A method for artifact correction of three-dimensional images

Abstract

The application discloses a kind of pseudo-correction methods of three-dimensional image, comprising the following steps: (1) respectively obtaining standard dark field image, air field gain calibration image, air field calibration image of different mean luminance, calibration image of different thickness water model and different thickness aluminum sheet, to obtain the reconstruction image of different mean luminance, different thickness water model and different thickness metal sheet according to this;(2) using least square method to fit each image point in each reconstruction image to obtain gain curve;(3) gain correction is carried out to three-dimensional image with the gain curve obtained.The application solves the problem of annular pseudo of 3D reconstruction image, and improves the image quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, in particular to a method for correcting artifacts of three-dimensional images. BACKGROUND

[0002] C-arm machines are increasingly used by major hospitals due to their low radiation dose, high spatial resolution and other advantages. At the same time, they are increasingly used in the field of orthopedic surgery robots to provide three-dimensional image guidance for doctors or surgery robots.

[0003] In use, it is necessary to ensure that the C-arm provides clear and accurate three-dimensional images. However, in practice, various image artifacts are introduced due to various reasons, resulting in poor image quality and thus misjudgment. Among them, ring artifacts are a common image artifact, mainly caused by bad points, bad lines and uneven gain of the detector. The commonly used correction methods include two categories of pre-processing and post-processing, wherein the pre-processing is mainly through the gain correction method under the condition of empty load; and the post-processing is mainly through extracting ring artifacts in the reconstructed image and then removing them. However, the two schemes currently cannot completely and effectively remove the ring artifacts. SUMMARY

[0004] The present application provides a method for correcting artifacts of three-dimensional images to solve the problem of ring artifacts in 3D reconstructed images, improve the image quality, and ensure the reliability of 3D images for medical diagnosis.

[0005] TECHNICAL SOLUTION

[0006] A method for correcting artifacts of three-dimensional images, comprising the steps of:

[0007] (1) obtaining a standard dark field image, an air field gain calibration image, an air field calibration image with different average brightness, a calibration image of different thickness water phantom and different thickness aluminum sheet, and obtaining a reconstructed image of different average brightness, different thickness water phantom and different thickness metal sheet according to the above;

[0008] (2) using the least square method to fit each image point in each of the above reconstructed images to obtain a gain curve;

[0009] (3) using the obtained gain curve to correct the gain of the three-dimensional image.

[0010] The reconstructed image of different average brightness, different thickness water phantom and different thickness metal sheet is specifically:

[0011] The air field calibration image with different average brightness, the calibration image of water phantom with different thickness and the calibration image of metal sheet with different thickness are subtracted from the standard dark field image respectively, and then the reconstructed images of the air field gain calibration image with different average brightness, the water phantom with different thickness and the metal sheet with different thickness are obtained after being divided by the air field gain calibration image and being processed by the inverse logarithm.

[0012] The gain curve is obtained by fitting each image point in the aforementioned reconstructed images by using the least square method, and the gain curve is specifically as follows:

[0013] The calculation formula of the gain calibration coefficient is as follows:

[0014] p ideal (i,j,k)=α(i,j)*p(i,j,k)+β(i,j)

[0015] Wherein, α(i,j) is the first term coefficient of the correction curve of the image point (i,j) obtained by solving, β(i,j) is the constant term coefficient of the correction curve of the image point (i,j) obtained by solving, i and j are the row and column coordinates of the image point respectively, k represents the kth image, p(i,j,k) represents the gray value of the kth image point (i,j), p ideal (i,j,k) represents the corrected gray value of the kth image point (i,j).

[0016] The cumulative sum s(i,j) of the residual square of the correction of the image point (i,j) is calculated as follows:

[0017]

[0018] The gain calibration coefficient is calculated according to the minimum value of the cumulative sum s(i,j) of the residual square of the correction of the image point (i,j) obtained by taking the partial derivative of the correction formula coefficients α(i,j) and β(i,j).

[0019] The gain curve is used to correct the gain of the three-dimensional image, including:

[0020] a) gain correction is performed on the air field gain calibration image according to the gain curve;

[0021] b) whether the uniformity of the image after gain correction meets the set requirement is judged;

[0022] If yes, the gain curve is used to correct the gain of the three-dimensional image;

[0023] If no, the gain curve is re-fitted for each image point in the aforementioned reconstructed images based on the deviation gray value of each image point of the air field gain calibration image before and after gain correction, and the step a) is repeated.

[0024] Whether the uniformity of the image after gain correction meets the set requirement is specifically:

[0025] The calculation formula of the uniformity of the image is:

[0026]

[0027] Wherein, C is the uniformity of the image, p0(i,j) is the gray value of the image point (i,j) in the image obtained after the air field gain calibration image is corrected by using the gain correction formula; N R and N C are the row and column numbers of the image points in the image, respectively;

[0028] If C is less than a set proportion threshold A, the uniformity of the gain-corrected image meets the set requirement.

[0029] A=0.2%.

[0030] The deviation gray value of each image point of the air field gain calibration image before and after the gain correction, is used to re-fit the gain curve of each image point in the aforementioned each reconstructed image, specifically as follows:

[0031] The gray value of each image point in each reconstructed image is added to the corresponding deviation gray value of each image point, and the gain curve is re-fitted with the updated image points.

[0032] The standard dark field image is obtained by continuously collecting a plurality of dark field images by the three-dimensional imaging device, and the mean value dark field image obtained by weighted averaging of the plurality of dark field images is used as the standard dark field image.

[0033] The air field gain calibration image with a set mean value brightness and the air field calibration image with different mean value brightnesses are obtained by:

[0034] The exposure voltage and the exposure current of the three-dimensional imaging device are set, so that the mean value brightness of the collected image is a set value and other different brightness air field images, wherein the set value is set according to the gain interval of the detector;

[0035] A plurality of images corresponding to the brightness are continuously collected by the three-dimensional imaging device, and the mean value air image obtained by weighted averaging of the plurality of air images is used as the air field calibration image corresponding to the brightness, wherein the mean value brightness of the image is the set value, and the air field calibration image corresponding to the set value is the air field gain calibration image.

[0036] The calibration images of the water phantom with different thicknesses and the metal sheet with different thicknesses are obtained by:

[0037] The exposure voltage and the exposure current of the three-dimensional imaging device are set, and a plurality of images of the water phantom with different thicknesses and the metal sheet with different thicknesses are collected by the three-dimensional imaging device, and the calibration images of the water phantom with different thicknesses and the metal sheet with different thicknesses are obtained by weighted averaging, respectively.

[0038] Beneficial effects: the present application solves the problem of ring artifact in 3D reconstructed image, improves the image quality, and ensures the reliability of 3D image for medical diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a flow chart of the present application;

[0040] Figure 2 is a dark field image collected by X-ray without exposure;

[0041] Figure 3 is an exposure image of X-ray in air at a specific kilovoltage and milliampere; in the figure, the specific kilovoltage and milliampere selected according to the characteristics of the X-ray tube and the detector are 100kv and 2mA respectively;

[0042] Figure 4 is an exposure image of X-ray through a water phantom of a specific thickness at a specific kilovoltage and milliampere; in the figure, the specific kilovoltage and milliampere selected are 100kv and 2mA respectively, and the thickness of the water phantom is 50mm;

[0043] Figure 5 is an exposure image of X-ray through an aluminum sheet of a specific thickness at a specific kilovoltage and milliampere; in the figure, the specific kilovoltage and milliampere selected are 100kv and 2mA respectively, and the thickness of the aluminum sheet is 1mm;

[0044] Figure 6 is a reconstructed air field slice image after flat field correction;

[0045] Figure 7 is a reconstructed air field slice image after the calibration method;

[0046] Figure 8 is a reconstructed phantom slice image without any correction;

[0047] Figure 9 is a reconstructed phantom slice image after flat field correction. DETAILED DESCRIPTION

[0048] The present application will be further illustrated below in combination with the drawings and specific embodiments.

[0049] The artifact correction method of the three-dimensional image of the present application is as shown in Figure 1 , which comprises the following steps:

[0050] (1) collecting dark field images, air field images of the same average brightness and air field images of different average brightness by C-arm scanning, and calculating to obtain standard dark field images, air field gain calibration images and air field calibration images of different average brightness respectively;

[0051] 500 dark field images are continuously collected by the C-arm machine, and the 500 dark field images are weighted and averaged to obtain a mean dark field image as a standard dark field image, as shown in Figure 2 ;

[0052] The exposure voltage of the C-arm machine is set to 100kV, and the amplitude of the exposure current is adjusted so that the mean brightness of the images collected by the C-arm machine is a set value (set according to the gain interval of the detector, and the present application is set to 13000), as shown in the example. Figure 3 500 images under this brightness are continuously collected by the C-arm machine, and the 500 air field images are weighted and averaged to obtain a mean air field image as an air field gain calibration image;

[0053] The exposure voltage of the C-arm machine is set to 100kV, and the amplitude of the exposure current is adjusted to obtain air field exposure images with mean brightness of 5000, 10000, 15000, 20000 and 30000 collected by the C-arm machine; 500 air field exposure images corresponding to the mean brightness are continuously collected by the C-arm machine, and air field calibration images with different mean brightness are obtained by weighted averaging;

[0054] (2) The exposure images of different thickness water films and aluminum sheets under the same kilovoltage and milliamperes are collected by the C-arm machine, and the corresponding calibration images are calculated;

[0055] The exposure voltage of the C-arm machine is set to 100kV, and the exposure current is the same as the current when the mean brightness is 30000, and 5mm, 10mm, 20mm, 30mm, 40mm and 50mm water models are respectively placed, and 500 images of different thickness water models are collected by the C-arm machine, as shown in the example. Figure 4 The water model calibration images of different thickness water models are obtained by weighted averaging;

[0056] The exposure voltage of the C-arm machine is set to 100kV, and the exposure current is the same as the current when the mean brightness is 30000, and 0.2mm, 0.5mm, 1mm, 2mm, 3mm and 5mm aluminum sheets are respectively placed, and 500 images of different thickness aluminum sheets are collected by the C-arm machine, as shown in the example. Figure 5 The aluminum sheet calibration images of different thickness aluminum sheets are obtained by weighted averaging;

[0057] In the present application, the aluminum sheet calibration image is used as an example; however, the present application is not limited to this, and the exposure images of different thickness metal sheets under the same kilovoltage and milliamperes can also be collected according to the above-mentioned manner, and the corresponding calibration images are calculated;

[0058] (3) according to the standard dark field image, air field gain calibration image, different mean luminance air field calibration image obtained in step (1) and the reconstruction images of different mean luminance, different thickness water phantom and different thickness aluminum sheet obtained respectively from each calibration image;

[0059] The air field calibration image of each mean luminance obtained in step (1) is subtracted from the standard dark field image, and after being divided by the air field gain calibration image obtained in step (1), the anti-log processing is carried out, thereby obtaining the reconstruction images of different mean luminance;

[0060] The water phantom calibration image of different thickness water phantom obtained in step (2) is subtracted from the standard dark field image obtained in step (1), and after being divided by the air field gain calibration image obtained in step (1), the anti-log processing is carried out, thereby obtaining the reconstruction images of different thickness water phantom;

[0061] The aluminum sheet calibration image of different thickness aluminum sheet obtained in step (2) is subtracted from the standard dark field image obtained in step (1), and after being divided by the air field gain calibration image obtained in step (1), the anti-log processing is carried out, thereby obtaining the reconstruction images of different thickness aluminum sheet;

[0062] In the present application, the image subtraction, division and anti-log processing refer to the corresponding operations on the corresponding pixel points of the images, for example, the image subtraction refers to the subtraction of the corresponding pixel points of the two images, and the division and the anti-log are also the same operations.

[0063] (4) according to the reconstruction images of different mean luminance, different thickness water phantom and different thickness aluminum sheet obtained in step (3), the gain curve fitting is carried out on each image point by using the least square method, thereby obtaining the gain calibration coefficient, and the specific process is as follows:

[0064] The calculation formula of the gain calibration coefficient is as follows:

[0065] p ideal (i,j,k)=α(i,j)*p(i,j,k)+β(i,j)

[0066] Wherein, alpha (i, j) is the first term coefficient of the correction curve of the image point (i, j) obtained by solving, beta (i, j) is the constant term coefficient of the correction curve of the image point (i, j) obtained by solving, i and j are the row and column coordinates of the image point respectively, k represents the kth image, p (i, j, k) represents the gray value of the kth image point (i, j), p ideal (i,j,k) represents the corrected gray value of the kth image point (i, j);

[0067] Therefore, the cumulative sum s (i, j) of the residual square of the correction of the image point (i, j) can be calculated:

[0068] Therefore, the cumulative sum s (i, j) of the residual square of the correction of the image point (i, j) can be calculated:

[0069] Taking partial derivative of the correction formula coefficients a(i,j) and β(i,j) respectively, we have:

[0070]

[0071] It can be seen that s(i,j) has the minimum value when the two partial derivatives are both 0, thus we have:

[0072]

[0073] wherein,

[0074] (5) Gain correction is performed on the air field gain calibration image obtained in step (1) by using the obtained gain correction coefficient, and the uniformity C of the gain corrected image is calculated, as shown in Figure 6 、 7 , Figure 6 It can be seen that there is obvious ring artifact in Figure 7 , Figure 6 , and the ring artifact in has been completely eliminated;

[0075] Figure 8 If C is less than 0.2%, gain correction is performed on the image obtained by three-dimensional scanning by using the gain correction coefficient obtained in step (4), and three-dimensional image reconstruction is performed to obtain the corrected three-dimensional image, as shown in 9 、 Figure 8 , Figure 9 It can be seen that there is obvious ring artifact in , and the ring artifact in

[0076] has been completely eliminated; otherwise, go to step (6);

[0077]

[0078] wherein, p0(i,j) is the gray value of the image point (i,j) in the image obtained by correcting the air field gain calibration image in step (1) by using the gain correction formula; N R and N C are the row and column numbers of the image points in the image;

[0079] (6) Calculate the gray value uniformity deviation of each image point of the air field gain calibration image before and after gain correction in step (5), and add the gray value of each image point in the reconstructed image of different mean brightness, different thickness water phantom and different thickness aluminum sheet obtained in steps (1) and (2) to the uniformity deviation, and then perform gain curve fitting on each image point by using the least square method to obtain gain calibration coefficients, and repeat step (5);

[0080] Specifically, the uniformity deviation image is calculated as follows:

[0081] p bias (i,j)=p0(i,j)-p air (i,j)

[0082] Wherein, p bias (i,j) is the gray value uniformity deviation of each image point (i,j) in the air field gain calibration image, p air (i,j) is the gray value of each image point (i,j) in the air field gain calibration image obtained in step (1).

[0083] The application solves the problem of unevenness of the image caused by the gain difference of different points of the flat panel detector, thereby causing the ring-shaped artifacts of the reconstructed three-dimensional image, effectively improves the image quality of the CT image, and ensures the reliability of the 3D image for medical diagnosis.

[0084] The preferred embodiments of the application are described in detail above, but the application is not limited to the specific details in the above-described embodiments, and various equivalent transformations (such as quantity, shape, position, etc.) can be performed on the technical solutions of the application within the technical concept range of the application, and these equivalent transformations all belong to the protection range of the application.

Claims

1. A method of artifact correction for a three-dimensional image, the method comprising: The method comprises the steps of: ​ (1) obtaining standard dark field images, air field gain calibration images, air field calibration images with different average luminance, calibration images of water models with different thickness and metal sheets with different thickness, and obtaining reconstructed images of water models with different thickness and metal sheets with different thickness according to the calibration images; (2) fitting each image point in the reconstructed images to obtain a gain curve by using a least square method; (3) performing gain correction on the three-dimensional image according to the obtained gain curve, comprising: a) performing gain correction on the air field gain calibration image according to the gain curve; b) judging whether the uniformity of the image after gain correction meets the set requirement; if yes, performing gain correction on the three-dimensional image according to the obtained gain curve; if no, re-fitting the gain curve of each image point in the reconstructed images based on the deviation gray value of each image point of the air field gain calibration images before and after gain correction, and repeating step a).

2. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: The reconstructed images of water models with different thickness and metal sheets with different thickness are obtained by: subtracting the air field calibration images with different average luminance, the calibration images of water models with different thickness and metal sheets with different thickness from the standard dark field images, and performing anti-log processing on the result of the division of the air field gain calibration image to obtain the reconstructed images of water models with different thickness and metal sheets with different thickness.

3. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: The gain curve is obtained by fitting each image point in the reconstructed images by using a least square method, and the calculation formula of the gain calibration coefficient is as follows: The calculation formula of the uniformity of the image is as follows: ； wherein, is a first order coefficient of the correction curve of the image point , is a constant coefficient of the correction curve of the image point , i, j are the row and column coordinates of the image point, respectively, and k represents the kth image. represents the gray value of the kth image point , represents the corrected gray value of the kth image point . The image point is thus calculated The cumulative sum of the corrected residual squares : ； The gain calibration coefficient is calculated based on the minimum value of the accumulated sum of the residual square of the correction and derivation of the image point correction .

4. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: if C is less than the set proportion threshold A, the uniformity of the image after gain correction meets the set requirement. The gain curve of each image point in the reconstructed images is re-fitted based on the deviation gray value of each image point of the air field gain calibration images before and after gain correction, and the gray value of each image point in each reconstructed image is added to the corresponding deviation gray value of each image point, and the gain curve is re-fitted based on the updated image points. ； Where C represents the uniformity of the image. , Image points in the image obtained after correcting the air field gain calibration image using the gain correction formula grayscale value; and These are the row and column numbers of the image points in the image; The standard dark field image is obtained by:

5. The artifact correction method of a three-dimensional image according to claim 4, characterized in that: A=0.2%。 6. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: continuously collecting a plurality of dark field images by a three-dimensional image device, and obtaining a mean dark field image by weighted averaging the plurality of dark field images, as the standard dark field image. The air field gain calibration image and the air field calibration image with different average luminance are obtained by:

7. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: setting the exposure voltage and the exposure current of the three-dimensional image, so that the average luminance of the collected images is a set value and other air field images with different luminance, wherein the set value is set according to the gain interval of the detector; 8. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: continuously collecting a plurality of images with corresponding luminance by the three-dimensional image, and obtaining a mean air image by weighted averaging the plurality of air images, as the air field calibration image with corresponding luminance, wherein the average luminance of the image is the air field calibration image corresponding to the set value, which is the air field gain calibration image. The calibration images of water models with different thickness and metal sheets with different thickness are obtained by: ​ 9. The artifact correction method of a three-dimensional image according to claim 1, characterized in that: ​ The exposure voltage and the exposure current of the three-dimensional image are set, images of water models with different thicknesses and metal sheets with different thicknesses are respectively collected through the three-dimensional image, and the calibration images of the water models with different thicknesses and the metal sheets with different thicknesses are respectively obtained through weighted average.

Images