Photon counting CT ring artifact correction method based on double-domain optimization

Abstract

The invention relates to a photon counting CT ring artifact correction method based on double-domain optimization. The method comprises the following steps: 1) pre-correcting a sinogram acquired by a photon counting CT; 2) for the pre-corrected sinogram, combining the correction parameters to finely adjust the response of the sinogram detector direction; 3) reconstructing the sinogram after fine tuning, converting the reconstructed CT image into an HU value, and converting the reconstructed CT image into a polar coordinate system; (4) calculating the absolute value of the gradient of the reconstructed image in the polar radial direction under the polar coordinate system, optimizing the correction coefficient in the step (2) by taking the minimization of the average value of the absolute value of the gradient of the whole image as a target, and carrying out iterative calculation until convergence; and 5) applying the correction coefficient obtained by final convergence to the sinogram subjected to pre-correction in the step 1, and performing image reconstruction to obtain a corrected CT image. According to the method, the polynomial correction coefficient is optimized through double-domain joint, and the problem of inconsistent detector response is further corrected, so that the residual ring artifact of the pre-corrected photon counting CT image is removed, and the image quality and the diagnosis accuracy are improved.

Description

Technical Field

[0001] The invention belongs to the field of image processing, and in particular relates to a photon counting CT annular artifact correction method based on dual-domain optimization. Background Art

[0002] Photon counting CT detectors do not require a physical grid to separate pixels, so the detector pixels are smaller, achieving high spatial resolution and improving contrast-to-noise ratio for clinical applications such as oncology, cardiology, and neurology. In these applications, accurate and reliable imaging is essential. However, due to different factors such as inconsistent detector response, actual threshold voltage of each pixel, and data acquisition errors, the acquired and reconstructed CT images contain high- and low-frequency ring artifacts. These artifacts appear as uneven concentric circles in CT images, which have nothing to do with the actual biological tissue structure and can seriously affect the doctor's diagnosis.

[0003] The ring artifacts in most photon counting CT images can be effectively corrected by physics-based projection domain pre-correction. However, due to factors such as correction error and drift of photon counting detector response over time, residual ring artifacts still exist in the pre-corrected CT images. The present invention aims to remove the residual ring artifacts in the pre-corrected photon counting CT images. Summary of the invention

[0004] In order to solve the problem of removing residual ring artifacts in photon counting CT images after physical pre-correction without damaging image quality, the present invention provides a photon counting CT ring artifact correction method based on dual-domain optimization.

[0005] To achieve the above technical objectives, the technical solution adopted by the present invention is: a photon counting CT ring artifact correction method based on dual-domain optimization, the method comprising the following steps:

[0006] Step 1: Pre-correct the sinusoidal image acquired by photon counting CT;

[0007] Step 2: For the pre-corrected sinogram, fine-tune the response of the sinogram detector direction in combination with the correction parameters;

[0008] Step 3: Reconstruct the fine-tuned sinusoidal image, convert the reconstructed CT image into HU value, and convert the reconstructed CT image into polar coordinate system;

[0009] Step 4: Calculate the absolute value of the gradient of the reconstructed image along the polar direction in the polar coordinate system, optimize the correction coefficient in step 2 with the goal of minimizing the average absolute value of the gradient of the entire image, and iterate until convergence.

[0010] Step 5: The correction coefficient obtained by the final convergence is applied to the pre-corrected sinusoidal image in step 1 to perform image reconstruction to obtain the corrected CT image.

[0011] Beneficial effects of the present invention:

[0012] Compared with the physics-based projection domain annular artifact correction method, the present invention utilizes the characteristic that annular artifacts appear as strip artifacts that are easy to identify and quantify in the polar coordinate system. The polynomial correction coefficients are continuously optimized by combining the projection domain and the image domain, and the correction coefficients are applied to the projection image, thereby removing the residual annular artifacts of the pre-corrected photon counting CT image, making the CT image quality higher and the tissue structure recognition better. At the same time, the present method has a wide range of applications, does not require energy spectrum information and training data, and can also be used in CT systems based on traditional energy integrating detectors. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the process of the present invention;

[0014] Figure 2 This is a diagram showing the appearance of ring artifacts in different coordinate systems;

[0015] Figure 3 This is a comparison of the real mouse data before and after the ring artifact correction at different energy thresholds. DETAILED DESCRIPTION

[0016] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0017] Example: Figure 1 As shown, the present invention is a photon counting CT ring artifact correction method based on dual-domain optimization, comprising the following steps:

[0018] Step 1: Pre-correct the sinusoidal images acquired by photon counting CT.

[0019] Step 2: For the pre-corrected sinogram, fine-tune the response of the sinogram detector direction in combination with the correction parameters. The correction coefficient uses a polynomial form that does not contain a zero-order term. The first-order term in the initial correction coefficient is 1, and the remaining high-order terms are set to 0. For pixels at the same detector position at different projection angles, use the same fine-tuning parameters for polynomial correction:

[0020]

[0021] The index i represents the projection angle direction, j represents the detector pixel direction, S i，j is the projected pixel value of the i-th row and j-th column, c p，j is the p-th order polynomial correction coefficient of the j-th column. The fine-tuned sinusoidal graph S′ is obtained by using the polynomial correction i，j .

[0022] Furthermore, the step 3 specifically includes the following process:

[0023] Step 3.1: The fine-tuned projection image S′ i，j Perform filtered back projection reconstruction to obtain the reconstructed image:

[0024] I μ (x, y) = R -1 (S′ i，j )

[0025] Where R -1 It is a filtered back-projection algorithm used to reconstruct CT images. μ (x, y) is the reconstructed CT image.

[0026] Step 3.2: Convert the reconstructed CT image to HU value:

[0027]

[0028] where μ w It is the reference value of the linear attenuation coefficient of water. HU The discretized image corresponding to (x, y) is represented as I HU (m, n), where m is the index along the horizontal direction and n is the index along the vertical direction.

[0029] Step 3.3: Perform polar coordinate transformation on the reconstructed image to obtain a polar coordinate image:

[0030] I P (θ, r) = C2P(I HU (x, y)

[0031] Where C2P(·) represents the transformation from Cartesian coordinate system to polar coordinate system. Figure 2 As shown, the reconstructed graph I HU (x, y) uses a Cartesian coordinate system and contains a ring artifact. The polar coordinate image I P (θ, r ) uses a polar coordinate system and contains the stripe artifacts corresponding to the ring artifacts. P The discretized image corresponding to (θ, r) is represented as I P (k, l), where k is the index along the angular direction and l is the index along the polar direction.

[0032] Furthermore, the step 4 specifically includes the following process:

[0033] Step 4.1: Calculate the absolute value of the gradient of the reconstructed image along the polar direction in the polar coordinate system.

[0034] Step 4.2: Optimize the correction coefficient in step 2 with the goal of minimizing the average value of the absolute value of the gradient of the entire image, and iterate until convergence:

[0035]

[0036] where ||·|| represents the L2 loss, K and L are the polar coordinate images I P (k, l) The number of all pixels along the angle and polar direction. The convergence criterion of the algorithm is that the absolute value of the gradient of the reconstructed image along the polar direction in the polar coordinate system no longer decreases significantly, and the HU difference of the absolute value of the gradient between the previous and next iterations is less than 0.5 HU within a certain range.

[0037] Furthermore, the step 5 specifically includes the following process:

[0038] Step 5.1: Apply the correction coefficients obtained by the final convergence to the pre-corrected sinusoidal graph in step 1:

[0039]

[0040] Step 5.2: Reconstruct the corrected sinusoidal image to obtain the corrected CT image:

[0041]

[0042] Step 5.3: Convert the reconstructed CT image to HU value:

[0043]

[0044] where μ w is the reference value of the linear attenuation coefficient of water.

[0045] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention. It should be noted that the above embodiments are not used to limit the protection scope of the present invention. Equivalent changes or substitutions made on the basis of the above technical solutions fall within the scope of protection of the claims of the present invention.

Claims

1. A photon counting CT ring artifact correction method based on dual-domain optimization, characterized in that: The method comprises the following steps: Step 1: Pre-correct the sinusoidal image acquired by photon counting CT. Step 2: For the pre-corrected sinogram, fine-tune the response of the sinogram detector direction in combination with the correction parameters. Step 3: Reconstruct the fine-tuned sinusoidal image, convert the reconstructed CT image into HU value, and convert the reconstructed CT image into polar coordinate system. Step 4: Calculate the absolute value of the gradient of the reconstructed image along the polar direction in the polar coordinate system, optimize the correction coefficient in step 2 with the goal of minimizing the average value of the absolute value of the gradient of the entire image, and iterate until convergence. Step 5: The correction coefficient obtained by the final convergence is applied to the pre-corrected sinusoidal image in step 1 to perform image reconstruction to obtain the corrected CT image.

2. The photon counting CT ring artifact correction method based on dual-domain optimization according to claim 1 is characterized in that: The correction coefficients in step 2 use a polynomial form that does not contain a zero-order term. The first-order term in the initial correction coefficient is 1, and the remaining high-order terms are set to 0. The same fine-tuning parameters are used for polynomial correction for pixels at the same detector position at different projection angles: The index i represents the projection angle direction, j represents the detector pixel direction, S i，j is the pixel value of the sinusoidal graph in the i-th row and j-th column, c p，j is the p-th order polynomial correction coefficient of the j-th column. The polynomial is used to correct the fine-tuned sinusoidal pixel value S′ i，j .

3. The photon counting CT ring artifact correction method based on dual-domain optimization according to claim 1, characterized in that: Step 3 specifically includes the following process: Step 3.1: The fine-tuned projection image S′ i，j Perform filtered back projection reconstruction to obtain the reconstructed image: I μ (x,y)=R -1 (S′ i，j ) Where R -1 It is a filtered back-projection algorithm used to reconstruct CT images. μ (x, y) is the reconstructed CT image, Step 3.2: Convert the reconstructed CT image to HU value: where μ w is the reference value of the linear attenuation coefficient of water, I HU The discretized image corresponding to (x, y) is represented as I HU (m, n), where m is the index along the horizontal direction and n is the index along the vertical direction, Step 3.3: Perform polar coordinate transformation on the reconstructed image to obtain a polar coordinate image: I P (θ,r)=C2P(I HU (x,y)) Where C2P(·) represents the transformation from Cartesian coordinate system to polar coordinate system, and the reconstructed graph I HU (x, y) uses a Cartesian coordinate system and contains a ring artifact. The polar coordinate image I P (θ, r) uses a polar coordinate system and contains the stripe artifact corresponding to the ring artifact, I P The discretized image corresponding to (θ, r) is represented as I P (k, l), where k is the index along the angular direction and l is the index along the polar direction.

4. The photon counting CT ring artifact correction method based on dual-domain optimization according to claim 1, characterized in that: Step 4 specifically includes the following process: Step 4.1: Calculate the absolute value of the gradient of the reconstructed image along the polar direction in the polar coordinate system. Step 4.2: Optimize the correction coefficient in step 2 with the goal of minimizing the average value of the absolute value of the gradient of the entire image, and iterate until convergence: where ||·|| represents the L2 loss, K and L are the polar coordinate images I P (k, l) the number of all pixels along the angular and epipolar directions, is the optimized polynomial correction coefficient. The convergence criterion of the algorithm is that the absolute value of the gradient of the reconstructed image along the polar direction in the polar coordinate system no longer decreases significantly, and the HU difference of the absolute value of the gradient of the two iterations before and after is less than 0.5 HU within a certain range.

5. The photon counting CT ring artifact correction method based on dual-domain optimization according to claim 1, characterized in that: Step 5 specifically includes the following process: Step 5.1: The final converged correction coefficient The pre-corrected sinusoidal graph S used in step 1 i，j , Step 5.2: The final corrected sinusoidal graph Perform image reconstruction to obtain corrected CT images Step 5.3: Convert the reconstructed CT image to HU value: where μ w is the reference value of the linear attenuation coefficient of water, It is the final corrected HU value to reconstruct the CT image.

Images