X-ray image correction method

Abstract

The present invention provides an X-ray image correction method, which is characterized in that: after a distortion cover is set in front of an X-ray machine to shoot an X-ray image, the X-ray image is corrected by an anti-distortion correction method; the distortion cover includes a shooting area and an identification area. This method can overcome the problem of irregular changes in X-rays due to the influence of magnetic fields, and the X-ray is corrected by an anti-distortion algorithm.

Description

Technical Field

[0001] The present invention relates to a medical image correction method, in particular to an X-ray image correction method. Background Art

[0002] The X-ray images taken by medical C-arm X-ray machines have image distortion. Distorted images affect medical workers' judgment of the condition and cannot be used in three-dimensional reconstruction and surgical navigation. The present invention mainly focuses on the processing method for the impact of X-ray distortion on later images, and there is no effective solution in the prior art. In order to improve the accuracy of surgical navigation and reduce the risk of medical personnel in surgery, early X-ray images are particularly important; this technology processes the X-ray images of early patients, eliminates noise, and corrects the X-ray distortion caused by the surrounding magnetic field during the shooting process, so that when loaded into the surgical navigation system, the X-ray image is more accurate, and the system can accurately and effectively recognize the image, and will not be unable to recognize the image due to problems with the X-ray image, providing more accurate path planning and surgical plans for subsequent surgical navigation, greatly improving the accuracy and safety of the surgery. Summary of the invention

[0003] The present invention aims to overcome the above-mentioned defects, overcome the problem of irregular changes of X-rays due to the influence of magnetic fields, and correct the X-rays through an anti-distortion algorithm.

[0004] The present invention provides an X-ray image correction method, which is characterized in that: after a distortion cover is arranged in front of an X-ray machine to shoot an X-ray film, the X-ray film is corrected by an anti-distortion correction method;

[0005] The distortion mask includes a shooting area and an identification area.

[0006] Furthermore, the present invention provides an X-ray image correction method, which is also characterized in that the specific method of the anti-distortion correction method is as follows:

[0007] S1. Identify points on the X-ray image and obtain point coordinate information;

[0008] S2. Generate a first grid based on the points on the X-ray image;

[0009] S3. After matching the first grid obtained in S2 with the point coordinate information of the X-ray image obtained in S1, noise points in the data are eliminated to obtain the set data of the identification points;

[0010] S4. Generate a second grid based on the identification points obtained in S3;

[0011] S5. Matching the second grid obtained in S4 with the set data of the identification points obtained in S3 to generate a transformation matrix of the image;

[0012] S6. After combining the transformation matrix obtained in S5 with the original X-ray image, the X-ray image is adjusted according to the transformation matrix to obtain a corrected image.

[0013] Furthermore, the present invention provides an X-ray image correction method, which is also characterized in that the specific steps of S1 are:

[0014] S1-1. Perform color space conversion on the X-ray image to convert the image into a grayscale image;

[0015] S1-2. Perform noise reduction on the grayscale image obtained in S1-1;

[0016] S1-3. Perform spot detection on the image processed by S1-2, identify the points on the image, and obtain the image point coordinate information.

[0017] Furthermore, the present invention provides an X-ray image correction method, which is also characterized in that the specific steps of S2 are:

[0018] S2-1. Calculate the estimated spacing of the identification points on the distortion mask in the image point coordinates according to the X-ray image pixel size (generally speaking, the image pixel size is two fixed parameters of the X-ray machine, namely, length and width) and the X-ray shooting range size on the distortion mask (the shooting range size is used to estimate the X-ray image size);

[0019] S2-2. Calculate the Euclidean distance between each point and its nearest neighboring point according to the image point coordinate information obtained in S1;

[0020] The nearest point calculation method is to calculate the Euclidean distance between each point and all other points, and find the minimum Euclidean distance. The point corresponding to the Euclidean distance is the nearest point.

[0021] The distance between two points, for example, the distance between point A (X0, Y0) and point B (X1, Y1) is

[0022] S2-3. According to the preset rule I, the Euclidean distance of S2-2 is compared with the estimated distance of S2-1, and the point distances that do not meet the requirements are removed;

[0023] S2-4. Based on the distances of the points that meet the requirements after screening in S2-3, calculate their average value, which is the average spacing;

[0024] S2-5. Calculate the Euclidean distance between each point in the positive Y direction and its nearest neighboring point according to the image point coordinate information obtained in S1;

[0025] S2-6. According to the preset rule II, the Euclidean distance of S2-5 is compared with the average distance of S2-4, and the vector between each point and its adjacent point in the positive Y direction that best meets the requirements is calculated, and the average value thereof is calculated, which is the average rotation vector;

[0026] S2-7. Generate a first grid based on the average spacing of S2-4 and the average rotation vector of S2-6;

[0027] The length and width of the first grid are both: (the number of points in the longest column on the distortion mask - 1) * average spacing. After the grid is generated, the grid is matched with the points on the captured image by rotating the vector.

[0028] The calculation method of the estimated spacing is: estimated spacing = X-ray image pixel size * actual spacing of identification points on the distortion mask / estimated actual size of the X-ray image;

[0029] The estimated actual size is estimated based on the X-ray shooting range. The point spacing on the anti-distortion cover is known, and the size is estimated based on the point spacing (for example: the longest column in the shooting range has 11 points, and the point spacing is 20MM, then the estimated size is 200MM).

[0030] The preset rule I means that the Euclidean distance and the estimated distance must comply with the following rules:

[0031] Euclidean distance < estimated distance * rated value I-1;

[0032] and / or

[0033] Euclidean distance>estimated distance*rated value I-2;

[0034] The preset rule II for selecting the neighboring points that best meet the requirements means that the Euclidean distance and the average spacing must meet the following rules:

[0035] Euclidean distance ≧ average distance * rated value II-1;

[0036] and / or

[0037] Euclidean distance ≦ average distance * rated value II-2.

[0038] Furthermore, the present invention provides an X-ray image correction method, which is also characterized in that the specific steps of S3 are:

[0039] S3-1. Calculate the Euclidean distance from each point in S1 to the nearest point on the first grid;

[0040] S3-2. According to the preset rule III, the average distance of S2-4 is compared with the Euclidean distance of S3-1 to determine whether the point is a noise point;

[0041] S3-3. After filtering out non-noise points, the set data of the identification points is formed;

[0042] The preset rule III means that the Euclidean distance and the average distance must comply with the following rules:

[0043] Euclidean distance > Rating III * Average distance.

[0044] Furthermore, the X-ray image correction method provided by the present invention is characterized in that the specific step of S4 is: generating a second grid according to the identification points on the anti-distortion cover;

[0045] The grid size of the second grid is: the number of points in the longest column on the distortion mask*the estimated spacing between the points.

[0046] The number of points in the longest column on the distortion mask refers to the points on the diameter of the entire distortion mask.

[0047] Furthermore, the present invention provides an X-ray image correction method, which is also characterized in that the specific steps of S5 are:

[0048] S5-1 calculates and accumulates the X coordinate value and the Y coordinate value of each point on the identification point set data obtained in S3, calculates the average value of the X and Y coordinates, and obtains the coordinates of the center point of the identification point set data;

[0049]

[0050] S5-2 matches the second grid with the identification point set data of S3, so that the center point of the second grid corresponds to the center point of the identification point set data obtained in S5-1, and sets the rotation gradient of the second grid, iteratively transforms the second grid, finds the nearest point on the second grid corresponding to each point on the identification point set data, and minimizes the sum of the distances between each point and the corresponding nearest point, thereby obtaining the set data of the corresponding points;

[0051] S5-3 uses the corresponding point set data obtained in S5-2 to establish a 4th-order polynomial relationship for each point, and generates a ploys matrix (a 15*N matrix, where N is the number of points in the set data);

[0052] Fourth-order polynomial: X 4 +Y 4 +XY 3 +X 2 Y 2 +X 3 Y+X 3 +Y 3 +XY 2 +X 2 Y+X 2 +Y 2+XY+X+Y+1 (where (X, Y) represents the coordinate information of the identification point);

[0053] S5-4 performs SVD singular value decomposition on the ploys matrix obtained in S5-3 to obtain three matrices w, u and vt (w: calculated singular value, u: calculated left singular value, vt: transposed matrix of right singular value vector);

[0054] S5-5 uses the identification point set data in S3 to construct a pts matrix (a 2*N matrix, where 2 represents the X and Y coordinate information of the point);

[0055] S5-6 combines the three matrices w, u, vt and the pts matrix to calculate the tf matrix (Transformation matrix) of the image; if but The s matrix is ​​an n*n matrix;

[0056] Calculation method: tf = vt t *s*u t *pts;

[0057] The transposed matrix of vt is vt t ;

[0058] The transposed matrix of u is u t ;

[0059] Transposed matrix: The new matrix obtained by swapping the rows and columns of a matrix is ​​called the transposed matrix.

[0060] Furthermore, the present invention provides an X-ray image correction method, which is also characterized in that the specific steps of S6 are:

[0061] S6-1. Perform color space conversion on the input original X-ray image to convert the image into a grayscale image;

[0062] S6-2. Generate a 4th-order fitting polynomial corresponding to each pixel in the image;

[0063] A 15*1 matrix is ​​generated for each recognition point using the polynomial, and then directly multiplied with the transformation matrix to calculate the corrected pixel value;

[0064] S6-3. Calculate the pixel corresponding to the current pixel through the transformation matrix and polynomial, and redraw the image;

[0065] S6-4. Output the corrected image.

[0066] Furthermore, the X-ray image correction method provided by the present invention is characterized in that it also includes a correction evaluation step S7;

[0067] The specific steps of the correction evaluation process S7 are as follows:

[0068] S7-1. Repeat the steps S1 and S2 on the corrected image to obtain the points on the image and the grid of the image;

[0069] S7-2. Match the points and the grid, calculate the Euclidean distance from each point to the nearest point in the grid, and calculate the average value to obtain the average spacing;

[0070] S7-3. Evaluate the distortion correction effect of the corrected image based on the average spacing. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 , an anti-distortion cover applied to the X-ray image correction method as mentioned in the embodiment;

[0072] Figure 2 , a method for using the anti-distortion cover as mentioned in the embodiment;

[0073] Figure 3 , a reference diagram of the usage status of the anti-distortion cover as mentioned in the embodiment. DETAILED DESCRIPTION

[0074] In order to achieve the X-ray correction effect mentioned in this embodiment, it is necessary to cooperate with the following Figure 1 The anti-distortion cover shown is used.

[0075] The anti-distortion hood is generally available in two sizes: 9 inches and 12 inches, depending on the usage environment and conditions.

[0076] The anti-distortion cover is provided with N evenly distributed identification points, and the distance between the points is 20 mm. The distance between the points can be adjusted arbitrarily according to the actual use environment.

[0077] During use, the anti-distortion cover is placed between the object to be filmed and the optical machine. Generally speaking, an imaging plate is provided at one end of the C-arm, and the anti-distortion cover is installed on the imaging plate, and an X-ray light source is provided at the other end of the C-arm. Different specifications of the C-arm do not affect the installation and correction effect of the anti-distortion cover.

[0078] The specific anti-distortion method is as follows:

[0079] 1. Obtain X-ray images taken before surgery;

[0080] 2. Identify points on the X-ray image and obtain coordinate information of the points; (the coordinates can be obtained through existing technology)

[0081] ①Convert the color space of the input X-ray image and convert the image into a grayscale image;

[0082] ②Perform median filtering on the grayscale image obtained in ① (deepen the color on the image) to reduce the impact of noise;

[0083] ③ Perform SimpleBlobDetector on the image processed by ② to identify the points on the image (also identify normal points and noise points) and obtain the image point coordinate information.

[0084] 3. Generate a first grid according to the points on the X-ray image;

[0085] ① According to the pixel size of the X-ray image and the shooting range of the X-ray on the distortion mask, the estimated spacing of the identification points on the distortion mask in the image point coordinates is calculated;

[0086] Estimated spacing calculation method: pixel size * actual spacing of identification points on the distortion mask / actual size estimated by the X-ray image = estimated spacing

[0087] ② Input the image point coordinate information obtained in step 2, calculate the Euclidean distance between each point in the image point coordinate information and its nearest neighbor point (using the nearest neighbor algorithm) by traversing the points, combine the obtained Euclidean distance with the estimated spacing obtained in the previous step, judge and remove the point distances that do not meet the requirements (the judgment criteria for the point distances that do not meet the requirements: Euclidean distance < estimated spacing * 0.8 or Euclidean distance > estimated spacing * 1.2), and calculate the average value of the point distances that meet the requirements, that is, the average spacing of the image point coordinate information;

[0088] ③ Input the image point coordinate information obtained in step 2, and calculate the vector of each point in the image point coordinate information and its neighboring point in the positive Y direction that best meets the requirements by traversing the points (the most suitable neighboring point judgment standard: first calculate the Euclidean distance between the two points, and then combine the average spacing obtained in ②. If the Euclidean distance ≧ average spacing * 0.8 and the Euclidean distance ≦ average spacing * 1.2, it is judged as a suitable neighboring point). Calculate the average value of the vector, that is, the average rotation vector of the image point coordinate information compared to the positive Y direction of the image;

[0089] ④Use the average spacing and average rotation vector obtained in ② and ③ to generate the first grid that matches this series of points

[0090] 4. Match the first grid with the X-ray image point coordinate information obtained in step 2 to obtain corresponding data, eliminate noise points in the data according to existing methods, and obtain the set data of the identification points;

[0091] ① By traversing the points, calculate the Euclidean distance from each point to the nearest point on the first grid

[0092] ② Use the Euclidean distance to determine whether the point is a noise point (noise point judgment standard: Euclidean distance>0.2*average spacing; the average spacing is obtained in step 3②). If it is a noise point, delete the point;

[0093] ③ Output the points after noise is removed, that is, the collection data of the identified points on the image;

[0094] 5. Generate a second grid based on the identification points on the anti-distortion mask, where the grid size is: the number of points in the longest column on the distortion mask multiplied by the estimated spacing between the points (the estimated spacing is obtained in step 3①);

[0095] 6. Match the second grid with the set data of the identification points obtained in step 4 to generate a transformation matrix of the image; (anti-distortion method)

[0096] ① Calculate the center point of the identification point;

[0097] Calculation method: Accumulate the set data of the identified points obtained in step 4, calculate the average value, and get the center point

[0098] That is, the X coordinate value and the Y coordinate value of each point on the identification point set data are calculated and added respectively, and the average value of the X and Y coordinates is calculated to obtain the coordinates of the center point of the identification point set data;

[0099]

[0100] ② Calculate the Euclidean distance from the identified point to the nearest point on the second grid;

[0101] That is, match the second grid with the S3 identification point set data, make the center point of the second grid correspond to the center point of the identification point set data obtained in ①, set the rotation gradient of the second grid, iteratively transform the second grid, find the nearest point on the second grid corresponding to each point on the identification point set data, and minimize the sum of the distances between each point and the corresponding nearest point, and obtain the set data of corresponding points;

[0102] ③ Combine ① and ② to align the identification points on the second grid and the set data obtained in step 4;

[0103] ④ Establish a 4th-order fitting polynomial relationship between the second grid and the identification points

[0104] Use the corresponding point set data obtained in S5-2 to establish a 4th-order polynomial relationship for each point and generate a ploys matrix (a 15*N matrix, where N is the number of points in the set data);

[0105] 4th order fitting polynomial relationship:

[0106] X 4 +Y 4 +XY 3+X 2 Y 2 +X 3 Y+X 3 +Y 3 +XY 2 +X 2 Y+X 2 +Y 2 +XY+X+Y+1 (where (X, Y) represents the coordinate information of the identification point)

[0107] Using this polynomial, a 15*1 matrix is ​​generated for each recognition point, and finally a 15*N matrix is ​​obtained.

[0108] Polynomial and fitting,The correction process is to fit the generated polynomial to the standard points on the second grid,,which is equivalent to obtaining a fitting function.

[0109] Perform SVD singular value decomposition on the ploys matrix to obtain three matrices w, u and vt (w: calculated singular value, u: calculated left singular value, vt: transposed matrix of the right singular value vector);

[0110] ⑤ Decompose the polynomial through SVD singular value decomposition and calculate the transformation matrix of the image;

[0111] Using the recognition point set data, construct a pts matrix (a 2*N matrix, where 2 represents the X and Y coordinate information of the point);

[0112] S5-6 combines the three matrices w, u, vt and the pts matrix to calculate the tf matrix (Transformation matrix) of the image; if but The s matrix is ​​an n*n matrix;

[0113] Calculation method: tf = vt t *s*u t *pts;

[0114] The transposed matrix of vt is vt;

[0115] The transposed matrix of u is u t ;

[0116] Transposed matrix: The new matrix obtained by swapping the rows and columns of a matrix is ​​called the transposed matrix.

[0117] 7. Combining the transformation matrix with the original X-ray image, the X-ray image is adjusted according to the transformation matrix to obtain a corrected image.

[0118] ①Convert the color space of the input original X-ray image and convert the image into a grayscale image;

[0119] ② Generate a 4th-order fitting polynomial corresponding to each pixel in the image;

[0120] 4th order fitting polynomial relationship:

[0121] X 4 +Y 4 +XY 3 +X 2 Y 2 +X 3 Y+X 3 +Y 3 +XY 2 +X 2 Y+X 2 +Y 2 +XY+X+Y

[0122] +1 (where (XY) represents the coordinate information of the identification point)

[0123] This polynomial is used to generate a 15*1 matrix for each identification point, which is then directly multiplied with the transformation matrix to calculate the corrected pixel value.

[0124] Polynomial and fitting, the correction process is to fit the generated polynomial to the standard points on the second grid, which is equivalent to obtaining a fitting function

[0125] ③ Calculate the pixel corresponding to the current pixel through the transformation matrix and polynomial, and redraw the image

[0126] Calculation method: It is a matrix multiplication, polynomial matrix * transformation matrix, which can be understood as fitting each pixel point on the image to the second grid, eliminating the distortion of the image. The corrected image is output.

[0127] 8. Evaluate through the corrected image; (verification)

[0128] ① Repeat steps 2 and 3 on the corrected image to obtain the points on the image and the grid of the image.

[0129] ②Match the points and the grid, calculate the Euclidean distance from each point to the nearest point in the grid, and calculate the average value to get the average spacing;

[0130] ③Use the obtained average spacing to evaluate the distortion correction effect of the corrected image. The smaller the value, the better the correction result.

Claims

1. A method for correcting an X-ray image, characterized in that: After the X-ray image is taken by setting a distortion hood in front of the X-ray machine, the X-ray image is corrected by the anti-distortion correction method; The distortion mask includes a shooting area and an identification area; The specific method of the anti-distortion correction method is as follows: S1. Identify points on the X-ray image and obtain point coordinate information; S2. Generate a first grid based on the points on the X-ray image; S3. After matching the first grid obtained in S2 with the point coordinate information of the X-ray image obtained in S1, noise points in the data are eliminated to obtain the set data of the identification points; S4. Generate a second grid based on the identification points obtained in S3; S5. Matching the second grid obtained in S4 with the set data of the identification points obtained in S3 to generate a transformation matrix of the image; S6. After combining the transformation matrix obtained in S5 with the original X-ray image, the X-ray image is adjusted according to the transformation matrix to obtain a corrected image.

2. The X-ray image correction method according to claim 1, characterized in that: The specific steps of S1 are: S1-1. Perform color space conversion on the X-ray image to convert the image into a grayscale image; S1-2. Perform noise reduction on the grayscale image obtained in S1-1; S1-3. Perform spot detection on the image processed by S1-2, identify the points on the image, and obtain the image point coordinate information.

3. The X-ray image correction method according to claim 1, characterized in that: The specific steps of S2 are: S2-1. Calculate the estimated spacing of the identification points on the distortion mask in the image point coordinates according to the pixel size of the X-ray image and the shooting range of the X-ray on the distortion mask; S2-2. Calculate the Euclidean distance between each point and its nearest neighboring point according to the image point coordinate information obtained in S1; S2-3. According to the preset rule I, the Euclidean distance of S2-2 is compared with the estimated distance of S2-1, and the point distances that do not meet the requirements are removed; S2-4. Based on the distances of the points that meet the requirements after screening in S2-3, calculate their average value, that is, the average spacing; S2-5. According to the image point coordinate information obtained in S1, calculate the Euclidean distance between each point in its positive Y direction and its nearest neighboring point; S2-6. According to the preset rule II, the Euclidean distance of S2-5 is compared with the average distance of S2-4, and the vector between each point and its adjacent point in the positive Y direction that best meets the requirements is calculated, and the average value thereof is calculated, which is the average rotation vector; S2-7. Generate a first grid based on the average spacing of S2-4 and the average rotation vector of S2-6; The calculation method of the estimated spacing is: estimated spacing = X-ray image pixel size * actual spacing of identification points on the distortion mask / estimated actual size of the X-ray image; The preset rule I means that the Euclidean distance and the estimated distance must comply with the following rules: Euclidean distance < estimated distance * rated value I-1; and / or Euclidean distance>estimated distance*rated value I-2; The preset rule II for selecting the neighboring points that best meet the requirements means that the Euclidean distance and the average spacing must meet the following rules: Euclidean distance ≧ average distance * rated value II-1; and / or Euclidean distance ≦ average distance * rated value II-2.

4. The X-ray image correction method according to claim 3, characterized in that: The specific steps of S3 are: S3-1. Calculate the Euclidean distance from each point in S1 to the nearest point on the first grid; S3-2. According to the preset rule III, the average distance of S2-4 is compared with the Euclidean distance of S3-1 to determine whether the point is a noise point; S3-3. After filtering out non-noise points, the set data of identification points is formed; The preset rule III means that the Euclidean distance and the average distance must comply with the following rules: Euclidean distance > Rating III * Average distance.

5. The X-ray image correction method according to claim 1, characterized in that: The specific steps of S4 are: generating a second grid according to the identification points on the anti-distortion mask; The grid size of the second grid is: the number of points in the longest column on the distortion mask*the estimated spacing between the points.

6. The X-ray image correction method according to claim 1, characterized in that: The specific steps of S5 are: S5-1. Calculate the X coordinate value and Y coordinate value of each point in the identification point set data obtained in S3, add them up respectively, calculate the average value of the X and Y coordinates, and obtain the coordinates of the center point of the identification point set data; S5-2. Match the second grid with the identification point set data of S3, so that the center point of the second grid corresponds to the center point of the identification point set data obtained in S5-1, and set the rotation gradient of the second grid, iteratively transform the second grid, find the nearest point on the second grid corresponding to each point on the identification point set data, and minimize the sum of the distances between each point and the corresponding nearest point, and obtain the set data of corresponding points; S5-3. Use the corresponding point set data obtained in S5-2 to establish a 4th-order polynomial relationship for each point and generate a ploys matrix; The 4th order polynomial is: X 4 +Y 4 +XY 3 +X 2 Y 2 +X 3 Y+X 3 +Y 3 +XY 2 +X 2 Y+X 2 +Y 2 +XY+X+Y+1 X, Y are the coordinate information of the identification point; S5-4 performs SVD singular value decomposition on the ploys matrix obtained in S5-3 to obtain three matrices w, u and vt; w: calculated singular value, u: calculated left singular value, vt: transposed matrix of right singular value vector; S5-5 constructs a pts matrix using the identification point set data in S3; S5-6 combines the three matrices w, u and vt with the pts matrix to calculate the tf matrix of the image; like but The s matrix is ​​an n*n matrix; Calculation method: tf = vt t *s*u t *pts; The transposed matrix of vt is vt t ; The transposed matrix of u is u t .

7. The X-ray image correction method as claimed in claim 1, characterized in that: The specific steps of S6 are: S6-1. Perform color space conversion on the input original X-ray image to convert the image into a grayscale image; S6-2. Generate a 4th-order fitting polynomial corresponding to each pixel in the image; A 15*1 matrix is ​​generated for each recognition point using the polynomial, and then directly multiplied with the transformation matrix to calculate the corrected pixel value; S6-3. Calculate the pixel corresponding to the current pixel point through the transformation matrix and polynomial, and redraw the image; S6-4. Output the corrected image.

8. The X-ray image correction method according to claim 1, characterized in that: Also includes a correction evaluation step S7; The specific steps of the correction evaluation process S7 are as follows: S7-1. Repeat the steps S1 and S2 on the corrected image to obtain the points on the image and the grid of the image; S7-2. Match the points and the grid, calculate the Euclidean distance from each point to the nearest point in the grid, and calculate the average value to obtain the average spacing; S7-3. Evaluate the distortion correction effect of the corrected image based on the average spacing.

Images