that of CT images, so the improvement of CBCT image quality has become an important research direction.

[0003] Deep learning is an important technical means to improve the quality of CBCT images. As a data-driven optimization algorithm, deep learning requires a large amount of data to train the algorithm model and improve the effectiveness and robustness of the model. The reality is that due to the privacy of the data, there are very few public data sets for improving the quality of low-dose CBCT images, so it is necessary to self-made simulated CBCT data to test the effectiveness of the algorithm, and use a small amount of real data to improve the robustness of the algorithm.

[0004] Whether the simulated CBCT data is close to the real data is the core of whether the algorithm can be truly applied to the real data, and it is also essential to further improve the robustness to the real data through the guidance of real data. The former determines whether the deep learning model is capable of improving the image quality of real CBCT data, and the latter focuses on further improving the quality of real CBCT images based on the former. Therefore, it is necessary to explore a CT data simulation method based on imaging physics drive and a CT image noise artifact suppression method based on unsupervised domain adaptation. Summary of the invention

[0005] In order to improve the quality of CBCT images, the present invention proposes a universal imaging physics-driven CT data simulation and CT image noise artifact suppression method. The entire process of the method is as follows: Figure 1 As shown in the figure, it can generate simulated CBCT data through real CT data according to the pixel values ​​of each tissue in the real CT image and the pixel values ​​of the corresponding parts of the real CBCT image, and then remove the noise and artifacts on the real data through the CT image noise artifact suppression method based on unsupervised domain adaptation.

[0006] In order to achieve the above-mentioned purpose, the technical solution of the present invention is as follows: The present invention provides a general imaging physics-driven CT data simulation and CT image noise artifact suppression method, and the specific steps are as follows:

[0007] Step 1: The real CT data is converted from CT value to pixel value.

[0008] Step 2: reconstruct the real CBCT projection data through the FDK function, whose parameters include the distance from the ray source to the detector, the distance from the ray source to the human body, the number of detectors, the size of the detector probe, the offset of the detector on the two-dimensional plane, the size of the generated image pixel, and the angle information corresponding to the projection, to obtain the real CBCT image under the pixel value.

[0009] Step 3, such as Figure 1 As shown, according to the real CT image under the pixel value obtained in step 1, check the average pixel value of the corresponding area of ​​bone, fat, muscle and air; according to the real CBCT image obtained in step 2, check the average pixel value of the corresponding area of ​​bone, fat, muscle and air, and calculate the function of mapping the pixel value of each part of the CT image to the pixel value of each part of the CBCT image. The pixel value of the real CT image is mapped to the pixel value corresponding to the real CBCT image to obtain the CT image after pixel value alignment, and the projection data is obtained by the forward projection function. The parameters of the forward projection function include the distance from the ray source to the detector, the distance from the ray source to the human body, the number of detectors, the size of the detector probe, the offset of the detector on the two-dimensional plane, the size of the generated image pixel point and the angle information corresponding to the projection. The above data are all consistent with the parameters in step 2 to ensure that the obtained projection is closer to the real CBCT projection. Then reconstruct it through the FDK function, and its parameters are consistent with the above parameters to obtain a simulated CBCT image.

[0010] Step 4, according to steps 1-3, obtain the standard dose CBCT simulation data and the corresponding low dose CBCT simulation data. According to step 2, obtain the standard dose CBCT real data and the corresponding low dose CBCT real data. It should be noted that the first two steps in the acquisition process of low dose CBCT simulation data are consistent with the standard dose CBCT simulation data. In step 3, the parameters of the forward projection and FDK function need to be modified. Each parameter is synchronously sampled by one k to simulate the image artifacts under sparse angles, and Poisson noise is added to the CBCT simulation projection after the forward projection to simulate the real noise caused by the lower radiation dose. The low dose CBCT real data is similar. One k is sampled on the projection and the same Poisson noise as the simulated low dose is added. On this basis, offline data enhancement is added, and the projection data after moving in three dimensions is randomly generated and reconstructed to obtain the paired low dose CBCT simulation data and standard dose CBCT simulation data after offline data enhancement.

[0011] Step 5: Based on the paired CBCT simulation data obtained in step 4, the low-dose CBCT simulation image is input into the supervised noise artifact removal backbone network, and the CBCT simulation image after the artifacts and noise are removed is output. Figure 3 As shown in the figure, the structure of the supervised noise artifact removal backbone network includes a downsampling path and an upsampling path, and the jump connection in the middle is used to transmit the detailed information of the image. At the same time, we set the number of channels of feature maps of different resolutions from small to large to 32-64-128-256 to occupy the GPU video memory and speed up the operation. The essence of the denoising network is to learn the difference between the input and the label. In this task, we introduced a residual module to transform the network from learning the data itself to learning the noise. Figure 3 The training process of the network is shown: We first use online data augmentation to expand the data, including random translation, rotation, etc., to prevent the network from overfitting and enhance the generalization ability of the network. Then the image is normalized and put into the supervised noise artifact removal backbone network to obtain the predicted output with improved image quality, and the supervised image quality improvement loss function is calculated together with the label. Then the gradient is calculated through the loss value, and the network parameters are updated using the stochastic gradient descent optimization method.

[0012] Step 6: Based on the low-dose CBCT real data and low-dose CBCT simulation data obtained in step 4, the two data are input into the backbone network respectively, the feature distribution learned by the encoder part is extracted, and it is upsampled to the original image size to speed up the operation by occupying the GPU video memory. After obtaining the corresponding feature map, it is input into the domain discriminator. The domain discriminator can extract the deep information of the feature map, perceive the distribution of the input features, and output the distribution of the input feature map. The distribution loss of the domain discriminator is calculated through the unsupervised domain adaptation loss function. Figure 3 The training process of the network is shown.

[0013]

[0014] Among them, I Source represents the simulated CBCT data, I Target represents the real CBCT data, Represents the feature extraction output and upsampled to the original image size, Φ Discriminator (·) represents the distribution of the domain discriminator output, It represents the discriminator Φ Discriminator right The gradient of , where:

[0015]

[0016] The unsupervised domain adaptation loss is added to the loss of the supervised noise artifact removal backbone network in step 5. The learning of real data features in unsupervised domain adaptation learning is used to guide the supervised noise artifact removal backbone network to adapt to the features of real data, thereby improving the effectiveness and robustness of the denoising network.

[0017] Compared with the prior art, the advantages of the present invention are as follows: (1) By collecting CT data of real human chest and applying offline data enhancement technology, the present method can generate data that is very close to the real CBCT data set in terms of tissue pixel value range, detail contrast and noise artifacts, thereby achieving accurate simulation effects. This means that in the field of medical imaging, we can obtain more realistic and reliable simulation data, which is helpful for the research and diagnosis of medical images; (2) The scheme adopts residual learning and multi-scale feature learning methods to effectively remove noise artifacts in CT images while keeping the details very close to the standard dose. This means that even in a noisy environment, we can maintain high image quality, which helps doctors to accurately diagnose patients; (3) Through the unsupervised domain transfer method, the transfer learning ability of the model from simulation data to real data is improved, thereby enhancing the generalization performance of the model. This means that our model can better adapt to real data, and even when there is a certain difference between real data and simulation data, the model can also show good stability; (4) The scheme runs on a GPU, so in actual testing, it only takes 190 milliseconds to divide the positioning line for a single positioning image, which has a fast computing speed. This means that our method not only has improvements in quality, but also has obvious advantages in computational efficiency, and can process medical imaging data more quickly, providing more efficient support for clinical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 , Figure 2 It is a flow chart of the present invention.

[0019] Figure 3 This is the network training graph.

[0020] Figure 4 This is the test result diagram. DETAILED DESCRIPTION

[0021] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0022] Example: Figure 1 , Figure 2As shown, the present invention provides a general imaging physics-driven CT data simulation and CT image noise artifact suppression method. The final test results are as follows Figure 4 As shown, the specific steps are as follows:

[0023] Step 1: The real CT data is converted from CT value to pixel value.

[0024] Step 2: reconstruct the real CBCT projection data through the FDK function, whose parameters include the distance from the ray source to the detector, the distance from the ray source to the human body, the number of detectors, the size of the detector probe, the offset of the detector on the two-dimensional plane, the size of the generated image pixel, and the angle information corresponding to the projection, to obtain the real CBCT image under the pixel value.

[0025] Step 3, such as Figure 1 As shown, according to the real CT image under the pixel value obtained in step 1, check the average pixel value of the corresponding area of ​​bone, fat, muscle and air; according to the real CBCT image obtained in step 2, check the average pixel value of the corresponding area of ​​bone, fat, muscle and air, and calculate the function of mapping the pixel value of each part of the CT image to the pixel value of each part of the CBCT image. The pixel value of the real CT image is mapped to the pixel value corresponding to the real CBCT image to obtain the CT image after pixel value alignment, and the projection data is obtained by the forward projection function. The parameters of the forward projection function include the distance from the ray source to the detector, the distance from the ray source to the human body, the number of detectors, the size of the detector probe, the offset of the detector on the two-dimensional plane, the size of the generated image pixel point and the angle information corresponding to the projection. The above data are all consistent with the parameters in step 2 to ensure that the obtained projection is closer to the real CBCT projection. Then reconstruct it through the FDK function, and its parameters are consistent with the above parameters to obtain a simulated CBCT image.

[0026] Step 4, according to steps 1-3, obtain the standard dose CBCT simulation data and the corresponding low dose CBCT simulation data. According to step 2, obtain the standard dose CBCT real data and the corresponding low dose CBCT real data. It should be noted that the first two steps in the acquisition process of low dose CBCT simulation data are consistent with the standard dose CBCT simulation data. In step 3, the parameters of the forward projection and FDK function need to be modified. Each parameter is synchronously sampled by one k to simulate the image artifacts under sparse angles, and Poisson noise is added to the CBCT simulation projection after the forward projection to simulate the real noise caused by the lower radiation dose. The low dose CBCT real data is similar. One k is sampled on the projection and the same Poisson noise as the simulated low dose is added. On this basis, offline data enhancement is added, and the projection data after moving in three dimensions is randomly generated and reconstructed to obtain the paired low dose CBCT simulation data and standard dose CBCT simulation data after offline data enhancement.

[0027] Step 5: Based on the paired CBCT simulation data obtained in step 4, the low-dose CBCT simulation image is input into the supervised noise artifact removal backbone network, and the CBCT simulation image after the artifacts and noise are removed is output. Figure 3 As shown in the figure, the structure of the supervised noise artifact removal backbone network includes a downsampling path and an upsampling path, and the jump connection in the middle is used to transmit the detailed information of the image. At the same time, we set the number of channels of feature maps of different resolutions from small to large to 32-64-128-256 to occupy the GPU video memory and speed up the operation. The essence of the denoising network is to learn the difference between the input and the label. In this task, we introduced a residual module to transform the network from learning the data itself to learning the noise. Figure 3 The training process of the network is shown: We first use online data augmentation to expand the data, including random translation, rotation, etc., to prevent the network from overfitting and enhance the generalization ability of the network. Then the image is normalized and put into the supervised noise artifact removal backbone network to obtain the predicted output with improved image quality, and the supervised image quality improvement loss function is calculated together with the label. Then the gradient is calculated through the loss value, and the network parameters are updated using the stochastic gradient descent optimization method.

[0028] Step 6: Based on the low-dose CBCT real data and low-dose CBCT simulation data obtained in step 4, the two data are input into the backbone network respectively, the feature distribution learned by the encoder part is extracted, and it is upsampled to the original image size to speed up the operation by occupying the GPU video memory. After obtaining the corresponding feature map, it is input into the domain discriminator. The domain discriminator can extract the deep information of the feature map, perceive the distribution of the input features, and output the distribution of the input feature map. The distribution loss of the domain discriminator is calculated through the unsupervised domain adaptation loss function. Figure 3 The training process of the network is shown.

[0029]

[0030] Among them, I Source represents the simulated CBCT data, I Target represents the real CBCT data, Represents the feature extraction output and upsampled to the original image size, Φ Discriminator (·) represents the distribution result of the domain discriminator output.

[0031] The unsupervised domain adaptation loss is added to the loss of the supervised noise artifact removal backbone network in step 5. The learning of real data features in unsupervised domain adaptation learning is used to guide the supervised noise artifact removal backbone network to adapt to the features of real data, thereby improving the effectiveness and robustness of the denoising network.

[0032] Effect evaluation:

[0033] The present invention discloses a universal imaging physics-driven CT data simulation and CT image noise artifact suppression method. The test results of this method are as follows Figure 4 As shown. The figure shows 3 groups of test images, including real CBCT images of different positions of the human chest. The test was conducted in the CMD command line in the Windows system. The test process is: first enter condaactivate environment to activate the environment configuration for running the code, then enter python data_simulated.py for data simulation, and finally enter python test.py for data artifact noise removal test. The average running time of the algorithm is about 190ms, which is fast. The simulation is very close to the real data, and the noise artifact removal is relatively accurate. This shows that this method can simulate data and remove noise artifacts quickly and with high quality.

[0034] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A general imaging physics-driven CT data simulation and CT image noise artifact suppression method, characterized in that: The specific steps are as follows: Step 1: Real CT image I CT Convert CT value to pixel value. Step 2: The real CBCT projection data p CBCT The real CBCT image is reconstructed by the FDK function. Step 3: The real CBCT image I under the pixel value obtained in step 2 CBCT , check the pixel values ​​corresponding to its bones, fat, muscles and air, and the real CT image I obtained in step 1 CT , check the pixel values ​​corresponding to the bone, fat, muscle and air, and calculate the function f that maps the CT image to the CBCT image CT_to_CBCT , according to the obtained mapping function f CT_to_CBCT , the real CT data I CT The pixel values ​​are mapped to the real CBCT data I CBCT The corresponding pixel value is obtained by using the mapped CT data I CT_simulated , the projection data p is obtained through the forward projection function cbct_simulated , and then reconstructed by FDK function to obtain the simulated CBCT data I CBCT_simulated , Step 4: According to steps 1-3, obtain the standard dose CBCT simulation data I CBCT_simulated_nd And the corresponding low-dose CBCT simulation data I CBCT_simulated_ld , Step 5: According to the paired CBCT simulation data obtained in step 3, input low-dose CBCT simulation data I CBCT_simulated_ld , through the denoising network, the standard dose CBCT simulation data I is calculated CBCT_simulated_nd The L1 loss between them is used to reduce the loss through network learning to remove noise and artifacts on low-dose data. Step 6: Introducing low-dose CBCT real data I CBCT_ld , and the input of the backbone network (low-dose CBCT simulation data I CBCT_simulated_ld ) to perform unsupervised domain adaptation learning, use the domain discriminator, use the WGAN_GP loss function to calculate the distribution loss of the two different domain features, add it to the loss of the supervised noise artifact removal backbone network, guide the network to learn the characteristics and data distribution of the real data domain, use the final training weight network, input low-dose CBCT real data I CBCT_ld , output I after removing artifact noise CBCT_output .

2. The method of CT data simulation and CT image noise artifact suppression driven by general imaging physics according to claim 1, characterized in that: The real CT data in step 1 comes from the public CT dataset. The CT data is converted from hu value to mu value.

3. The method of CT data simulation and CT image noise artifact suppression driven by general imaging physics according to claim 1, characterized in that: In step 2, the real CBCT is reconstructed according to the given reconstruction parameters through the FDK function of the third-party library TIGRE.

4. The method of CT data simulation and CT image noise artifact suppression driven by general imaging physics according to claim 1, characterized in that: In step 3, the pixel values ​​are checked by opening the image using ImageJ software, zooming in on the bone, fat, muscle, and air areas, selecting and calculating their corresponding average values, and obtaining the corresponding pixel values ​​of each tissue.

5. The method of general imaging physics-driven CT data simulation and CT image noise artifact suppression according to claim 1, characterized in that: The parameters used in the forward projection function mentioned in step 3 are consistent with the reconstruction parameters of the real CBCT data in step 2.

6. The method of CT data simulation and CT image noise artifact suppression driven by general imaging physics according to claim 1, characterized in that: The parameters used by the FDK function in step 3 are consistent with the reconstruction parameters of the real CBCT data in step 2.

7. The method of claim 1, wherein: Step 4 Low-dose CBCT simulation data I CBCT_simulated_ld Before the FDK function is used for reconstruction, one-kth of the projections are selected and a certain amount of Poisson noise is added to the projection data. Finally, the FDK function is used to reconstruct the simulated low-dose CBCT data.

8. The method of claim 1, wherein: The specific implementation form of unsupervised domain adaptation in step 6 is: D target =Φ Discriminator (feature target ) D source =Φ Discriminator (feature source ) Among them, I source , I target Represents low-dose CBCT simulation data and low-dose CBCT real data, It means extracting the corresponding features and upsampling them to the original image size, Φ Discriminator It represents the domain discriminator, which is implemented by multi-layer convolution and is used to output the distribution of input features, D target , D source It represents the real feature distribution and simulated feature distribution output by the domain discriminator respectively. Finally, the unsupervised domain adaptation loss function is used to calculate the loss of unsupervised domain adaptation. in, and represents the expected distribution of source and target domain features, λ gp represents the weight of the gradient guidance part, Representation discriminator Φ Discriminator right The gradient of , where:

Images