A knee joint X-ray image quality control method and device

Abstract

The present invention provides a knee joint X-ray image quality control method and device. The method comprises: obtaining the anteroposterior and lateral X-ray images of the knee joint; detecting key points located in the anteroposterior and lateral positions of each bone of the knee joint for positioning and eliminating the influence of occlusion; performing image segmentation based on the key points, scoring each quality control index by calculating the relevant parameters characterizing the quality control index, and obtaining a quantitative score for the knee joint X-ray image after summarization. The present invention can conveniently realize the positioning of each bone on the knee joint by marking a certain number of key points in the anteroposterior and lateral positions of each bone of the knee joint, thereby simplifying the calculation of quality control indicators; since the segmentation, classification and other operations involved in the calculation of quality control indicators are based on key points, most of the background area can be removed, thereby reducing the memory usage and data calculation amount, and improving the running speed; by setting a small number of redundant key points, the problem of difficulty in calculating quality control parameters due to occlusion can be effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical imaging, and in particular relates to a knee joint X-ray image quality control method and device. Background Art

[0002] Medical imaging technology (X-ray, CT, MRI, etc.) is the foundation of modern medical diagnostic technology. As a basic auxiliary diagnostic tool, it can help doctors quickly locate the condition and make accurate and professional judgments quickly. Digital Radiography (DR) is one of the most commonly used auxiliary means by doctors due to its low price, low radiation to patients, and fast imaging, especially for third-tier cities or some relatively backward cities.

[0003] Due to the backwardness of local equipment, the small number of radiologists and the uneven professional level, the images taken are often not so perfect, and the quality of the images directly affects the final doctor's judgment and the determination of the subsequent treatment plan. Traditionally, the determination of image quality is completed by senior radiologists with the naked eye. This quality control method usually has the following disadvantages: there are few doctors, and even fewer senior doctors; doctors have limited time, and the efficiency of naked eye review is too low; it is difficult to achieve complete uniformity of standards among different doctors.

[0004] The knee joint is a very important joint in the human body. It has a complex structure and is relatively flexible. The actual image shooting process requires relatively high technical skills from doctors. The invention patent with application number 202010431833.9 and the name of "Quality Control Method for Medical Images and Computer-readable Storage Medium" proposes an image quality control method based on instance segmentation. This method has the following disadvantages: it is necessary to annotate instance-level segmentation labels for each bone block (tibia, fibula, patella), which requires a huge amount of annotation and requires a lot of data; it is necessary to convert the segmentation results into corresponding linear measurement results, such as how to convert the segmentation results of the tibia and femur into corresponding angle relationships, which is a relatively complex and error-prone process; for severely occluded bones, the performance of the model segmentation directly affects the final quality of the image, such as the overlapping part of the fibula head and the tibia is often blurred and difficult to locate. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a knee joint X-ray image quality control method and device.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions.

[0007] In a first aspect, the present invention provides a knee joint X-ray image quality control method, comprising the following steps:

[0008] Obtain anteroposterior and lateral x-rays of the knee;

[0009] Mark the key points of each bone in the knee joint in the anteroposterior and lateral positions for positioning and avoiding occlusion;

[0010] Image segmentation is performed based on the key points, and each quality control indicator is scored by calculating relevant parameters characterizing the quality control indicator, and the quantitative score of the knee joint X-ray image is obtained after summarization.

[0011] Furthermore, the method also includes the following preprocessing steps: data collection, data screening and desensitization, labeling and review, and data conversion.

[0012] Furthermore, the key points of the anteroposterior position include: the key point of the mid-section of the femur, the key point of the femoral condyle, the key point of the tibial tuberosity, the key point of the mid-section of the tibia, the key point of the fibula neck, the key point of the mid-section of the fibula, the key point of the fibula head, the key point of the overlap between the tibia and the fibula head, and the key point of the patella; the key points of the lateral position include: the key point of the mid-section of the femur, the key point of the femoral condyle, the key point of the tibial tuberosity, the key point of the mid-section of the tibia, the key point of the fibula neck, the key point of the mid-section of the fibula, the key point of the fibula head, the key point of the overlap between the tibia and the fibula head, the patella, the key point of the posterior edge of the medial and lateral femoral condyles, the key point of the distal condyle of the medial and lateral femoral condyles, and the key point of the patellar surface of the medial and lateral femoral condyles.

[0013] Furthermore, the scoring method for whether the surrounding soft tissue is intact in the anteroposterior position includes:

[0014] Detect the edge of the input image based on the edge detection operator;

[0015] Determine whether the soft tissue is completely wrapped based on the gradient changes at the edge;

[0016] If yes, the score is S1; otherwise, the score is 0.

[0017] Furthermore, the scoring method for the degree of overlap of the medial and lateral femoral condyles in the lateral view includes:

[0018] Using the lateral patella key point I, the posterior edge key point J of the medial and lateral femoral condyle, and the distal condyle key point K of the medial and lateral femoral condyle, the bilateral shadow area of ​​the medial and lateral femoral condyle is segmented from the input lateral image;

[0019] Normalize the segmented region image to a fixed size;

[0020] Calculate the ratio of the double-sided shadow area in the segmented region to the segmented region area;

[0021] If the ratio is greater than the set threshold, the score is S2; otherwise, the score is 0.

[0022] Furthermore, the scoring method for the degree of overlap between the patella and femur in the lateral view includes:

[0023] Take the key point K of the distal condyle of the medial and lateral femoral condyle as the center and draw a square with the distance between 4 times K and the key point L of the patellar surface of the medial and lateral femoral condyle as the side length;

[0024] The square area is cut out from the input lateral image to obtain a distribution position structure diagram of the patella and femur;

[0025] The confidence of the overlap between the patella and femur is obtained using a classification network;

[0026] If the confidence level is less than the set threshold, the score is S3; otherwise, the score is 0.

[0027] Furthermore, the scoring method for the overlap between the fibular head and the tibia includes:

[0028] The direction of the fibula is determined by the straight line EF passing through the key point E of the fibula neck and the key point F of the mid-section of the fibula;

[0029] Calculate the distances L1, L2, L3 from the fibular head key points G1, G2 and the overlapping key point H of the tibia and fibular head to the straight line EF respectively;

[0030] Calculate the fibular head size L f : If there is no occlusion, L f =L1+L2; If one of G1 and G2 is blocked, calculate using symmetry, L f =2*L1 or L f =2*L2;

[0031] Calculate the length Lc of the overlapping part of the fibula head and the tibia: if G1 and H are on both sides of the straight line EF, Lc = L1 + L3; if G1 and H are on the same side of the straight line EF, Lc = L1-L3; if G1 is blocked, Lc = L2-L3;

[0032] Calculate the overlap ratio of the fibular head to the tibia Lc / L f ;

[0033] If the overlap ratio is within the set threshold range, the score is S4; otherwise, the score is 0.

[0034] Furthermore, the scoring method for the lateral lower limb knee flexion angle includes:

[0035] Draw a straight line AB through the key point A of the mid-femoral section and the key point B of the femoral condyle in the lateral position;

[0036] Draw a straight line CD through the key point C of the lateral tibial tuberosity and the key point D of the mid-tibia, intersecting the straight line AB at point O;

[0037] Calculate the size of the angle BOC;

[0038] If the angle BOC is within the set threshold range, the score is S5; otherwise, the score is 0.

[0039] Furthermore, the method also includes a quantitative score correction step: detecting whether there are multiple knee joints, artifacts, foreign bodies, postoperative changes, and other body positions in the input image, and if so, deducting a certain score from the knee joint quantitative score; the artifacts include pants artifacts and knee pad artifacts.

[0040] In a second aspect, the present invention provides a knee joint X-ray image quality control device, comprising:

[0041] An image acquisition module, used to acquire anteroposterior and lateral X-ray images of the knee joint;

[0042] Key point detection module, used to detect the key points located at the front and side positions of each bone of the knee joint for positioning and eliminating occlusion effects;

[0043] The indicator scoring module is used to perform image segmentation based on the key points, score each quality control indicator by calculating the relevant parameters characterizing the quality control indicator, and obtain a quantitative score for the knee joint X-ray image after summarizing.

[0044] Compared with the prior art, the present invention has the following beneficial effects.

[0045] The present invention can conveniently realize the positioning of each bone on the knee joint by setting a certain number of key points in the anteroposterior and lateral positions of each bone of the knee joint, thereby simplifying the calculation of quality control indicators; since the segmentation, classification and other operations involved in the calculation of quality control indicators are performed based on key points, most of the background area can be cropped out, thereby reducing memory usage and data calculation amount, and improving the running speed; by setting a small number of redundant key points, the problem of difficulty in calculating quality control parameters due to occlusion can be effectively solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flowchart of a knee joint X-ray image quality control method according to an embodiment of the present invention.

[0047] Figure 2 Schematic diagram of the distribution of key points of the knee joint in anteroposterior position.

[0048] Figure 3 Schematic diagram of the distribution of key points of the lateral knee joint.

[0049] Figure 4 Schematic diagram of the overlap calculation between the fibular head and tibia.

[0050] Figure 5 Schematic diagram for calculating the knee flexion angle of the lower limbs.

[0051] Figure 6 Schematic diagram of double-sided shadow segmentation.

[0052] Figure 7 The present invention is a block diagram of a knee joint X-ray image quality control device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention clearer and more understandable, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0054] Figure 1 This is a flow chart of a knee joint X-ray image quality control method according to an embodiment of the present invention, comprising the following steps:

[0055] Step 101, obtaining anteroposterior and lateral X-ray images of the knee joint;

[0056] Step 102, detecting the key points located at the frontal and lateral positions of each bone of the knee joint for positioning and eliminating the influence of occlusion;

[0057] Step 103, performing image segmentation based on the key points, scoring each quality control indicator by calculating relevant parameters characterizing the quality control indicator, and obtaining a quantitative score for the knee joint X-ray image after summarization.

[0058] In this embodiment, step 101 is mainly used to obtain X-ray images of the knee joint. Medical images such as X-ray, CT, and MRI are the basis of modern medical diagnostic technology. As basic auxiliary diagnostic tools, they can help doctors quickly locate the condition, thereby quickly making accurate and professional judgments. X-ray imaging is one of the most commonly used auxiliary means by doctors due to its low price, low radiation to patients, and rapid imaging. Due to the need for quality control parameter calculation, the knee joint images input in this embodiment include images in both the frontal and lateral directions (equivalent to the commonly said front view and left or right view).

[0059] In this embodiment, step 102 is mainly used to mark key points. In order to simplify the complexity of knee joint image quality control parameter calculation, this embodiment sets a certain number of key points in the frontal and lateral positions of each bone of the knee joint, such as Figure 2 , 3A, B, C, D, etc. in the figure. Setting key points is mainly used for locating each bone. For example, if two key points are determined on a bone, the direction of the bone can be represented by a straight line connecting the two key points. Setting some redundant key points with special positions (such as symmetry about a bone) and using the redundant key points for calculation when one of the key points is blocked can solve the problem of difficulty in calculating quality control parameters due to bone occlusion (overlap). The following embodiments will give the position distribution of some specific key points.

[0060] In this embodiment, step 103 is mainly used to quantitatively score the knee joint X-ray image. This embodiment calculates the parameters related to each quality control index based on the marked key points, and then scores each quality control index based on the calculated parameters, and finally summarizes them to obtain the quantitative score of the knee joint X-ray image. In order to reduce the burden of image processing, reduce memory usage, and improve the running speed, image segmentation is performed before parameter calculation, and then calculations are performed based on the segmented image. This embodiment performs image segmentation based on key points, and most of the background area can be cropped out, further reducing memory usage and data calculation. With key points, the calculation of related parameters becomes as simple as solving elementary geometry and algebra exercises, involving calculations including the distance from a point to a straight line, the angle between two straight lines, the ratio, etc. The scoring of each quality control index parameter is assigned different weights, i.e., scores, according to its importance, such as 10 points, 15 points, etc. The quality control index parameters that need to be scored mainly include whether the surrounding soft tissues in the anteroposterior position are intact, the degree of overlap between the femur and the medial and lateral condyles in the lateral position, the degree of overlap between the patella and the femur in the lateral position, the lateral lower limb knee flexion angle, the degree of overlap between the fibula head and the tibia, whether the anteroposterior patella is located in the middle of the femur, etc. The following embodiments will provide specific calculation and scoring methods for several quality control index parameters.

[0061] As an optional embodiment, the method also includes the following preprocessing steps: data collection, data screening and desensitization, labeling and review, and data conversion.

[0062] This embodiment provides a technical solution for preprocessing. Before inputting knee joint X-ray images, some necessary preprocessing is required. It mainly includes: (1) Data collection: Collect knee joint X-ray image fracture data that conforms to the DICOM (Digital Imaging and Communications in Medicine) specification from major hospitals; (2) Data screening and desensitization: Preliminarily screen out some unqualified data. At the same time, in order to protect the privacy of users, it is necessary to perform necessary desensitization work on each DICOM data source to ensure that key information is not leaked; (3) Doctor labeling and review: This is a very critical step in the preprocessing process. Generally, more than two experienced experts are required to label the fracture contour, fracture site and key point information of bone blocks of an X-ray image data at the same time. Then, the labeling results with consistent labeling results are directly entered into the database. For images with inconsistent labeling (two experts have different opinions), it is necessary to invite a third party more authoritative expert to review, and then all the labeling results that have been reviewed and have no objections are entered into the database for subsequent use; (4) Data conversion: Perform preliminary preprocessing work on the DICOM data that meets the requirements and has been labeled to ensure that the subsequent model can be read and used more conveniently. It mainly includes the setting of window width and window level and the conversion of DICOM format. The setting of window width and window level can directly use the default values ​​preset in DICOM or use some adaptive algorithms (such as clustering with a large number of empirical results) to find the most suitable window width and window level. The format conversion mainly converts DICOM format data into an easier-to-use image format (jpg, png, etc.) based on window width and window level, and also organizes the image annotation information into a unified format for subsequent use.

[0063] As an optional embodiment, the key points of the frontal position include: the key point of the middle section of the femur, the key point of the femoral condyle, the key point of the tibial tuberosity, the key point of the middle section of the tibia, the key point of the fibula neck, the key point of the middle section of the fibula, the key point of the fibula head, the key point of the overlap between the tibia and the fibula head, and the key point of the patella; the key points of the lateral position include: the key point of the middle section of the femur, the key point of the femoral condyle, the key point of the tibial tuberosity, the key point of the middle section of the tibia, the key point of the fibula neck, the key point of the middle section of the fibula, the key point of the fibula head, the key point of the overlap between the tibia and the fibula head, the key point of the patella, the key point of the posterior edge of the medial and lateral femoral condyles, the key point of the distal condyle of the medial and lateral femoral condyles, and the key point of the patellar surface of the medial and lateral femoral condyles.

[0064] This embodiment provides the position distribution of some specific key points. Since the calculation of quality control index parameters involves knee joint images in both anteroposterior and lateral directions, the key points of this embodiment include anteroposterior key points and lateral key points. There are 10 anteroposterior key points in total, and their specific position distribution is shown in Table 1 and Table 2. Figure 2As shown in Table 2; the lateral key points are slightly different from the frontal key points. There are 13 lateral key points in total, and their specific location distribution is shown in Table 2 and Figure 3 As shown. Two key points are determined for the tibia and fibula respectively. When a certain part is blocked, it can be supplemented by another part, so as to better measure the angle of knee flexion of the lower limb (the angle between the femur and tibia or the femur and fibula). Two points are used to determine the left and right boundaries of the fibula head. When there is occlusion on one side, the size of the fibula head can be deduced from the other point by using symmetry, so as to better measure the occlusion ratio of the fibula head. It is worth noting that the position and number of key points defined in this embodiment are only a preferred implementation method, and do not limit or exclude other feasible implementation methods-key points with different positions and numbers.

[0065] Table 1 Distribution of positive key points

[0066]

[0067] Table 2 Distribution of key points in lateral position

[0068]

[0069] As an optional embodiment, the scoring method for whether the surrounding soft tissue is intact in the anteroposterior position includes:

[0070] Detect the edge of the input image based on the edge detection operator;

[0071] Determine whether the soft tissue is completely wrapped based on the gradient changes at the edge;

[0072] If yes, the score is S1; otherwise, the score is 0.

[0073] Starting from this embodiment, several typical methods for calculating and scoring quality control index parameters are given in succession. The calculation of most index parameters of the present invention is based on key points, but whether the soft tissue is completely wrapped in this embodiment is an exception and does not require the help of key points. This embodiment mainly determines whether the soft tissue around the positive image is intact. Taking into account the complexity of the actual situation, this embodiment uses a traditional edge detection algorithm to judge the integrity of the soft tissue. Specifically, the edge of the input image is detected by using a mature edge detection operator, such as the Sobel operator. Then, whether the soft tissue is completely wrapped is determined by whether the gradient change at the edge is continuous. If the judgment result is that the soft tissue is completely wrapped, the score for this item is S1; otherwise, the score is 0. S1 is a score set according to the importance of the indicator. The more important the indicator, the higher the score. The same is true for S2 to S5 in the following embodiments.

[0074] As an optional embodiment, the scoring method for the degree of overlap of the medial and lateral femoral condyles in the lateral view includes:

[0075] Using the lateral patella key point I, the posterior edge key point J of the medial and lateral femoral condyle, and the distal condyle key point K of the medial and lateral femoral condyle, the bilateral shadow area of ​​the medial and lateral femoral condyle is segmented from the input lateral image;

[0076] Normalize the segmented region image to a fixed size;

[0077] Calculate the ratio of the double-sided shadow area in the segmented region to the segmented region area;

[0078] If the ratio is greater than the set threshold, the score is S2; otherwise, the score is 0.

[0079] This embodiment provides a technical solution for scoring the degree of overlap of the medial and lateral condyles of the femur in the lateral view. This embodiment converts the problem of overlap of the medial and lateral condyles of the femur into a problem of measuring the ratio of bilateral shadows, effectively bypassing the need for a large number of samples in different overlapping states in traditional classification methods. Another advantage of this conversion method is that the overlap ratio can be linearly measured. This embodiment uses the lateral key points I (patella key point), J (posterior edge key point of the medial and lateral condyles of the femur), and K (distal condyle key point of the medial and lateral condyles of the femur) to cut out the medial and lateral condyles from the original image. Compared with directly using the original image for segmentation, it can significantly reduce background interference and memory usage, and can be completed with only a small network and video memory. The purpose of normalizing the cut-out medial and lateral condyle structures is to eliminate the problem of inaccurate final proportions caused by different sizes of the cut-out areas. By adjusting the length and width to a fixed size, it is helpful to improve the accuracy of the calculation of the bilateral shadow proportion. Finally, the proportion is compared with the set threshold, and a score is given based on the comparison result. The schematic diagram of bilateral shadow segmentation is shown in FIG. Figure 6 shown.

[0080] As an optional embodiment, the scoring method for the degree of overlap between the patella and the femur in the lateral view includes:

[0081] Take the key point K of the distal condyle of the medial and lateral femoral condyle as the center and draw a square with the distance between 4 times K and the key point L of the patellar surface of the medial and lateral femoral condyle as the side length;

[0082] The square area is cut out from the input lateral image to obtain a distribution position structure diagram of the patella and femur;

[0083] The confidence of the overlap between the patella and femur is obtained using a classification network;

[0084] If the confidence level is less than the set threshold, the score is S3; otherwise, the score is 0.

[0085] This embodiment provides a technical solution for scoring the degree of overlap between the patella and the femur in the lateral position. This embodiment mainly uses a classification network to determine whether the patella in the lateral position overlaps with the femur. This embodiment uses the lateral key points K and L for calculation. Specifically, with K as the center position and a square with a side length of 4 times the distance between L and K, the corresponding area in the original image is cut to obtain a structural diagram of the distribution position of the patella and femur. Subsequently, a simple classification network can be used to determine whether the two overlap. For example, when the confidence of the classification network is greater than 0.5, it is determined that the current output is overlapping, and scoring is performed accordingly.

[0086] As an optional embodiment, the scoring method for the overlap between the fibular head and the tibia includes:

[0087] The direction of the fibula is determined by the straight line EF passing through the key point E of the fibula neck and the key point F of the mid-section of the fibula;

[0088] Calculate the distances L1, L2, L3 from the fibular head key points G1, G2 and the overlapping key point H of the tibia and fibular head to the straight line EF respectively;

[0089] Calculate the fibular head size L f : If there is no occlusion, L f =L1+L2; If one of G1 and G2 is blocked, calculate using symmetry, L f =2*L1 or L f =2*L2;

[0090] Calculate the length Lc of the overlapping part of the fibula head and the tibia: if G1 and H are on both sides of the straight line EF, Lc = L1 + L3; if G1 and H are on the same side of the straight line EF, Lc = L1-L3; if G1 is blocked, Lc = L2-L3;

[0091] Calculate the overlap ratio of the fibular head to the tibia Lc / L f ;

[0092] If the overlap is within the set threshold range, the score is S4; otherwise, the score is 0.

[0093] This embodiment provides a technical solution for scoring the overlap between the fibula head and the tibia. This embodiment converts the overlap problem into a problem of calculating the distance from a point to a straight line, which greatly simplifies the calculation complexity. This embodiment provides a solution to the occlusion problem by using the symmetry of redundant key points (G1 or G2), further verifying the benefits of key point annotation. The calculation process diagram is shown in Figure 4 As shown, no further detailed description is given here.

[0094] As an optional embodiment, the scoring method of the lateral lower limb knee flexion angle includes:

[0095] Draw a straight line AB through the key point A of the mid-femoral section and the key point B of the femoral condyle in the lateral position;

[0096] Draw a straight line CD through the key point C of the lateral tibial tuberosity and the key point D of the mid-tibia, intersecting the straight line AB at point O;

[0097] Calculate the size of the angle BOC;

[0098] If the angle BOC is within the set threshold range, the score is S5; otherwise, the score is 0.

[0099] This embodiment provides a technical solution for scoring the lateral lower limb knee flexion angle. This embodiment uses two key points on the relevant bones for positioning and converts the lower limb knee flexion angle into the angle between two straight lines AB and CD. This is another example of using key points to simplify calculations. The schematic diagram of the angle calculation process is shown in FIG. Figure 5 The threshold range of the lower limb knee flexion angle in this embodiment is generally set to 130-140 degrees.

[0100] The previous embodiments provide several typical methods for calculating and scoring quality control indicators. The calculation and scoring methods for several other quality control indicators (such as whether the long axis of the limb in the positive position is parallel to the long axis of the film, whether the patella in the positive position is located in the middle of the femur, etc.) are similar to the methods given above. They are all implemented through simple geometric algebraic calculations using the marked key points. Specific technical solutions will not be given one by one here.

[0101] As an optional embodiment, the method also includes a quantitative score correction step: detecting whether there are multiple knee joints, artifacts, foreign objects, postoperative changes, and other body positions in the input image, and if so, deducting a certain score from the knee joint quantitative score; the artifacts include pants artifacts and knee pad artifacts.

[0102] This embodiment provides a technical solution for quantitative scoring correction. The quality control indicators given above are all indicators that have a relatively obvious impact on the quality of knee joint X-ray images, but there are many factors that affect image quality. After scoring and summarizing these typical indicators, the influence of other factors must also be considered, that is, scoring correction is performed. The other factors mainly include whether there are artifacts, foreign matter, postoperative changes, etc. If artifacts, foreign matter, and postoperative changes are detected in the input image, a certain score is deducted from the summarized quantitative score to make the score more objective and accurate.

[0103] Figure 7 The present invention is a schematic diagram of a knee joint X-ray image quality control device according to an embodiment of the present invention, wherein the device comprises:

[0104] An image acquisition module 11 is used to acquire anteroposterior and lateral X-ray images of the knee joint;

[0105] A key point labeling module 12 is used to detect key points located at the front and side positions of various bones of the knee joint for positioning and eliminating occlusion effects;

[0106] The indicator scoring module 13 is used to perform image segmentation based on the key points, score each quality control indicator by calculating the relevant parameters representing the quality control indicator, and obtain a quantitative score for the knee joint X-ray image after summarization.

[0107] The device of this embodiment can be used to perform Figure 1 The technical solution of the method embodiment shown has similar implementation principles and technical effects, which will not be repeated here.

[0108] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A knee joint X-ray image quality control method, characterized in that: The following steps are involved: Obtain anteroposterior and lateral x-rays of the knee; Detect the key points of the frontal and lateral positions of each bone in the knee joint for positioning and eliminating occlusion effects; Image segmentation is performed based on the key points to crop out the background area, and each quality control indicator is scored by calculating the relevant parameters characterizing the quality control indicator, and the quantitative score of the knee joint X-ray image is obtained after summarization; the relevant parameters include the distance from the point to the straight line, the angle between the two straight lines, and the ratio; the scoring of the quality control indicator parameters is assigned different weights, i.e., scores, according to their importance, and the quality control indicator parameters include at least one of whether the surrounding soft tissue in the anteroposterior view is intact, the degree of overlap between the femur and the medial and lateral condyles in the lateral view, the degree of overlap between the patella and the femur in the lateral view, the lateral lower limb knee flexion angle, the degree of overlap between the fibular head and the tibia, and whether the patella in the anteroposterior view is located in the middle of the femur.

2. The knee joint X-ray image quality control method according to claim 1, characterized in that: The method also includes the following preprocessing steps: data collection, data screening and desensitization, labeling and review, and data conversion.

3. The knee joint X-ray image quality control method according to claim 1, characterized in that: The key points of the frontal position include: the key point of the mid-section of the femur, the key point of the femoral condyle, the key point of the tibial tuberosity, the key point of the mid-section of the tibia, the key point of the fibula neck, the key point of the mid-section of the fibula, the key point of the fibula head, the key point of the overlap between the tibia and the fibula head, and the key point of the patella; the key points of the lateral position include: the key point of the mid-section of the femur, the key point of the femoral condyle, the key point of the tibial tuberosity, the key point of the mid-section of the tibia, the key point of the fibula neck, the key point of the mid-section of the fibula, the key point of the fibula head, the key point of the overlap between the tibia and the fibula head, the key point of the patella, the key point of the posterior edge of the medial and lateral femoral condyles, the key point of the distal condyle of the medial and lateral femoral condyles, and the key point of the patellar surface of the medial and lateral femoral condyles.

4. The knee joint X-ray image quality control method according to claim 3, characterized in that: The scoring methods for whether the surrounding soft tissue is intact include: Detect the edge of the input image based on the edge detection operator; Determine whether the soft tissue is completely wrapped based on the gradient changes at the edge; If yes, the score is S1; otherwise, the score is 0.

5. The knee joint X-ray image quality control method according to claim 3, characterized in that: Scoring of the degree of overlap of the medial and lateral femoral condyles in the lateral view includes: Using the lateral patella key point I, the posterior edge key point J of the medial and lateral femoral condyle, and the distal condyle key point K of the medial and lateral femoral condyle, the bilateral shadow area of ​​the medial and lateral femoral condyle is segmented from the input lateral image; Normalize the segmented region image to a fixed size; Calculate the ratio of the double-sided shadow area in the segmented region to the segmented region area; If the ratio is greater than the set threshold, the score is S2; otherwise, the score is 0.

6. The knee joint X-ray image quality control method according to claim 3, characterized in that: The degree of overlap between the patella and femur in the lateral view is scored as follows: Take the key point K of the distal condyle of the medial and lateral femoral condyle as the center and draw a square with the distance between 4 times K and the key point L of the patellar surface of the medial and lateral femoral condyle as the side length; The square area is cut out from the input lateral image to obtain a distribution position structure diagram of the patella and femur; The confidence of the overlap between the patella and femur is obtained using a classification network; If the confidence level is less than the set threshold, the score is S3; otherwise, the score is 0.

7. The knee joint X-ray image quality control method according to claim 3, characterized in that: Scoring of the fibular head-to-tibia overlap includes: The direction of the fibula is determined by the straight line EF passing through the key point E of the fibula neck and the key point F of the mid-section of the fibula; Calculate the distances L1, L2, L3 from the key points G1, G2 of the fibula head and the key point H of the overlap of the tibia and fibula head to the straight line EF respectively; Calculate the fibular head size L f : If there is no occlusion, L f =L1+L2; If one of G1 and G2 is blocked, calculate using symmetry, L f =2*L1 or L f =2*L2; Calculate the length Lc of the overlapping part of the fibula head and tibia: if G1 and H are on both sides of the straight line EF, Lc=L1+L3; if G1 and H are on the same side of the straight line EF, Lc=L1-L3; if G1 is blocked, Lc=L2-L3; Calculate the overlap ratio of the fibular head to the tibia Lc / L f ; If the overlap ratio is within the set threshold range, the score is S4; otherwise, the score is 0.

8. The knee joint X-ray image quality control method according to claim 3, characterized in that: The scoring method for the lateral lower limb knee flexion angle includes: Draw a straight line AB through the key point A of the mid-femoral section and the key point B of the femoral condyle in the lateral position; Draw a straight line CD through the key point C of the lateral tibial tuberosity and the key point D of the mid-tibia, intersecting the straight line AB at point O; Calculate the size of the angle BOC; If the angle BOC is within the set threshold range, the score is S5; otherwise, the score is 0.

9. The knee joint X-ray image quality control method according to claim 1, characterized in that: The method also includes a quantitative score correction step: detecting whether there are multiple knee joints, artifacts, foreign bodies, postoperative changes, and other body positions in the input image, and if so, deducting a certain score from the knee joint quantitative score; the artifacts include pants artifacts and knee pad artifacts.

10. A knee joint X-ray image quality control device, characterized in that: include: An image acquisition module, used to acquire anteroposterior and lateral X-ray images of the knee joint; Key point detection module, used to detect the key points located at the front and side positions of each bone of the knee joint for positioning and eliminating occlusion effects; The indicator scoring module is used to perform image segmentation based on the key points to crop out the background area, score each quality control indicator by calculating the relevant parameters characterizing the quality control indicator, and obtain a quantitative score for the knee joint X-ray image after summarizing; the relevant parameters include the distance from the point to the straight line, the angle between the two straight lines, and the ratio; the scoring of the quality control indicator parameters is assigned different weights, i.e., scores, according to their importance, and the quality control indicator parameters include at least one of whether the surrounding soft tissue in the anteroposterior position is intact, the degree of overlap between the femur and the medial and lateral condyles in the lateral position, the degree of overlap between the patella and the femur in the lateral position, the lateral lower limb knee flexion angle, the degree of overlap between the fibular head and the tibia, and whether the anteroposterior position patella is located in the middle of the femur.

Images