A Method for Correcting Beam Hardening Artifacts in Spectral CT Based on Image Inpainting

Abstract

This invention discloses a method for correcting beam hardening artifacts in spectral CT based on image inpainting. The method includes: scanning the object under test to obtain original projection data; using an image inpainting algorithm to eliminate high-attenuation material trajectories in the original projection data as input data; reprojecting the data using a template image obtained through threshold segmentation and an initial model spectrum, and iteratively updating the model spectrum parameters with the input data; reprojecting the segmented image of the original projection data using the estimated energy spectrum, and obtaining the artifact portion by subtracting the pre-set single-energy spectrum reprojection monochromatic projection data from the reprojected multi-energy projection data to correct the original projection data; finally, filtering and backprojecting the corrected projection data to obtain the corrected CT image. This invention solves the problem that software methods for beam hardening artifact correction have poor correction effects on CT images of objects containing high-attenuation materials.

Description

Technical Field

[0001] This invention relates to the field of X-ray CT imaging technology, and specifically to a method for correcting beam hardening artifacts in spectral CT based on image restoration. Background Technology

[0002] Spectral CT technology, also known as computed tomography (CT), uses X-rays to scan cross-sectional areas of an object under test. After a flat-panel detector receives the X-rays passing through the cross-section, the signal is converted into a digital signal, and finally, a computer processes the data to obtain the X-ray absorption value. This technology allows for the examination of tomographic images of the internal structure of an object without contact or damage. It is currently widely used in the medical and industrial fields.

[0003] Normally, X-rays emitted from an X-ray source have a continuous spectrum. When multi-energy X-rays pass through the object under test, low-energy X-rays are more easily absorbed by high-attenuation materials, such as metals like iron, leading to beam hardening and producing beam hardening artifacts. These artifacts mainly manifest as cupping artifacts and streak artifacts, resulting in the loss of structural information about the object and severely reducing diagnostic value.

[0004] Currently, there are increasingly more methods for beam hardening artifact correction, mainly divided into hardware and software methods.

[0005] Hardware methods involve adding correction tools to the CT system, such as filters and water bag correction. However, hardware methods reduce the signal-to-noise ratio and have limited suppression effects, failing to completely solve the beam hardening problem.

[0006] Software methods, based on the mechanism of beam hardening, correct projected or reconstructed data. These methods primarily include polynomial fitting, iterative correction, dual-energy methods, and deep learning. However, current software methods only show good correction results for objects that have already been penetrated. When the object under test contains highly attenuating materials, commonly used software methods often perform poorly.

[0007] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0008] To overcome the shortcomings of existing technologies, a method for correcting beam hardening artifacts in spectral CT based on image restoration is provided to address the problem that software methods for beam hardening artifact correction are not effective for CT images of objects containing highly attenuated materials.

[0009] To achieve the above objectives, a method for correcting beam hardening artifacts in spectral CT based on image inpainting is provided, comprising the following steps:

[0010] Obtain the original projection data of the cross-section of the test object containing high-attenuation material;

[0011] An image inpainting algorithm is used to remove the high-attenuation material portion from the original projection data to obtain preprocessed projection data;

[0012] The preprocessed projection data is reconstructed using a reconstruction algorithm to obtain a first reconstructed image;

[0013] The first reconstructed image is segmented to obtain a template image;

[0014] A set of continuous energy spectra is initialized using energy spectrum generation software as the initial model spectrum, and multi-energy reprojection is calculated with the template image.

[0015] The parameters of the initial model spectrum are optimized by minimizing the square of the L2 norm of the multi-energy reprojection and the preprocessed projection data using an iterative algorithm to obtain the estimated energy spectrum of the multi-energy reprojection that is closest to the preprocessed projection data.

[0016] The original projection data is reconstructed using the reconstruction algorithm to obtain a second reconstructed image;

[0017] The second reconstructed image is segmented to obtain a segmented image, and the estimated energy spectrum and the segmented image are used to calculate the reprojection data containing high-attenuation material;

[0018] Acquire the monoenergetic projection data of the object under test, and calculate the beam hardening artifact data based on the monoenergetic projection data and the reprojection data.

[0019] Based on the original projection data and the partial data of beam hardening artifacts, beam hardening artifact correction data is calculated.

[0020] The beam hardening artifact correction data is reconstructed using the reconstruction algorithm to obtain a beam hardening artifact correction CT image.

[0021] Furthermore, the image restoration algorithm is an image linear interpolation algorithm.

[0022] Furthermore, the reconstruction algorithm is a filtered back-projection reconstruction algorithm.

[0023] Furthermore, the initial model spectrum and the template image are used to calculate the multi-energy reprojection using a first formula, which is:

[0024]

[0025] Where S(E) is the continuous energy spectrum, τ m (E) is the mass attenuation coefficient of the material corresponding to the pixel in the template image, A m (u) is the line integral of the mass density of the corresponding material.

[0026] Furthermore, the square of the L2 norm is the objective function of the iterative algorithm, and the objective function is:

[0027]

[0028] The objective function is optimized using the downhill simplex method.

[0029] The beneficial effect of the present invention is that the method for correcting beam hardening artifacts in spectral CT based on image restoration does not require a complex experimental environment or procedure, nor a large amount of laboratory data. It can not only accurately estimate the energy spectrum with only a set of CT system projection data, but also use the estimated energy spectrum to correct beam hardening artifacts in spectral CT. Attached Figure Description

[0030] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0031] Figure 1 This is a flowchart of a method for correcting beam hardening artifacts in spectral CT based on image restoration, according to an embodiment of the present invention.

[0032] Figure 2 This is a schematic diagram of the original projection data in an embodiment of the present invention.

[0033] Figure 3 This is a set of energy spectrum diagrams generated using SpekCalc software according to an embodiment of the present invention.

[0034] Figure 4 This is a schematic diagram of the estimated energy spectrum and the reference energy spectrum in an embodiment of the present invention.

[0035] Figure 5a The image shown is the original CT image from an embodiment of the present invention.

[0036] Figure 5b This is a corrected CT image according to an embodiment of the present invention. Detailed Implementation

[0037] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0038] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0039] Reference Figure 1 As shown, this invention provides a method for correcting beam hardening artifacts in spectral CT based on image restoration, comprising the following steps:

[0040] S1. Obtain the original projection data of the cross-section of the object under test containing high-attenuation material.

[0041] A 360° CT scan of the object under test is performed to acquire the raw projection data R of the cross-sectional plane of the object containing high-attenuation material. a ,like Figure 2 As shown.

[0042] Original projection data R a It is a sinusoidal image, with the horizontal axis representing the number of scanning angles of the CT system.

[0043] In this embodiment, a GE Maxiray 125CT system is used to perform a CT scan on the object under test at 80Kvp conditions to obtain the raw projection data.

[0044] In this embodiment, the actual mold uses plexiglass as the base material and iron rods as the embedded material.

[0045] S2. Use an image inpainting algorithm to remove the high-attenuation material trajectories in the original projection data to obtain preprocessed projection data.

[0046] Preprocessed projection data is used as input data R for the iterative algorithm. u .

[0047] As a preferred implementation method, the image restoration algorithm is an image linear interpolation algorithm.

[0048] Specifically, a linear interpolation image inpainting algorithm is used to eliminate high-attenuation material trajectories in the original projection data.

[0049] Specifically, high-attenuation materials, such as metals like iron, appear as highlighted areas in the original projection data, such as... Figure 2 The highlighted trajectory.

[0050] Specifically, linear interpolation image inpainting algorithms predict pixel values ​​for high-attenuation material trajectories using known surrounding pixels. Interpolation needs to be performed simultaneously in two directions: first horizontal interpolation, then vertical interpolation.

[0051] S3. Use a reconstruction algorithm to reconstruct the preprocessed projection data to obtain the first reconstructed image.

[0052] The reconstruction algorithm is the FBP reconstruction algorithm, and the specific FBP reconstruction algorithm is the filtered back projection reconstruction algorithm.

[0053] For input data R u The FBP reconstruction algorithm is used to reconstruct the image, and the first reconstructed image is obtained. Then, the material is segmented by pixels using a thresholding technique to obtain a template image for the iterative algorithm.

[0054] S4. Segment the first reconstructed image to obtain the template image.

[0055] Thresholding segmentation is used to segment the pixels of the first reconstructed image to obtain a template image for the iterative algorithm.

[0056] In this embodiment, the threshold segmentation adopts adaptive threshold segmentation, which distinguishes the pixel regions corresponding to each material based on the image pixel values.

[0057] S5. Use energy spectrum generation software to initialize a set of continuous energy spectra as the initial model spectrum, and calculate the multi-energy reprojection with the template image.

[0058] A set of continuous energy spectra was initialized using the energy spectrum generation software SpekCalc as the initial model spectrum S. es (E), such as Figure 3 As shown.

[0059] In this example, the energy spectrum generation software SpekCalc initializes a set of energy spectra by setting a voltage value of 80Kvp and different aluminum filter thicknesses, and obtains the normalized energy spectrum by using a summation normalization method.

[0060] Specifically, summation normalization refers to summing the number of photons corresponding to each energy level in an energy spectrum, and then dividing the number of photons at each energy level by the summation value.

[0061] In this example, the initial model spectrum is obtained by weighting and summing the initial energy spectra. The weighted sum formula for the initial model spectrum is:

[0062]

[0063] And initialize c1=1 to a set of energy spectra S i (E) Initialize to the initial model spectrum.

[0064] The multi-energy reprojection R is obtained by calculating the initial model spectrum and template image using the first formula. p The first formula is:

[0065]

[0066] Where S(E) is the continuous energy spectrum, τ m (E) is the mass attenuation coefficient of the material corresponding to the pixel in the template image, A m (u) is the line integral of the mass density of the corresponding material.

[0067] The mass decay factor and mass density were obtained from the National Institute of Standards and Technology (NIST).

[0068] S6. The parameters of the initial model spectrum are optimized by minimizing the square of the L2 norm of the multi-energy reprojection and preprocessed projection data through an iterative algorithm to obtain the estimated energy spectrum of the multi-energy reprojection that is closest to the preprocessed projection data.

[0069] Specifically, the multi-energy reprojection R is minimized through an iterative algorithm. p and input data R u The square of the L2 norm is used to optimize the energy spectrum model S. es The parameter c of (E) is used to obtain the multi-energy reprojection R. p Closest to input data R u The estimated energy spectrum.

[0070] The square of the L2 norm is the objective function of the iterative algorithm, and the objective function is:

[0071]

[0072] The objective function is optimized using the downhill simplex method.

[0073] Specifically, the square of the L2 norm is the sum of squared errors, representing the predicted value R. p and the true value R u The gap between them.

[0074] Specifically, the downhill simplex method is an optimization algorithm for unconstrained optimization problems that uses operations such as reflection, expansion, and contraction to find the minimum value of the objective function. The downhill simplex method does not rely on the derivative of the objective function, making its implementation relatively simple. Combining the downhill simplex method with an objective function based on the square of the L2 norm can effectively solve minimization problems.

[0075] In this example, the estimated energy spectrum obtained through the optimization algorithm is evaluated using the MSE (Mean Sequence of Effect) metric. The MSE metric is used to assess the accuracy of the estimated energy spectrum; the MSE calculation formula is as follows:

[0076]

[0077] Estimated energy spectrum, reference energy spectrum, and calculated MSE values ​​are as follows: Figure 4 As shown.

[0078] S7. Use the reconstruction algorithm to reconstruct the original projection data to obtain the second reconstructed image.

[0079] The original projection data R a The second reconstructed image is obtained by applying the FBP reconstruction algorithm.

[0080] S8. Segment the second reconstructed image to obtain a segmented image, and use the estimated energy spectrum and the segmented image to calculate the reprojection data containing high-attenuation material.

[0081] Thresholding is performed on the second reconstructed image, and the estimated energy spectrum is used to calculate the reprojection data R containing the high-attenuation material using the segmented image. ′ p .

[0082] The first formula is also used to calculate the reprojection data containing highly attenuated materials:

[0083]

[0084] Calculated.

[0085] S9. Obtain the monoenergetic projection data of the object under test, and calculate the beam hardening artifact data based on the monoenergetic projection data and reprojection data.

[0086] Calculate the monoenergetic projection data R using a preset monoenergetic constant. m The beam hardening artifact data δ is obtained by subtracting the single-energy projection data and the reprojection data containing highly attenuating materials. BHC .

[0087] The formula for using unidirectional projection data is:

[0088]

[0089] The calculations show that S and τ are preset constants.

[0090] In this example, the energy spectrum is estimated by summation and normalization, so the preset monoenergetic constant is 1, and the mass decay coefficient of the corresponding material is the decay coefficient value at the energy level of 0.08 MeV.

[0091] S10. Based on the original projection data and the partial data of beam hardening artifacts, calculate and obtain the beam hardening artifact correction data.

[0092] Using the original projection data R a Subtracting beam hardening artifacts from the data δ BHC Thus, beam hardening artifact correction data R are obtained. c .

[0093] S11. The beam hardening artifact correction data is reconstructed using a reconstruction algorithm to obtain beam hardening artifact correction CT images.

[0094] In this example, the corrected CT image is compared with the original CT image, such as... Figure 5a The original CT image. Figure 5b This is the corrected CT image.

[0095] Specifically, the corrected CT image is obtained by reconstructing the projection data after beam hardening artifact correction using FBP, while the original CT image is obtained by reconstructing the original projection data using FBP.

[0096] The present invention provides a method for correcting beam hardening artifacts in spectral CT based on image restoration. It does not require a complex experimental environment or procedure, nor does it require a large amount of laboratory data. It can not only accurately estimate the energy spectrum with only a set of CT system projection data, but also use the estimated energy spectrum to correct beam hardening artifacts in spectral CT.

[0097] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for correcting beam hardening artifacts in spectral CT based on image inpainting, characterized in that, Includes the following steps: Obtain the original projection data of the cross-section of the test object containing high-attenuation material; An image inpainting algorithm is used to remove the high-attenuation material portion from the original projection data to obtain preprocessed projection data; The preprocessed projection data is reconstructed using a reconstruction algorithm to obtain a first reconstructed image; The first reconstructed image is segmented to obtain a template image; A set of continuous energy spectra is initialized using energy spectrum generation software as the initial model spectrum, and multi-energy reprojection is calculated with the template image. The parameters of the initial model spectrum are optimized by minimizing the square of the L2 norm of the multi-energy reprojection and the preprocessed projection data using an iterative algorithm to obtain the estimated energy spectrum of the multi-energy reprojection that is closest to the preprocessed projection data. The original projection data is reconstructed using the reconstruction algorithm to obtain a second reconstructed image; The second reconstructed image is segmented to obtain a segmented image, and the estimated energy spectrum and the segmented image are used to calculate the reprojection data containing high-attenuation material; Acquire the monoenergetic projection data of the object under test, and calculate the beam hardening artifact data based on the monoenergetic projection data and the reprojection data. Based on the original projection data and the partial data of beam hardening artifacts, beam hardening artifact correction data is calculated. The beam hardening artifact correction data is reconstructed using the reconstruction algorithm to obtain a beam hardening artifact correction CT image.

2. The method for correcting beam hardening artifacts in spectral CT based on image restoration according to claim 1, characterized in that, The image restoration algorithm is an image linear interpolation algorithm.

3. The method for correcting beam hardening artifacts in spectral CT based on image restoration according to claim 1, characterized in that, The reconstruction algorithm is a filtered back-projection reconstruction algorithm.

4. The method for correcting beam hardening artifacts in spectral CT based on image restoration according to claim 1, characterized in that, The initial model spectrum and the template image are used to calculate the multi-energy reprojection using a first formula, which is: ； in, It is a continuous energy spectrum. This represents the mass attenuation coefficient of the material corresponding to each pixel in the template image. This is the line integral of the mass density of the corresponding material.

Images