Method and system for improving microdose ct image quality

Abstract

The present application relates to the technical field of image data processing, in particular to a method and system for improving the CT image quality of micro-radiation dose, comprising: acquiring a plurality of continuous frames of CT scan images of the same part of a detected person, screening a plurality of selected matching points from feature points in each frame of CT scan image, screening a plurality of key points from the selected matching points, acquiring the motion vector of the key points in each frame of CT scan image according to the matching of the key points in adjacent frames of CT scan image, embedding the motion vector of the key points in all frames of CT scan image as prior knowledge into a statistical iterative reconstruction algorithm, and outputting a reconstructed CT scan image through the updated statistical iterative reconstruction algorithm. The present application improves the CT image reconstruction quality of micro-radiation dose by acquiring the motion vector of the accurate key points in the CT scan image.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, in particular to a method and system for improving the quality of micro-radiation dose CT images. BACKGROUND

[0002] Micro-radiation dose CT images refer to a new type of CT imaging technology that can ensure image quality or even improve image quality at extremely low radiation dose through various imaging algorithm techniques. It can significantly reduce the radiation dose in CT examination while ensuring image quality, thereby reducing the radiation damage to patients. The realization of micro-radiation CT mainly relies on image reconstruction algorithms, that is, an initial scanning image is first constructed, different angle integrations are performed on the initial image to obtain simulated projection images, the differences between the simulated projection images and the actual projection images are compared, and the initial image is adjusted by inversely propagating the differences, and then the process is repeatedly performed to continuously reduce the error, and finally the micro-radiation dose CT image is obtained.

[0003] However, when scanning the human body, various organs in the abdominal region of the human body are always in certain rhythmic movements (such as intestinal peristalsis and heartbeats). Due to the rhythmic movements of the organs of the human body, some motion artifacts will always appear during scanning. Therefore, motion vectors are constructed by comparing the differences between corresponding feature points in multiple images, and then the motion of these organs is compensated by using the motion vectors to eliminate the motion artifacts.

[0004] Existing problems: The existing selection and matching of feature points are mainly performed by using the optical flow method and the deformation registration method. The corner point detection algorithm, such as the SIFT (Scale Invariant Feature Transform) algorithm, is used to determine the feature points at different times, and then the nearest neighbor search and block matching technology are used to match the feature points between adjacent images. However, since the micro-radiation dose CT image can be more accurately positioned during scanning, the motion artifacts are further enlarged in the projection image due to the projection relationship. This makes it difficult to accurately position the determined feature points and the matching relationship between the feature points, resulting in errors in the final obtained motion vectors, and thus the predetermined artifacts cannot be eliminated, which may result in poor quality of the micro-radiation dose CT image. SUMMARY

[0005] The present application provides a method and system for improving the quality of micro-radiation dose CT images to solve the existing problems.

[0006] The method and system for improving the quality of micro-radiation dose CT images provided by the present application adopt the following technical solutions:

[0007] One embodiment of the present application provides a method for improving the quality of micro-radiation dose CT images, which includes the following steps:

[0008] Acquire several consecutive frames of CT scan images of the same part of the person being examined;

[0009] Using any frame of a CT scan image as the target image, select several frames of images with the same frequency from the CT scan images other than the target image; obtain several feature points in each frame of the CT scan image, and select several candidate matching points from the feature points in the target image;

[0010] Based on the matching results of candidate matching points in the target image in the same frequency image, several key points are selected from the candidate matching points in the target image;

[0011] Based on the matching results of key points in adjacent CT scan images, the motion vectors of key points in each CT scan image are obtained; the motion vectors of key points in all CT scan images are embedded as prior knowledge into the statistical iterative reconstruction algorithm to obtain the updated statistical iterative reconstruction algorithm; each CT scan image is input into the updated statistical iterative reconstruction algorithm to output the reconstructed CT scan image.

[0012] Furthermore, the specific steps for selecting several frames of images with the same frequency from the CT scan images other than the target image are as follows:

[0013] In any row of any frame of a CT scan image, the gray values ​​of all pixels are counted from left to right to form a gray value sequence.

[0014] All frames of CT scan images other than the target image are designated as reference images;

[0015] Compare the target image with the first The mean of the inversely proportional normalized DTW distances between the reference image and the grayscale value sequences of all identical rows is denoted as the distance between the target image and the reference image. Overall graph similarity of frame reference images;

[0016] Based on the overall similarity between the target image and each frame of the reference image, the co-frequency image of the target image is determined.

[0017] Furthermore, the specific steps for determining the co-frequency images of the target image based on the overall similarity between the target image and each frame of the reference image are as follows:

[0018] Reference images whose overall image similarity is greater than or equal to a preset similarity threshold are denoted as images with the same frequency as the target image.

[0019] Furthermore, the specific steps for selecting several candidate matching points from the feature points in the target image are as follows:

[0020] For any feature point in the target image In the Obtaining feature points from frames of images with the same frequency The feature point that has the closest Euclidean distance to the position coordinates of the feature point is denoted as the feature point. The matching point, if the feature point With feature points When the Euclidean distance between the location coordinates of the matching point and the target point is less than a preset distance threshold, the feature point will be... Record them as candidate matching points.

[0021] Furthermore, the specific steps for selecting several key points from the candidate matching points in the target image are as follows:

[0022] In the target image, each candidate matching point is used as the initial cluster center. Based on the Euclidean distance between pixels, all pixels are clustered only once to obtain the connected region corresponding to each candidate matching point.

[0023] For any connected region To connect regions From the center point to the connected region The vector formed by the Euclidean distance and direction of the center points of each adjacent connected region is denoted as the connected region. The neighborhood vector of the corresponding candidate matching point;

[0024] Based on the method of obtaining the neighborhood vector of each candidate matching point in the target image, obtain the neighborhood vector of the matching point in each frame of the same frequency image for each candidate matching point;

[0025] According to the target image, the first The first candidate matching point The neighborhood vector and the first The candidate matching point is at the _th ... The cosine similarity of each neighborhood vector of a matching point in a frame of the same frequency image is used to determine the first [frame name] in the target image. The first candidate matching point The maximum similarity of neighboring vectors;

[0026] Based on the maximum similarity, from the target image, the first... Select matching vectors from the neighborhood vectors of candidate matching points;

[0027] The first in the target image The mean of the maximum similarity of all matching vectors of the nth candidate matching point is denoted as the mean of the maximum similarity of the vectors of the nth candidate matching point in the target image. The candidate matching point is at the _th ... Structural matching degree in frames with the same frequency;

[0028] Based on the degree of structural matching of each candidate matching point in the target image in each frame of the same frequency image, several key points are selected.

[0029] Furthermore, the determination of the first in the target image The first candidate matching point The specific steps involved in finding the maximum similarity of n neighborhood vectors are as follows:

[0030] The first in the target image The first candidate matching point The neighborhood vector and the first The candidate matching point is at the _th ... The maximum cosine similarity among all neighborhood vectors of a matching point in a frame of the same frequency image is denoted as the i-th similarity among the target image's neighborhood vectors. The first candidate matching point The maximum similarity of neighboring vectors.

[0031] Furthermore, the step of selecting the first image from the target image based on the maximum similarity... The specific steps for selecting matching vectors from the neighborhood vectors of candidate matching points are as follows:

[0032] For the target image, the first For each candidate matching point, the neighborhood vectors with the maximum similarity greater than or equal to a preset vector threshold are denoted as matchable vectors.

[0033] Furthermore, the specific steps for selecting several key points based on the structural matching degree of each candidate matching point in the target image in each frame of the same frequency image are as follows:

[0034] The product of the preset quantity coefficient and the number of images with the same frequency is recorded as the quantity threshold;

[0035] For the target image, the first The structural matching degree of the candidate matching point in all frames of same frequency images is used to determine the matching points. Images with a structural matching degree greater than or equal to a preset matching threshold are denoted as matchable images. When the number of matchable images exceeds a threshold, the target image is selected as the first matching point. There are 10 candidate matching points, denoted as key points.

[0036] Furthermore, the specific steps for obtaining the motion vectors of key points in each frame of CT scan image are as follows:

[0037] Obtain the first... Each key point in a frame of a CT scan image is at the 1st Key points on a frame CT scan image that were successfully matched;

[0038] Get the frame CT scan image position coordinates of the key point to the frame CT scan image position coordinates of the key point to the frame CT scan image frame CT scan image motion vector of the key point.

[0039] The application further provides a system for improving the quality of a CT image with a micro-radiation dose, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and the processor executes the computer program stored in the memory to realize the steps of the method for improving the quality of a CT image with a micro-radiation dose.

[0040] The technical scheme of the application has the following beneficial effects:

[0041] In the embodiment of the application, a plurality of frames of CT scan images of the same part of a human body are acquired, a plurality of candidate matching points are selected from feature points in each frame of CT scan image, and a plurality of key points are selected from the candidate matching points in each frame of CT scan image, so that the key points that can maintain the original structure in the motion are selected according to the motion law of the organ tissue, the accuracy of the selection of the key points is improved, and the accuracy of the motion vector analysis of the subsequent key points is ensured. The motion vectors of the key points in each frame of CT scan image are acquired through the matching of the key points in adjacent frames of CT scan image, the motion vectors of the key points in all frames of CT scan image are embedded into a statistical iterative reconstruction algorithm as prior knowledge, an updated statistical iterative reconstruction algorithm is acquired, the motion vector is used to correct the statistical iterative reconstruction algorithm, and the high-quality reconstruction of the subsequent images is ensured. Each frame of CT scan image is input into the updated statistical iterative reconstruction algorithm, and a reconstructed CT scan image is output. Thus, the motion vectors of the accurate key points in the CT scan image are acquired, the problem of the feature point matching error caused by the similar texture is effectively avoided by combining the structural features of the organ, the efficiency of eliminating the motion artifacts of the CT scan image is greatly improved, and the quality of the CT image reconstruction with a micro-radiation dose is improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only show some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0043] Figure 1 FIG. 1 is a flow chart of a method for improving CT image quality at a low radiation dose according to an embodiment of the present application;

[0044] Figure 2 FIG. 2 is a schematic diagram of a CT machine structure;

[0045] Figure 3 FIG. 3 is a CT image reconstructed without removing motion artifacts;

[0046] Figure 4 FIG. 4 is a CT image reconstructed after removing motion artifacts. DETAILED DESCRIPTION

[0047] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined object, the following describes in detail the specific implementation, structure, features and effects of a method and system for improving CT image quality at a low radiation dose according to the present application, with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0049] The following describes in detail the specific scheme of a method and system for improving CT image quality at a low radiation dose according to the present application, with reference to the accompanying drawings.

[0050] Referring to FIG. 1, Figure 1 which shows a flow chart of steps of a method for improving CT image quality at a low radiation dose according to an embodiment of the present application, the method includes the following steps:

[0051] Step S001: Obtain a plurality of CT scan images of the same part of a subject.

[0052] In this embodiment, a CT machine is used to collect a plurality of CT scan images of the same part of a subject. The CT scan images are grayscale images.

[0053] The specific acquisition process is as follows: first, the whole machine system is powered on, and the current state of each hard component is checked by using a self-checking program, and the CT machine is calibrated and calibrated. Then, the laser positioning system at the scanning head is used to position the detected organ, and the exposure parameters of the radiation source, the acquisition parameters of the area array detector, and the operation parameters of each axis motion mechanism are input. Then, according to the setting of the scanning parameters, the corresponding scanning motion is carried out, and synchronously, the shaped X-ray beam irradiates the human body and interacts with the human tissue, and the energy-changed photons enter the specific metal grid in a certain direction and reach the corresponding position of the two-dimensional area array detector to be counted, thereby obtaining the corresponding two-dimensional array. These two-dimensional arrays are arranged to obtain the CT scanning two-dimensional image. After the scanning is completed, the axis motion mechanism is returned to the original position, and the person being detected gets off the bed. The structure diagram of the CT machine is shown in Figure 2 The CT image reconstructed without removing the motion artifact is shown in Figure 3

[0054] It should be noted that: when CT scanning is performed, the scanned person lies on the bed, and the bed drives the scanned person to the predetermined position, and then the scanning head releases a micro-radiation dose of X-rays. The X-rays pass through the human body and interact with the internal tissues of the human body, and then are received by the special grid in the detector to form a two-dimensional scanning image of the corresponding position. During scanning, due to the movement of various internal tissues of the human body (such as peristalsis of the intestines and stomach), the internal organs and their edges will become blurred due to movement during X-ray scanning. The edges of different organs originally have their unique boundary characteristics, but due to the existence of motion artifacts and the existence of a certain inclination angle of X-rays when irradiating the human tissue, the edge part of each organ tissue in the two-dimensional scanning image becomes wider, so that the positioning of the original feature points becomes inaccurate. However, since the movement of the human organ tissue has a certain regularity, it will have a certain periodicity, and although the edge part of the organ becomes wider due to movement and inclination, when the organ tissue moves to the same position, it is the image of the same organ, which makes the common repeated structural features in multiple images more obvious, which can more accurately reflect the characteristics of the original organ edge. Accordingly, the feature points and feature edges in each scanning image can be constructed, and then the changes of the feature points between continuous time points are investigated, thereby constructing the motion vector of each pixel point.

[0055] Step S002: selecting a number of same-frequency images from the CT scanning images except the target image, taking any one CT scanning image as the target image; and selecting a number of candidate matching points from the feature points in the target image.

[0056] Preferably, in an embodiment of the present application, the method for obtaining the candidate matching points comprises:

[0057] ​The SIFT algorithm is used to obtain several feature points in each frame of CT scan image.

[0058] It should be noted that the SIFT (Scale-Invariant Feature Transform) algorithm is a well-known technique, and its specific method will not be described here. Regardless of the movement of the problem organ, its surface will always have some wrinkles or bends. Although these wrinkles and bends may change position due to movement, their characteristic structures always remain. For example, there are always certain intersection points at the contact points of two organs. Therefore, the SIFT algorithm can be used to process numerous scanned 2D images to obtain the feature corner points in the scanned images. First, the scanned images are Gaussian blurred to construct a scale space and generate a corresponding difference Gaussian pyramid. Then, in each layer, the grayscale values ​​of the current pixel and its neighbors are compared to mark local extrema. Next, the principal curvature ratio of these extrema is calculated using the Hessian matrix. If the ratio exceeds a threshold (preset to 10:1 in this embodiment, used as an example for description), it is determined to be an edge artifact (such as the stripe artifacts on the edge of a rib) and is removed. This yields the selectable feature points in the scanned 2D image.

[0059] It should be further explained that during a CT scan, human organs are constantly moving, and these organs always follow certain patterns. After a certain period of time, they will return to their original positions. However, since there are many organs in the human abdomen, different organ tissues may be covered in the same cross section (the target location of the CT scan). Therefore, it is necessary to look for images that are one cycle apart based on the location where the organs have moved to, in order to eliminate some candidate points (optional feature points) that are significantly different.

[0060] In any row of any frame of a CT scan image, the gray values ​​of all pixels are counted from left to right to form a gray value sequence.

[0061] Any frame of a CT scan image is designated as the target image, and all other frames of CT scan images are designated as reference images.

[0062] Acquire target image and the first DTW distance between grayscale value sequences in the same row of the frame reference image The inverse proportional normalized value is used to compare the target image with the first... The mean of the inversely proportional normalized DTW distances between the reference image and the grayscale value sequences of all identical rows is denoted as the distance between the target image and the reference image. Overall graph similarity of the frame reference image.

[0063] It should be noted that: in this embodiment, the DTW (Dynamic Time Warping) algorithm is used to obtain the DTW distance of two gray value sequences, and the smaller the DTW distance is, the more similar the two gray value sequences are. It is a known technology, and the specific method is not introduced here. Again The inverse proportional relationship and the normalization processing are presented The implementer can set the inverse proportional function and the normalization function according to the actual situation.

[0064] According to the above manner, the overall image similarity of the target image and each frame of reference image can be obtained.

[0065] The preset similarity threshold is 0.75, and the preset distance threshold is 5. This is described as an example.

[0066] The reference image with the overall image similarity greater than or equal to the preset similarity threshold is recorded as the same frequency image of the target image.

[0067] It should be noted that: when the overall image similarity is greater than or equal to the preset similarity threshold, it is considered that the target image and the reference image correspond to the scanning of the organ tissue which just moves to the same approximate position, that is, the same frequency image, and thus a plurality of frames of the same frequency image of the target image are obtained.

[0068] For any feature point in the target image , the feature point with the Euclidean distance closest to the position coordinates of the feature point in the first frame of the same frequency image is obtained, which is recorded as the matching point of the feature point . If the Euclidean distance between the position coordinates of the feature point and the matching point of the feature point is less than the preset distance threshold, the feature point is recorded as the candidate matching point.

[0069] It should be noted that: in this embodiment, a rectangular coordinate system is constructed with the top point of the lower left corner of each frame of CT scan image as the origin, a square with the horizontal right as the horizontal axis, and the vertical upward as the positive direction. The position coordinates of the feature points in each frame of CT scan image are obtained in the rectangular coordinate system. If there are multiple feature points with the Euclidean distance closest to the position coordinates of the feature point in the first frame of the same frequency image, one of them is randomly taken as the matching point of the feature point . If the Euclidean distance between the position coordinates of the feature point and the matching point of the feature point is greater than or equal to the preset distance threshold, it is considered that the feature point does not exist. The two features are possibly small at the same point on the organ, and it is difficult to match them as matching points, and they are recorded as discarded points, and vice versa, recorded as matching points.

[0070] In the above manner, a plurality of matching points in the target image and matching points of each matching point in each frame of the same frequency image can be obtained.

[0071] Step S003: According to the matching of the matching points in the target image in the same frequency image, a plurality of key points are selected from the matching points in the target image.

[0072] It should be noted that: when the organs are moving, each organ has its unique structure, and its position is fixed, that is, the adjacent organs are fixed. This leads to the fact that the scanning images of different organs have their unique texture features. During the movement of the organs, because a large number of organs are moving, some organs are deformed to a certain extent, which makes it necessary to select some invariant feature points from a large number of optional feature points, that is, no matter how the movement is, these feature points always have a relatively fixed angle structure, so that the feature points at the same position in the two scanning images can be better matched.

[0073] Preferably, in an embodiment of the present application, the method for obtaining the key points comprises:

[0074] In the target image, each matching point is used as an initial clustering center, the Euclidean distance between the position coordinates of the pixels is used as the clustering distance, and the K-means clustering algorithm is used to perform clustering assignment only once to obtain a connected region corresponding to each initial clustering center, that is, each matching point corresponds to a connected region, and only one matching point exists in each connected region.

[0075] It should be noted that: because a large number of matching points are distributed in different organs, it is difficult to describe their adjacent relationship through fixed positional relationship, so each matching point is used as an initial clustering center to perform single clustering assignment. The K-means clustering algorithm is a known technology, and the single clustering assignment is: for each pixel in the target image, calculate the Euclidean distance between the pixel and all initial clustering centers, and assign the pixel to the nearest initial clustering center. In this way, each initial clustering center will correspond to a region, that is, a region composed of all pixels assigned to the center. And because the clustering process will gather spatially close pixels together, a continuous region is formed in the image, that is, an initial clustering center corresponds to a connected region.

[0076] Based on the method of obtaining the connected region corresponding to each candidate matching point in the target image, obtain the connected region corresponding to the matching point in each frame of the same frequency image for each candidate matching point.

[0077] It should be noted that because organs move to different positions in different scanned images, different scanned images will have different numbers of candidate matching points when detecting feature points and selecting candidate matching points. In order to better describe the location information and surrounding structural information of each candidate feature point, it is necessary to describe the structure near the candidate matching point based on its segmented region and the relative positions of its surrounding segmented regions.

[0078] In the target image, for any connected region To connect regions From the center point to the connected region The vector formed by the Euclidean distance and direction of the center points of each adjacent connected region is denoted as the connected region. The corresponding neighborhood vector of the candidate matching point. Thus, several neighborhood vectors are obtained for each candidate matching point.

[0079] Based on the method of obtaining the neighborhood vector of each candidate matching point in the target image, obtain the neighborhood vector of the matching point in each frame of the same frequency image for each candidate matching point.

[0080] It should be noted that if the candidate matching point and its corresponding matching point in the target image and the same frequency image have similar neighborhood features—that is, in several major directions, there are one or more feature points at the same or similar positions, and the texture structure at these feature points is also similar—then it can be said that the candidate matching point and its corresponding matching point are at the same location on the organ. Based on this, the structural matching degree between the candidate matching point and its corresponding matching point in the target image and the same frequency image is constructed.

[0081] Calculate the first in the target image The first candidate matching point The neighborhood vector and the first The candidate matching point is at the _th ... The cosine similarity of each neighborhood vector of the matching point in the same frame of the same frequency image will be used to determine the cosine similarity of the first neighbor vector in the target image. The first candidate matching point The neighborhood vector and the first The candidate matching point is at the _th ... The maximum cosine similarity among all neighborhood vectors of a matching point in a frame of the same frequency image is denoted as the i-th similarity among the target image's neighborhood vectors. The first candidate matching point The maximum similarity of neighboring vectors.

[0082] It should be noted that the calculation of cosine similarity between two vectors is a well-known technique, and the specific method will not be introduced here. The value of cosine similarity is between -1 and 1, where a cosine similarity of 1 means that the two vectors are completely identical.

[0083] Using the above method, the maximum similarity of each neighborhood vector of each candidate matching point in the target image can be obtained.

[0084] The preset vector threshold is 0.5, the preset matching threshold is 0.7, and the preset quantity coefficient is 50%. This will be used as an example for explanation.

[0085] For the target image, the first For each candidate matching point, the neighborhood vectors with the maximum similarity greater than or equal to a preset vector threshold are denoted as matchable vectors.

[0086] The first in the target image The mean of the maximum similarity of all matching vectors of the nth candidate matching point is denoted as the mean of the maximum similarity of the vectors of the nth candidate matching point in the target image. The candidate matching point is at the _th ... Structural matching degree in frames with the same frequency.

[0087] The product of the preset quantity coefficient and the number of images with the same frequency is denoted as the quantity threshold.

[0088] For the target image, the first The structural matching degree of the candidate matching point in all frames of same frequency images is used to determine the matching points. Images with a structural matching degree greater than or equal to a preset matching threshold are denoted as matchable images. When the number of matchable images exceeds a threshold, the target image is selected as the first matching point. There are 10 candidate matching points, denoted as key points.

[0089] Using the method described above, several key points in the target image can be obtained.

[0090] It should be noted that: when the target image contains the first... The greater the structural matching degree of the candidate matching point in each frame of the same frequency image, the more it indicates that the... The greater the probability that a candidate matching point is the same point on an organ as its matching point in each frame of the same frequency image, the more matching images that can be matched are filtered. When the number of matching images exceeds half of the total number of images in the same frequency image, it indicates that the candidate matching point has a greater structural matching degree in a large number of images in the same frequency image, and the more critical the candidate matching point is.

[0091] Using the above method, several key points can be obtained from each frame of a CT scan image.

[0092] Step S004: obtaining the motion vector of the key point in each frame of CT scan image according to the matching result of the key point in the adjacent frame of CT scan image; embedding the motion vector of the key point in all frames of CT scan image into the statistical iterative reconstruction algorithm as prior knowledge to obtain an updated statistical iterative reconstruction algorithm; inputting each frame of CT scan image into the updated statistical iterative reconstruction algorithm to output a reconstructed CT scan image.

[0093] Preferably, in one embodiment of the present application, the method for obtaining the reconstructed CT scan image comprises:

[0094] For the CT scan images of adjacent frames, the motion vector of the key point in the first frame of CT scan image is obtained by a Brute-Force method. The motion vector of the key point in the first frame of CT scan image is obtained by a Brute-Force method. The motion vector of the key point in the first frame of CT scan image is obtained by a Brute-Force method.

[0095] It should be noted that the Brute-Force is a basic feature matching method, which is a known technology and will not be described in detail. The above-mentioned key points are screened in each scan image, which can better describe the position information of the organ at the corresponding moment. Therefore, in this embodiment, the two scan images are matched by the Brute-Force, and the grid division method (known technology) is combined to constrain the space and accelerate the matching speed. That is, all the contrast nodes in the grid of the two adjacent frames of scan images are traversed, the Euclidean distance between the contrast node corresponding descriptors is calculated, the Euclidean distance between a plurality of descriptors is obtained, and if the Euclidean distance between two descriptors is less than a certain threshold (for example, 0.4 times the average Euclidean distance, which is described as an example), it is considered that the two key points are matched, which is a known matching process. If the key point in the first frame of CT scan image does not exist in the first frame of CT scan image, the subsequent analysis of the key point is not performed. The motion vector of the key point in the first frame of CT scan image is obtained by a Brute-Force method. The motion vector of the key point in the first frame of CT scan image is obtained by a Brute-Force method.

[0096] The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained. The position coordinates of the key point in the first frame of CT scan image are obtained.

[0097] Thus, the motion vector of each key point in each frame of CT scan image is obtained, and in the embodiment, the motion vector of the key point in the last frame of CT scan image is not analyzed.

[0098] The motion vector of the key point in all frames of CT scan image is embedded into a statistical iterative reconstruction algorithm (SIR) as prior knowledge, and an updated statistical iterative reconstruction algorithm is obtained.

[0099] Each frame of CT scan image is input into the updated statistical iterative reconstruction algorithm, and a reconstructed CT scan image is output.

[0100] It should be noted that the statistical iterative reconstruction algorithm is an image reconstruction method based on a statistical model, which gradually approximates the real image through iterative optimization. In each iteration, the algorithm updates the image estimate using prior knowledge and observed data until a certain stopping condition is met, which is a known technology and will not be described in detail. Embedding the motion vector of the key point in each frame of image into the statistical iterative reconstruction algorithm (SIR) as prior knowledge is a known feasible method, which can use the motion information of the key point to improve the reconstruction quality of the video, i.e. by modifying the cost function of the algorithm to consider the constraint of the motion vector, so as to use these prior knowledge in the reconstruction process, and correct the motion-aware system matrix from the static projection described by the traditional SIR system matrix. Using the motion information of the key point can help the algorithm better understand the dynamic changes of the scan image content, thereby improving the quality of the reconstructed image. Therefore, in the embodiment, by obtaining the accurate motion vector of the key point, the enhancement effect of the reconstructed CT scan image is ensured. The CT image reconstructed after removing the motion artifact is shown in FIG. 2. Figure 4

[0101] The application further provides a CT image quality improvement system with micro-radiation dose, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program stored in the memory to realize the steps of the aforementioned CT image quality improvement method with micro-radiation dose.

[0102] Thus, the application is completed.

[0103] ​To sum up, in the embodiment of the present application, a plurality of CT scanning images of the same part of a human body are acquired, a plurality of candidate matching points are selected from feature points in each CT scanning image, a plurality of key points are selected from the candidate matching points in each CT scanning image, motion vectors of the key points in each CT scanning image are acquired according to key point matching in adjacent CT scanning images, the motion vectors of the key points in all CT scanning images are embedded into a statistical iterative reconstruction algorithm as prior knowledge, an updated statistical iterative reconstruction algorithm is acquired, each CT scanning image is input into the updated statistical iterative reconstruction algorithm, and a reconstructed CT scanning image is output. The motion vectors of the accurate key points in the CT scanning image are acquired, and the CT image reconstruction quality of the micro radiation dose is improved.

[0104] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of improving the quality of a microdose CT image, characterized by, The method comprises the following steps: Obtain a plurality of continuous CT scan images of the same part of the human body; Select a plurality of same-frequency images from the CT scan images other than the target image, and select a plurality of candidate matching points from the feature points in the target image; Select a plurality of key points from the candidate matching points in the target image according to the matching of the candidate matching points in the target image in the same-frequency images; Obtain the motion vector of the key points in each CT scan image according to the matching result of the key points in the adjacent CT scan images, embed the motion vector of the key points in all CT scan images into the statistical iterative reconstruction algorithm as prior knowledge, obtain the updated statistical iterative reconstruction algorithm, input each CT scan image into the updated statistical iterative reconstruction algorithm, and output a reconstructed CT scan image; The specific steps of selecting a plurality of key points from the candidate matching points in the target image comprise the following steps: In the target image, take each candidate matching point as an initial clustering center, perform clustering assignment on all pixel points only once according to the Euclidean distance between the pixel points, and obtain a connected region corresponding to each candidate matching point; For any one connected region , the center point of the connected region to the Euclidean distance and direction of each adjacent connected region center point of the connected region , the vector formed by the center point of the connected region corresponding to the neighborhood vector of the candidate matching point; According to the acquisition method of the neighborhood vector of each candidate matching point in the target image, obtain the neighborhood vector of the matching point of each candidate matching point in each same-frequency image; According to the cosine similarity of each neighborhood vector of the first candidate matching point in the target image and the matching point in the same frequency image of the first frame, the maximum similarity of the first neighborhood vector of the first candidate matching point in the target image is determined. ​​​​​​ According to the maximum similarity, a matchable vector is selected from a neighborhood vector of a candidate matching point in the target image. a matchable vector is selected from a neighborhood vector of a candidate matching point in the target image. The first in the target image The mean of the maximum similarity of all matching vectors of the nth candidate matching point is denoted as the mean of the maximum similarity of the vectors of the nth candidate matching point in the target image. The candidate matching point is at the _th ... Structural matching degree in frames with the same frequency; According to the size of the structural matching degree of each candidate matching point in the target image in each same-frequency image, select a plurality of key points.

2. The method of claim 1, wherein the microdose CT image quality is improved. The specific steps of selecting a plurality of same-frequency images from the CT scan images other than the target image comprise the following steps: In any row of any CT scan image, sequentially count the gray values of all pixel points from left to right to form a gray value sequence; All CT scan images other than the target image are recorded as reference images; The overall image similarity of the target image and the first frame reference image is denoted as The average of the inverse proportional normalized values of the DTW distances of the gray value sequences of all the same lines of the target image and the first frame reference image is denoted as the overall image similarity of the target image and the first frame reference image. The average of the inverse proportional normalized values of the DTW distances of the gray value sequences of all the same lines of the target image and the first frame reference image is denoted as the overall image similarity of the target image and the first frame reference image. According to the size of the overall image similarity between the target image and each reference image, determine the same-frequency images of the target image.

3. The method of claim 2, wherein the microdose CT image quality is improved. The specific steps of determining the same-frequency images of the target image according to the size of the overall image similarity between the target image and each reference image comprise the following steps: The reference image with an overall image similarity greater than or equal to a preset similarity threshold is recorded as the same-frequency image of the target image.

4. The method for improving the quality of a CT image of a microdose according to claim 1, wherein, The specific steps of selecting a plurality of candidate matching points from the feature points in the target image comprise the following steps: For any one feature point in the target image , in the first frame, the feature point with the Euclidean distance closest to the position coordinate of the feature point is obtained in the same frequency image of the frame, and is recorded as the matching point of the feature point , if the Euclidean distance between the position coordinate of the feature point and the matching point of the feature point is less than a preset distance threshold, the feature point is recorded as a candidate matching point.

5. The method for improving the quality of a CT image of a microdose according to claim 1, wherein, The determination of the target image in the first The first candidate matching point The specific steps involved in finding the maximum similarity of n neighborhood vectors are as follows: The first in the target image The first candidate matching point The neighborhood vector and the first The candidate matching point is at the _th ... The maximum cosine similarity among all neighborhood vectors of a matching point in a frame of the same frequency image is denoted as the i-th similarity among the target image's neighborhood vectors. The first candidate matching point The maximum similarity of neighboring vectors.

6. The method for improving the quality of a CT image of a microdose according to claim 1, wherein, The step of screening the matchable vector from the neighborhood vectors of the candidate matching points in the target image according to the maximum similarity comprises the following specific steps: The step of screening the matchable vector from the neighborhood vectors of the candidate matching points in the target image according to the maximum similarity comprises the following specific steps: For the target image, the first For each candidate matching point, the neighborhood vectors with the maximum similarity greater than or equal to a preset vector threshold are denoted as matchable vectors.

7. The method for improving the quality of a CT image of a microdose according to claim 1, wherein, The specific steps of selecting a plurality of key points from the candidate matching points in the target image according to the size of the structural matching degree of each candidate matching point in the target image in each same-frequency image comprise the following steps: The product of the preset number coefficient and the number of same-frequency images is recorded as a number threshold. For the structure matching degree of the first candidate matching point in the target image in all frame same-frequency images, the same-frequency image with a structure matching degree greater than or equal to a preset matching threshold is recorded as a matchable same-frequency image, and when the number of matchable same-frequency images is greater than a number threshold, the first candidate matching point in the target image is recorded as a key point.

8. The method for improving the quality of a microdose CT image according to claim 1, wherein, The specific steps of obtaining the motion vector of the key points in each CT scan image comprise the following steps: Obtaining the first key point on the frame CT scan image by a brute force matching method key points matched successfully on the frame CT scan image​ acquire the position coordinates of the first key point in the first CT scan image frame CT scan image acquire the position coordinates of the first key point in the first CT scan image frame CT scan image acquire the position coordinates of the first key point in the first CT scan image frame CT scan image frame CT scan image acquire the position coordinates of the first key point in the first CT scan image 9. A system for improving the quality of a microdose CT image, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program comprises the following steps of: The computer program is executed by the processor to realize the steps of the method for improving the quality of the CT image with a small radiation dose according to any one of claims 1-8.

Images