CN104574416A - Low-dose energy spectrum CT image denoising method - Google Patents

Abstract

A method for denoising a low-dose spectral CT image, comprising: (1) acquiring low-energy CT projection data and high-energy CT projection data of an imaging object under low-dose radiation, and respectively performing low-energy CT projection data and high-energy CT projection data Projection data for CT image reconstruction to obtain low-energy CT images and high-energy CT images , where H represents high energy, and L represents low energy; (2) According to the matrix decomposition model satisfied by the reconstructed data in step (1), construct a mathematical model for spectral CT image denoising; (3) use the generalized total variable As a regularization prior, combine the mathematical model obtained in step (2) to construct an objective function for image denoising; (4) use splitting The Bregman algorithm is solved to complete the denoising of the energy spectrum CT image. The invention realizes the denoising of the energy spectrum CT image by using the base material decomposition model satisfied by the high and low energy images in the energy spectrum CT and combining the energy spectrum CT image information and the base material image information.

Description

A low-dose energy spectrum CT Image Denoising Methods

technical field

The invention relates to an image processing method of medical images, in particular to a low-dose energy spectrum CT image denoising method.

Background technique

With the rapid development of CT technology, dual-energy CT scanning technology based on spectral integration detectors and photon counting detection technology based on energy-resolution detectors have enabled spectral CT imaging to be realized. Spectral CT is one of the development directions of CT imaging technology in the future, because spectral CT can not only obtain the information of the internal attenuation coefficient of the material, but also obtain the information of the material composition through reconstruction. Spectral CT can transfer from traditional morphological diagnosis to functional diagnosis. For example, it can find lesions that cannot be found by conventional CT, realize ultra-early detection of tumors, and perform qualitative diagnosis and quantitative analysis of tumors. In addition, spectral CT can solve many defects in conventional CT imaging, such as removing beam hardening and metal artifacts.

Spectral CT imaging can only be applied clinically under low-dose conditions, so it is necessary to find an efficient low-dose imaging method. There are two main types of existing methods for realizing low-dose spectral CT imaging. Among them, it is the simplest method to reduce the tube current (mA) and tube voltage (kV) as much as possible during the data acquisition process. The reduction of tube current will lead to a substantial increase in the intensity of photon noise in the energy spectrum projection data and the influence of electronic noise will be more prominent; changing the tube voltage can affect the penetration of X-rays into human tissues, thereby affecting the image quality of various tissues. The other is to use the statistical reconstruction method. Taking advantage of its physical model accuracy and insensitivity to noise, it can reconstruct images in the case of irregular sampling and data loss, improve the noise of the final image, and improve the spatial resolution of the reconstructed image. Due to the large amount of spectral CT projection data, this method has a large amount of calculation and a very long reconstruction time, which is difficult to meet the requirements of real-time interaction in clinical practice.

Therefore, in view of the deficiencies of the existing technologies, a low-dose spectral CT image denoising method is provided, which can improve the density measurement accuracy of the base material and can realize high-quality imaging of the spectral CT image under the low-dose scanning protocol.

Contents of the invention

The purpose of the present invention is to avoid the deficiencies of the prior art and provide a low-dose energy spectrum CT image denoising method, which can improve the image quality of the matrix material density image, and can realize the high-quality energy spectrum CT image under the low-dose scanning protocol imaging.

The above object of the present invention is achieved through the following technical means.

A method for denoising a low-dose spectral CT image is provided, comprising the following steps,

(1) Obtain low-energy CT projection data and high-energy CT projection data of the imaging object under low-dose radiation, and perform CT image reconstruction on the low-energy CT projection data and high-energy CT projection data respectively to obtain low-energy CT images and high-energy CT images , where H represents high energy, L represents low energy;

(2) Construct a mathematical model for spectral CT image denoising according to the matrix decomposition model satisfied by the reconstructed data in step (1);

(3) Using generalized total variation as a regularization prior, combined with the mathematical model obtained in step (2) to construct an objective function for image denoising;

(4) The objective function for spectral CT image denoising constructed in step (3) is solved by using the split Bregman algorithm to complete the energy spectral CT image denoising.

Preferably, the matrix decomposition model in the above step (2) is:

The Mass Absorption Function of Matter to X-Photons Expressed by the mass absorption function of any two species, the base species pair: ,in and are the mass absorption functions of the two substances, are the densities of the required substrates, respectively, and The value of is independent of the energy of the X photon;

According to the matrix material decomposition model, for the high-energy CT projection data and low-energy CT projection data of the spectral CT in step (1), the expression of the mass absorption function of the corresponding material is: ,

Define the substance mass absorption function matrix , the matrix mass absorption function matrix , the base mass density matrix ;

C is directly obtained by calculating the inverse matrix, the formula is , define the inverse matrix form of the matrix mass absorption matrix A .

Preferably, in the above step (3), the second-order generalized total variation is used as a priori, and the definition of the second-order generalized total variation is:

;

in is a non-negative weighting coefficient; Auxiliary parameters introduced for generalized total variation, and take Represents a symmetric gradient operator, where Represents the gradient operator, Indicates the matrix transpose operation;

The objective function for image denoising constructed in the step (3) Specifically: , where X represents the spectral CT image obtained after denoising, Y is the measured spectral CT image data, and is a regularization parameter used to characterize the generalized full variational regularization strength.

Preferably, the specific calculation process of the split Bregman algorithm in the above step (4) is:

Introduce formula A, formula B and formula C for iterative solution,

,in is an incoming vector value, Represents the residual, n represents the number of iteration steps;

The specific iterative process is carried out according to the following steps:

(4.1) Let n = 0,

(4.2) According to the formulas A and B , solve by the primal dual algorithm ;

(4.3) will step (4.1) obtained Substitute into formula C to solve ;

(4.4) Determine whether the iteration is terminated

Judging whether n is equal to N, if n is equal to N, the iteration is terminated, and the current result is used as the energy spectrum CT image after denoising;

If n is less than N, go to step (4.5);

(4.5) Let n = n +1, return to step (4.2).

Preferably, the above step (1) is also provided with a registration processing step, specifically:

It is judged whether the obtained low-energy CT projection data and high-energy CT projection data have a position offset, and if there is a position offset, the low-energy CT projection data and the high-energy CT projection data are registered by a data registration method.

The low-dose spectral CT image denoising method of the present invention includes the following steps: (1) Acquiring low-energy CT projection data and high-energy CT projection data of the imaging object under low-dose radiation, and respectively performing low-energy CT projection data and high-energy CT projection data High-energy CT projection data for CT image reconstruction to obtain low-energy CT images and high-energy CT images , where H represents high energy, and L represents low energy; (2) According to the matrix decomposition model satisfied by the reconstructed data in step (1), construct a mathematical model for spectral CT image denoising; (3) use the generalized total variable As a regularization prior, combine the mathematical model obtained in step (2) to construct an objective function for image denoising; (4) use splitting The Bregman algorithm is solved to complete the denoising of the energy spectrum CT image. The invention realizes the denoising of the energy spectrum CT image by using the base material decomposition model satisfied by the high and low energy images in the energy spectrum CT and combining the energy spectrum CT image information and the base material image information. The present invention can use low-dose emission while still ensuring high-quality energy spectrum CT denoising images, the method of the present invention has good robustness, and has good effects in both noise elimination and artifact suppression.

Description of drawings

The present invention will be further described by using the accompanying drawings, but the content in the accompanying drawings does not constitute any limitation to the present invention.

Fig. 1 is a schematic flowchart of the method for denoising a low-dose spectral CT image of the present invention.

Fig. 2 is a schematic diagram of an ideal phantom without artifacts and noise; among them, Fig. 2(a) is a schematic diagram of an image without any artifacts and noise under the ideal Clock phantom at 80kVp; Fig. 2(b) is an ideal Clock phantom Schematic diagram of the image without any artifacts and noises under 140kVp.

Fig. 3 is a schematic diagram of the low-dose data of the Clock phantom directly reconstructed by the FBP algorithm; among them, Fig. 3 (a) is a schematic diagram of the image directly reconstructed by the FBP algorithm of the low-dose data of the Clock phantom at 80kVp; Fig. 3 (b) are schematic diagrams of the images of the Clock phantom directly reconstructed from the 140kVp low-dose data using the FBP algorithm.

Fig. 4 is the image schematic diagram obtained by adopting the denoising method of the present invention for the ideal Clock phantom low-dose data; Wherein, Fig. 4 (a) is the image obtained by adopting the denoising method of the present invention for the low-dose data of the Clock phantom under 80kVp, Fig. 4(b) is an image obtained by applying the denoising method of the present invention to the low-dose data of the Clock phantom at 140kVp.

Fig. 5 is a schematic diagram of the water-based map and bone-based map obtained by the decomposition method of the ideal Clock phantom based on the image domain decomposition method; among them, Fig. 5(a) is a schematic diagram of the water-based map, and Fig. 5(b) is a schematic diagram of the bone-based map.

Figure 6 is a schematic diagram of the water-based map and bone-based map obtained by the low-dose Clock phantom based on the image domain decomposition method; wherein, Figure 6(a) is a schematic diagram of the water-based map, and Figure 6(b) is a schematic diagram of the bone-based map.

Figure 7 is a schematic diagram of the water-based map and bone-based map obtained based on the image domain decomposition method after the denoising method of the present invention is used to obtain the results; wherein, Figure 7(a) is a schematic diagram of the water-based map, and Figure 7(b) is a bone-based map schematic diagram.

Fig. 8 is a horizontal midline sectional view of the image part, wherein Fig. 8(a) is a horizontal midline sectional view of an 80kVp image part, and Fig. 8(b) is a horizontal midline sectional view of a 140kVp image part.

Figure 9 is a partial horizontal midline cross-sectional view of the water-based map and a bone-based map, wherein Figure 9(a) is a partial horizontal mid-line cross-sectional view of the water-based map, and Figure 9(b) is a partial horizontal mid-line cross-sectional view of the bone-based map.

Detailed ways

The present invention is further described in conjunction with the following examples.

Example 1 .

A low-dose spectral CT image denoising method, as shown in Figure 1, comprises the following steps,

(1) Obtain low-energy CT projection data and high-energy CT projection data of the imaging object under low-dose radiation, and perform CT image reconstruction on the low-energy CT projection data and high-energy CT projection data respectively to obtain low-energy CT images and high-energy CT images , where H represents high energy, L represents low energy;

(2) Construct a mathematical model for spectral CT image denoising according to the matrix decomposition model satisfied by the reconstructed data in step (1);

(3) Using generalized total variation as a regularization prior, combined with the mathematical model obtained in step (2) to construct an objective function for image denoising;

(4) The objective function for spectral CT image denoising constructed in step (3) is solved by using the split Bregman algorithm to complete the energy spectral CT image denoising.

Preferably, the above step (1) is also provided with a registration processing step, specifically: judging whether there is a positional offset between the obtained low-energy CT projection data and high-energy CT projection data, and adopting data registration when there is a positional offset The method of low-energy CT projection data and high-energy CT projection data is registered.

Among them, the matrix decomposition model in step (2) is:

The Mass Absorption Function of Matter to X-Photons Expressed by the mass absorption function of any two species, the base species pair: ,in and are the mass absorption functions of the two substances, are the densities of the required substrates, respectively, and The value of is independent of the energy of the X photon;

According to the matrix material decomposition model, for the high-energy CT projection data and low-energy CT projection data of the spectral CT in step (1), the expression of the mass absorption function of the corresponding material is: ,

Define the substance mass absorption function matrix , the matrix mass absorption function matrix , the base mass density matrix ;

C is directly obtained by calculating the inverse matrix, the formula is , define the inverse matrix form of the matrix mass absorption matrix A .

Among them, in step (3), the second-order generalized total variation is used as a priori, and the definition of the second-order generalized total variation is:

;

in is a non-negative weighting coefficient; Auxiliary parameters introduced for generalized total variation, and take Represents a symmetric gradient operator, where Represents the gradient operator, Represents a matrix transpose operation.

The objective function for image denoising constructed in step (3) Specifically: , where X represents the spectral CT image obtained after denoising, Y is the measured spectral CT image data, and is a regularization parameter used to characterize the generalized full variational regularization strength.

The specific calculation process of the split Bregman algorithm in step (4) is:

Introduce formula A, formula B and formula C for iterative solution,

,in is an incoming vector value, Represents the residual, and n represents the number of iteration steps.

The specific iterative process is carried out according to the following steps:

(4.1) Let n = 0,

(4.2) According to the formulas A and B , solve by the primal dual algorithm ;

(4.3) will step (4.1) obtained Substitute into formula C to solve ;

(4.4) Determine whether the iteration is terminated

Judging whether n is equal to N, if n is equal to N, the iteration is terminated, and the current result is used as the energy spectrum CT image after denoising;

If n is less than N, go to step (4.5);

(4.5) Let n = n +1, return to step (4.2).

The invention realizes the denoising of the energy spectrum CT image by using the base material decomposition model satisfied by the high and low energy images in the energy spectrum CT and combining the energy spectrum CT image information and the base material image information. The present invention can use low-dose emission while still ensuring high-quality energy spectrum CT denoising images, the method of the present invention has good robustness, and has good effects in both noise elimination and artifact suppression.

Example 2 .

Taking the digital phantom data simulated by computer as an example to describe the specific implementation process of the method of the present invention, as shown in FIG. 1 , the implementation process of this embodiment is as follows.

(1) Use the Clock digital phantom to simulate and generate low-dose spectral CT projection data for verification and evaluation of the algorithm of the present invention. In this embodiment, the distances from the X-ray source of the simulated CT machine to the rotation center and the detector are: 570.00mm and 1040.00mm respectively, the number of detection elements is 672, the size is 1.407mm, and the number of detection angular samples for one rotation for 1160. The Clock phantom image size is 512×512. The projection data of 80kVp and 140kVp with the size of 1160×672 were respectively generated by CT system simulation. The variance of the electronic noise of the system is 10.0.

(2) Data reconstruction: Use the obtained system parameters to correct the detection data, perform logarithmic transformation, and perform filtered back-projection reconstruction.

(3) Construct image denoising model: Mathematically model the base material decomposition model satisfied by the reconstructed spectral CT image data obtained in step (2), complete the design of the generalized total variational prior term, and construct the A Constrained Objective Function for Spectral CT Image Denoising , , where X represents the spectral CT image obtained after denoising, Y is the measured spectral CT image data, and is a regularization parameter, in the embodiment of the present invention, , , used to characterize the generalized total variational regularization strength.

The specific form of the matrix decomposition model is:

The Mass Absorption Function of Matter to X-Photons Expressed by the mass absorption function of any two species, the base species pair: ,in and are the mass absorption functions of the two substances, are the densities of the required substrates, respectively, and The value of is independent of the energy of the X photon;

According to the matrix material decomposition model, for the high-energy CT projection data and low-energy CT projection data of the spectral CT in step (1), the expression of the mass absorption function of the corresponding material is: ,

Define the substance mass absorption function matrix , the matrix mass absorption function matrix , the base mass density matrix ;

C is directly obtained by calculating the inverse matrix, the formula is , define the inverse matrix form of the matrix mass absorption matrix A .

The specific process of constructing the above-mentioned generalized total variation regularization prior is: use the second-order generalized total variation as the prior, and its definition is: ;in , is a non-negative weighting coefficient; the generalized total variation introduces auxiliary parameters , and take .

(3) Complete denoising: On the basis of the related model constructed in step (3), the split Bregman algorithm is used for image denoising processing. The specific calculation process is as follows:

Introduce formula A, formula B and formula C for iterative solution,

,in is an incoming vector value, Represents the residual, and n represents the number of iteration steps.

The specific iterative process is carried out according to the following steps:

(4.1) Let n = 0,

(4.2) According to the formulas A and B , solve by the primal dual algorithm ;

(4.3) will step (4.1) obtained Substitute into formula C to solve ;

(4.4) Determine whether the iteration is terminated

Judging whether n is equal to N, if n is equal to N, the iteration is terminated, and the current result is used as the energy spectrum CT image after denoising;

If n is less than N, go to step (4.5);

(4.5) Let n = n +1, return to step (4.2).

In order to verify the effect of the reconstruction method of the present invention, the results of this example are shown in Figure 2-Figure 7, in which: Figure 2 (a) and Figure 2 (b) are the ideal Clock phantom at 80kVp and 140kVp without Image of any artifacts and noise. Figure 3(a) and Figure 3(b) are the images obtained by directly reconstructing the Clock phantom at 80kVp and 140kVp low-dose data using the FBP algorithm, respectively. It can be seen that the reconstructed image has serious statistical noise due to the reduction of the dose. Figure 4(a) and Figure 4(b) are the water base map and bone base map reconstructed by the ideal Clock phantom based on the projection domain decomposition method, respectively. Figure 5(a) and Figure 5(b) are the water base map and bone base map reconstructed by the low-dose Clock phantom based on the projection domain decomposition method. Similarly, the noise in the original high and low energy images leads to the density of the base material There is also severe noise in the image. Figure 6(a) and Figure 6(b) are the water-based map and bone-based map obtained by the reconstruction method of the present invention respectively. It can be seen from the two reconstructed images in Figure 6 that the results obtained by using the method of the present invention to reconstruct are suppressing The effect of noise and artifacts is obvious.

Figures 7(a) and 7(b) show the horizontal midline profiles corresponding to the matrix material reconstruction images in Figures 4, 5, and 6. Since the entire profile contains 512 pixels, it is difficult to distinguish them all. For each method, only one section is intercepted when it is only displayed. For the water-based map, the interval is [189,320]. For bone-based maps, the interval is [147,189]. It can be seen from Fig. 7 that in the water-based map and bone-based map, the reconstruction value of the method of the present invention is closer to the ideal value regardless of the background area or the target area.

The invention realizes the denoising of the energy spectrum CT image by using the base material decomposition model satisfied by the high and low energy images in the energy spectrum CT and combining the energy spectrum CT image information and the base material image information. The present invention can use low-dose emission while still ensuring the generation of high-quality spectral CT denoising images, the method of the present invention has good robustness, and has excellent performance in both noise elimination and artifact suppression .

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than limit the protection scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that Modifications or equivalent replacements are made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims (5)

1. a low-dose energy spectrum CT image denoising method, is characterized in that: comprise the steps, (1) Obtain low-energy CT projection data and high-energy CT projection data of the imaging object under low-dose radiation, and perform CT image reconstruction on the low-energy CT projection data and high-energy CT projection data respectively to obtain low-energy CT images and high-energy CT images , where H represents high energy, L represents low energy; (2) Construct a mathematical model for spectral CT image denoising according to the matrix decomposition model satisfied by the reconstructed data in step (1); (3) Using generalized total variation as a regularization prior, combined with the mathematical model obtained in step (2) to construct an objective function for image denoising; (4) The objective function for spectral CT image denoising constructed in step (3) is solved by using the split Bregman algorithm to complete the energy spectral CT image denoising. 2. The low-dose energy spectrum CT image denoising method according to claim 1, characterized in that: The matrix decomposition model in the step (2) is: The Mass Absorption Function of Matter to X-Photons Expressed by the mass absorption function of any two species, the base species pair: ,in and are the mass absorption functions of the two substances, are the densities of the required substrates, respectively, and The value of is independent of the energy of the X photon; According to the matrix material decomposition model, for the high-energy CT projection data and low-energy CT projection data of the spectral CT in step (1), the expression of the mass absorption function of the corresponding material is: , Define the substance mass absorption function matrix , the matrix mass absorption function matrix , the base mass density matrix ; C is directly obtained by calculating the inverse matrix, the formula is , define the inverse matrix form of the matrix mass absorption matrix A . 3. The low-dose energy spectrum CT image denoising method according to claim 2, characterized in that: In the step (3), the second-order generalized total variation is used as a priori, and the definition of the second-order generalized total variation is: ; in is a non-negative weighting coefficient; Auxiliary parameters introduced for generalized total variation, and take Represents a symmetric gradient operator, where Represents the gradient operator, Indicates the matrix transpose operation; The objective function for image denoising constructed in the step (3) Specifically: , where X represents the spectral CT image obtained after denoising, Y is the measured spectral CT image data, and is a regularization parameter used to characterize the generalized full variational regularization strength. 4. according to the low-dose energy spectrum CT image denoising method according to claim 3, it is characterized in that: The specific calculation process of the split Bregman algorithm in the step (4) is: Introduce formula A, formula B and formula C for iterative solution, ,in is an incoming vector value, Represents the residual, n represents the number of iteration steps; The specific iterative process is carried out according to the following steps: (4.1) Let n = 0, (4.2) According to the formulas A and B , solve by the primal dual algorithm ; (4.3) will step (4.1) obtained Substitute into formula C to solve ; (4.4) Determine whether the iteration is terminated Judging whether n is equal to N, if n is equal to N, the iteration is terminated, and the current result is used as the energy spectrum CT image after denoising; If n is less than N, go to step (4.5); (4.5) Let n = n +1, return to step (4.2). 5. The low-dose energy spectrum CT image denoising method according to claim 1, characterized in that: The step (1) is also provided with a registration processing step, specifically: It is judged whether the obtained low-energy CT projection data and high-energy CT projection data have a position offset, and if there is a position offset, the low-energy CT projection data and the high-energy CT projection data are registered by a data registration method.