Cardiovascular image recognition method, device and system based on OCT

Abstract

The present invention relates to the technical field of image denoising, and in particular to a method, device and system for cardiovascular image recognition based on OCT, the method comprising: acquiring cardiovascular images based on OCT, and performing light and dark distribution analysis and processing on each pixel in the cardiovascular images; clustering the pixels in the cardiovascular images, and performing edge detection on the target clustering clusters; performing structural texture analysis and processing on all target edges in the target clustering clusters; determining image information coefficients; performing morphological closing operations on the cardiovascular images, and performing superpixel segmentation to determine reference information coefficients; determining target information differences, using the target information differences as a scale factor for Wiener filtering, and performing Wiener filtering on the cardiovascular images to identify the target images. The present invention realizes denoising of cardiovascular images, thereby realizing recognition of cardiovascular images, and improving the effects of denoising and recognizing cardiovascular images.

Description

Technical Field

[0001] The present invention relates to the technical field of image denoising, and in particular to an OCT-based cardiovascular image recognition method, device and system. Background Art

[0002] Cardiovascular images based on OCT (Optical Coherence Tomography) often produce speckle noise during the acquisition process due to tiny irregular reflections between the laser beam and the tissue surface, which affects the subsequent identification of tissue structures in cardiovascular images. Therefore, it is very important to denoise cardiovascular images. At present, when denoising images, the commonly used method can be: performing Wiener filtering on the image to achieve image denoising, wherein the image signal-to-noise ratio is used to set the scale factor of the Wiener filter, and the scale factor is used to adjust the filtering strength of the Wiener filter.

[0003] However, when the image signal-to-noise ratio is used to set the scale factor of the Wiener filter and Wiener filtering is performed on cardiovascular images, the following technical problems often occur:

[0004] It is often difficult to obtain an accurate image signal-to-noise ratio in the absence of a clear reference image, which results in a low accuracy in the scale factor setting of the Wiener filter, which in turn leads to a poor denoising effect on cardiovascular images, and thus leads to a poor recognition effect on cardiovascular images. Summary of the invention

[0005] In order to solve the technical problem that the recognition effect of cardiovascular images is poor due to the poor denoising effect of cardiovascular images, the present invention proposes a cardiovascular image recognition method, device and system based on OCT.

[0006] In a first aspect, the present invention provides a cardiovascular image recognition method based on OCT, the method comprising:

[0007] Obtain cardiovascular images based on OCT, and perform light and dark distribution analysis on each pixel in the cardiovascular images to obtain the light and dark coefficient corresponding to each pixel;

[0008] According to the brightness coefficients corresponding to the pixels, the pixels in the cardiovascular image are clustered to obtain target clusters, and edge detection is performed on each target cluster to obtain the target edge;

[0009] Perform structural texture analysis on all target edges in each target cluster to obtain the structural texture coefficient corresponding to each target cluster;

[0010] According to the differences between the structural texture coefficients corresponding to all target clusters, the image information coefficient corresponding to the cardiovascular image is determined;

[0011] Perform morphological closing operation on cardiovascular images and perform superpixel segmentation to obtain target superpixel blocks. Determine the reference information coefficient based on the edge pixel density corresponding to all target superpixel blocks, where the edge pixel density is represented by the proportion of edge pixels in the target superpixel block.

[0012] According to the image information coefficient and the reference information coefficient, the target information difference is determined, the target information difference is used as the scale factor of Wiener filtering, and the cardiovascular image is subjected to Wiener filtering to obtain the target image, and the target image is identified.

[0013] In combination with the first aspect above, in a possible implementation, performing light and dark distribution analysis processing on each pixel in the cardiovascular image to obtain a light and dark coefficient corresponding to each pixel includes:

[0014] Target maximum points are screened out from the cardiovascular image, and the brightness coefficient corresponding to each pixel point in the cardiovascular image is determined based on the grayscale brightness distribution between all target maximum points and their corresponding reference points within a second preset neighborhood corresponding to each pixel point in the cardiovascular image, wherein the reference point corresponding to the target maximum point is the target maximum point closest to the target maximum point.

[0015] In combination with the first aspect above, in a possible implementation manner, screening out the target maximum point from the cardiovascular image includes:

[0016] Any pixel point in the cardiovascular image is determined as a marked pixel point. If the grayscale value corresponding to the marked pixel point is greater than the grayscale values ​​corresponding to each pixel point in its first preset neighborhood, the marked pixel point is determined as a target maximum point.

[0017] In combination with the first aspect above, in a possible implementation, determining the shading coefficient corresponding to each pixel in the cardiovascular image according to the grayscale shading distribution between all target maximum points and their corresponding reference points in the second preset neighborhood corresponding to each pixel in the cardiovascular image includes:

[0018] Determine any pixel point in the cardiovascular image as a marked pixel point, and determine any target maximum point in a second preset neighborhood corresponding to the marked pixel point as a candidate maximum point;

[0019] Taking the candidate maximum point and its corresponding reference point as endpoints, a target connecting line segment corresponding to the candidate maximum point is drawn;

[0020] Determine the mean of the grayscale values ​​corresponding to all non-maximum points on the target line segment corresponding to the candidate maximum point as the dark grayscale factor corresponding to the candidate maximum point, wherein the non-maximum point is a pixel point in the cardiovascular image other than the target maximum point;

[0021] Determine the average of the grayscale value corresponding to the candidate maximum point and the grayscale value corresponding to the corresponding reference point as the bright grayscale factor corresponding to the candidate maximum point;

[0022] Determine the difference between the bright grayscale factor and the dark grayscale factor corresponding to the candidate maximum point as the initial bright and dark factor corresponding to the candidate maximum point;

[0023] Determine the distance between the candidate maximum point and its corresponding reference point as the reference distance corresponding to the candidate maximum point;

[0024] The brightness coefficient corresponding to the marked pixel point is determined according to the initial brightness factors and reference distances corresponding to all target maximum points in the second preset neighborhood corresponding to the marked pixel point, wherein the initial brightness factor is positively correlated with the brightness coefficient, and the reference distance is negatively correlated with the brightness coefficient.

[0025] In combination with the first aspect, in a possible implementation, performing structural texture analysis on all target edges in each target cluster to obtain a structural texture coefficient corresponding to each target cluster includes:

[0026] Connecting two ends of each target edge to obtain a straight line segment corresponding to each target edge, and determining a straight line direction corresponding to the straight line segment corresponding to each target edge as a reference direction corresponding to each target edge;

[0027] The information entropy of the reference direction corresponding to all target edges in each target cluster is determined as the target information entropy corresponding to each target cluster;

[0028] The average length of the straight line segments corresponding to all target edges in each target cluster is determined as the average target length corresponding to each target cluster;

[0029] The structural texture coefficient corresponding to each target cluster is determined according to the minimum distance between target edges in each target cluster, the angle between the straight line segments corresponding to the target edges in each target cluster, the target information entropy corresponding to each target cluster and the target average length.

[0030] In combination with the first aspect, in a possible implementation, determining the structural texture coefficient corresponding to each target cluster according to the minimum distance between target edges in each target cluster, the angle between the straight line segments corresponding to the target edges in each target cluster, the target information entropy corresponding to each target cluster, and the target average length includes:

[0031] Determine any target cluster as a marked cluster, determine any target edge in the marked cluster as a marked edge, and determine the target edge closest to the marked edge as a reference edge;

[0032] Determine the minimum distance between the marked edge and the reference edge as the target distance corresponding to the marked edge;

[0033] Determine the angle between the straight line segment corresponding to the marked edge and the straight line segment corresponding to the reference edge as the target angle corresponding to the marked edge;

[0034] Determining a substructure feature corresponding to the marked edge according to a target distance and a target angle corresponding to the marked edge, wherein both the target distance and the target angle are negatively correlated with the substructure feature;

[0035] According to the target information entropy and target average length corresponding to the marked clustering cluster, and the substructure features corresponding to all target edges in the marked clustering cluster, the structural texture coefficient corresponding to the marked clustering cluster is determined, wherein the target information entropy, target average length and substructure features are all positively correlated with the structural texture coefficient.

[0036] In combination with the first aspect above, in a possible implementation manner, determining the image information coefficient corresponding to the cardiovascular image according to the difference between the structural texture coefficients corresponding to all target clusters includes:

[0037] The absolute value of the difference between the structural texture coefficients corresponding to each two target clusters is determined as the target difference between each two target clusters;

[0038] The mean of the target differences between all target clusters is determined as the image information coefficient corresponding to the cardiovascular image.

[0039] In combination with the first aspect, in a possible implementation manner, determining the reference information coefficient according to the edge pixel densities corresponding to all target superpixel blocks includes:

[0040] The absolute value of the difference between the edge pixel densities corresponding to each two target super-pixel blocks is determined as the reference difference between each two target super-pixel blocks;

[0041] The mean of the reference differences between all target superpixel blocks is determined as the reference information coefficient.

[0042] The present invention has the following beneficial effects:

[0043] The cardiovascular image recognition method based on OCT of the present invention realizes the denoising of cardiovascular images, thereby realizing the recognition of cardiovascular images, solves the technical problem of poor recognition effect of cardiovascular images due to poor denoising effect of cardiovascular images, and improves the effect of denoising cardiovascular images and the effect of cardiovascular image recognition. Generally speaking, speckle noise is often caused by the phase distortion of light due to tissue unevenness, thereby causing interference to form light spot noise with interlaced light and dark. Therefore, the brightness coefficient corresponding to the quantized pixel point can often characterize the intensity of the pixel point affected by the speckle noise. Since the intensity of the influence of speckle noise on different tissue structures is often different, clustering the pixels in the cardiovascular image based on the brightness coefficient can often achieve the classification of different tissue structures, and different target clusters can often characterize different tissue structures. The structural texture coefficient corresponding to the quantized target cluster cluster can characterize the structural texture characteristics of the target cluster cluster, and the quantized image information coefficient can characterize the contrast between vascular structure areas when noise exists. Generally speaking, the edge pixel density of different tissue structures in the cardiovascular system is often different, and the presence of noise often interferes with the contrast between different tissue structures. The morphological closing operation can close the tissue structure area with a larger edge pixel density, partially close the tissue structure area with an edge pixel density in the middle and partially open it, and keep the tissue structure area with a smaller edge pixel density unchanged, thereby suppressing the influence of noise on different tissue structure areas to a certain extent, reducing the difference within different tissue structure areas, and improving the contrast between different tissue structure areas. Therefore, the quantized reference information coefficient can characterize the contrast between vascular structure areas after suppressing noise. The quantized target information difference can often characterize the impact of noise on cardiovascular structure. Therefore, compared with the traditional filtering method and the scale factor of Wiener filtering set by the image signal-to-noise ratio, the present invention uses the target information difference as the scale factor of Wiener filtering, and performs Wiener filtering on cardiovascular images, which can improve the filtering effect, thereby improving the effect of denoising cardiovascular images, and further improving the effect of identifying cardiovascular images. Among them, the traditional filtering method can be median filtering, maximum filtering or mean filtering. Traditional filtering methods often have certain limitations when dealing with speckle noise. It is often difficult to correctly distinguish between speckle noise and the texture of cardiovascular tissue structure, resulting in the non-smooth texture of cardiovascular tissue structure being filtered out. Spot noise is also called speckle noise.

[0044] In a second aspect, a cardiovascular image recognition device based on OCT is provided, comprising a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes the method in the first aspect or any possible implementation of the first aspect.

[0045] In a third aspect, the present invention provides a cardiovascular image recognition system based on OCT, the system comprising:

[0046] An acquisition processing module acquires cardiovascular images based on OCT, and performs light and dark distribution analysis on each pixel in the cardiovascular images to obtain a light and dark coefficient corresponding to each pixel;

[0047] A cluster detection module is used to cluster the pixels in the cardiovascular image according to the brightness coefficients corresponding to the pixels to obtain target clusters, and to perform edge detection on each target cluster to obtain target edges;

[0048] A structural texture analysis module is used to perform structural texture analysis on all target edges in each target cluster to obtain a structural texture coefficient corresponding to each target cluster;

[0049] An information coefficient determination module is used to determine the image information coefficient corresponding to the cardiovascular image according to the difference between the structural texture coefficients corresponding to all target clusters;

[0050] An operation segmentation determination module is used to perform morphological closing operations on cardiovascular images and perform superpixel segmentation to obtain target superpixel blocks, and determine a reference information coefficient according to the edge pixel density corresponding to all target superpixel blocks, wherein the edge pixel density represents the proportion of edge pixel points in the target superpixel block;

[0051] A filter identification module is determined, which is used to determine the target information difference according to the image information coefficient and the reference information coefficient, use the target information difference as the scale factor of the Wiener filter, perform Wiener filtering on the cardiovascular image, obtain the target image, and identify the target image.

[0052] In a fourth aspect, a computer program product is provided, comprising: a computer program code, which, when executed on a computer, enables the computer to execute the method in the first aspect or any possible implementation of the first aspect.

[0053] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the method in the above-mentioned first aspect or any possible implementation manner of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 A flowchart of a cardiovascular image recognition method based on OCT of the present invention;

[0056] Figure 2 A schematic diagram of a flow chart for determining a structural texture coefficient of the present invention;

[0057] Figure 3 It is a schematic diagram of the structural texture coefficient refinement process of the present invention;

[0058] Figure 4 A schematic diagram of the composition structure of an OCT-based cardiovascular image recognition system of the present invention;

[0059] Figure 5 The figure is a schematic diagram of the structure of a computer device of the present invention. DETAILED DESCRIPTION

[0060] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0061] Unless defined otherwise, 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 invention belongs.

[0062] refer to Figure 1 , shows the process of some embodiments of a cardiovascular image recognition method based on OCT of the present invention. The cardiovascular image recognition method based on OCT comprises the following steps:

[0063] Step S1, obtaining a cardiovascular image based on OCT, and performing light and dark distribution analysis processing on each pixel in the cardiovascular image to obtain a light and dark coefficient corresponding to each pixel.

[0064] Among them, the cardiovascular image based on OCT is also called cardiovascular OCT image. The cardiovascular image based on OCT can be a grayscale image of the cardiovascular acquired by OCT (Optical Coherence Tomography).

[0065] It should be noted that cardiovascular images based on OCT are high-resolution images that can display tiny structures and details inside the human cardiovascular system. They are usually images of cardiovascular cross-sections that can display the cross-sectional structure of blood vessels. However, the images obtained may be affected by various factors, such as blood flow, arterial pulse, etc., which may cause speckle noise in the image, resulting in image blur or distortion, thus affecting the accuracy and interpretability of the image. In order to improve the examination effect of doctors on cardiovascular images, it is often necessary to denoise cardiovascular images.

[0066] As an example, this step may include the following steps:

[0067] The first step is to obtain cardiovascular images based on OCT.

[0068] For example, optical coherence tomography technology can be used to acquire cardiovascular imaging of the patient to be examined as an OCT-based cardiovascular image, wherein the patient to be examined may be a patient who needs to be examined.

[0069] It should be noted that for patients who need to be examined, the doctor will introduce the guide into the patient's vascular system by puncture or intravascular cannulation to determine its position and direction in OCT imaging. When the guide is in the appropriate position, the doctor will introduce the optical probe through the guide. The probe contains a light source and optical elements, which can transmit the light beam to the blood vessel site of interest and receive the reflected light signal; the optical probe will scan the blood vessel wall at a very fast speed and record the reflected light signal at each position. These signals are then converted into digital signals and transmitted to the computer system for processing; the computer system will process the received light signals and use OCT to reconstruct the structural image of the inside of the blood vessel, which contains detailed information on the internal structure such as the blood vessel wall, plaque, and thrombus.

[0070] The second step, screening out the target maximum points from the cardiovascular images, may include the following sub-steps:

[0071] In the first sub-step, any pixel point in the cardiovascular image is determined as a marked pixel point.

[0072] In the second sub-step, if the grayscale value corresponding to the marked pixel point is greater than the grayscale values ​​corresponding to each pixel point in its first preset neighborhood, the marked pixel point is determined as the target maximum point; that is, if the grayscale values ​​corresponding to each pixel point in the first preset neighborhood corresponding to the marked pixel point are all less than the grayscale value corresponding to the marked pixel point, the marked pixel point is determined as the target maximum point.

[0073] The first preset neighborhood may be a preset neighborhood, for example, the first preset neighborhood may be an eight-neighborhood, that is, a 3×3 neighborhood.

[0074] The third step is to determine the brightness coefficient corresponding to each pixel in the cardiovascular image according to the grayscale brightness distribution between all target maximum points and their corresponding reference points in the second preset neighborhood corresponding to each pixel in the cardiovascular image.

[0075] Among them, the second preset neighborhood can be a pre-set neighborhood, and the area of ​​the second preset neighborhood can be larger than the area of ​​the first preset neighborhood. For example, if the first preset neighborhood is an eight-neighborhood neighborhood, the second preset neighborhood can be a twenty-four-neighborhood neighborhood, that is, a 5×5 neighborhood. The reference point corresponding to the target maximum point can be the target maximum point closest to the target maximum point. It should be noted that if there are multiple target maximum points closest to the target maximum point, a target maximum point can be randomly selected from these multiple target maximum points as a reference point.

[0076] For example, determining the brightness coefficient corresponding to each pixel in a cardiovascular image may include the following sub-steps:

[0077] In the first sub-step, any pixel point in the cardiovascular image is determined as a marked pixel point, and any target maximum point in a second preset neighborhood corresponding to the marked pixel point is determined as a candidate maximum point.

[0078] In the second sub-step, the target line segment corresponding to the candidate maximum point is drawn with the candidate maximum point and its corresponding reference point as endpoints.

[0079] Among them, the two endpoints of the target line segment corresponding to the candidate maximum point can be: the candidate maximum point and the reference point corresponding to the candidate maximum point.

[0080] In the third sub-step, the average of the grayscale values ​​corresponding to all non-maximum points on the target line segment corresponding to the candidate maximum point is determined as the dark grayscale factor corresponding to the candidate maximum point.

[0081] Among them, the non-maximum point can be a pixel point other than the target maximum point in the cardiovascular image.

[0082] In the fourth sub-step, the average of the grayscale value corresponding to the candidate maximum point and the grayscale value corresponding to the corresponding reference point is determined as the bright grayscale factor corresponding to the candidate maximum point.

[0083] In the fifth sub-step, the difference between the bright grayscale factor and the dark grayscale factor corresponding to the candidate maximum point is determined as the initial bright and dark factor corresponding to the candidate maximum point.

[0084] In the sixth sub-step, the distance between the candidate maximum point and its corresponding reference point is determined as the reference distance corresponding to the candidate maximum point.

[0085] The seventh sub-step is to determine the brightness coefficient corresponding to the marked pixel point according to the initial brightness factors and reference distances corresponding to all target maximum points in the second preset neighborhood corresponding to the marked pixel point.

[0086] The initial brightness factor may be positively correlated with the brightness coefficient, and the reference distance may be negatively correlated with the brightness coefficient.

[0087] For example, the formula for determining the brightness coefficient corresponding to the pixel point in the cardiovascular image can be:

[0088] ;

[0089] in, It is the first The brightness coefficient corresponding to each pixel. It is the serial number of the pixel in the cardiovascular image. It is the first The number of target maximum points in the second preset neighborhood corresponding to the pixel points. It is the first The sequence number of the target maximum point in the second preset neighborhood corresponding to the pixel point. It is an exponential function with a natural constant as its base. It is the first In the second preset neighborhood corresponding to the pixel point, The dark grayscale factor corresponding to the maximum point of the target; The mean of the grayscale values ​​corresponding to all non-maximum points on the target line segment corresponding to the target maximum point. It is the first In the second preset neighborhood corresponding to the pixel point, The bright grayscale factor corresponding to the maximum point of the target; The grayscale value corresponding to the target maximum point and the average grayscale value corresponding to the corresponding reference point. Generally speaking, the bright grayscale factor is often higher than the dark grayscale factor. It should be noted that when The smaller the time, the more likely it is that The smaller the , it is often said that the greater the difference between the bright grayscale factor and the dark grayscale factor, so, It is The initial brightness factor corresponding to the maximum point of the target can be used to characterize the The difference between the bright grayscale factor and the dark grayscale factor corresponding to the maximum point of the target. It is the first In the second preset neighborhood corresponding to the pixel point, The reference distance corresponding to the maximum point of the target; that is, The distance between a target maximum point and its corresponding reference point.

[0090] It should be noted that, since the grayscale values ​​at the speckle noise are interlaced, the higher the degree of interlacing at the pixel point, the greater the noise impact on the pixel point. The larger the The larger the grayscale change around the target maximum point, the greater the grayscale change around the target maximum point. The target maximum point can represent the point of light and dark change. The smaller the time, the more likely it is that The smaller the distance between the maximum point of a target and its corresponding reference point, the faster the speed of light and dark changes. The larger the The greater the degree of brightness change in the second preset neighborhood corresponding to the pixel point, and the faster the brightness change speed, the greater the brightness change in the second preset neighborhood corresponding to the pixel point. The more pixels there are, the more likely they are to be affected by speckle noise.

[0091] Optionally, if there is no target maximum point in the second preset neighborhood corresponding to the pixel point, it often means that the brightness around the pixel point is relatively uniform and there is no large change in brightness, so the brightness coefficient corresponding to the pixel point can be directly set to 0.

[0092] Step S2, clustering the pixels in the cardiovascular image according to the brightness coefficients corresponding to the pixels to obtain target clusters, and performing edge detection on each target cluster to obtain target edges.

[0093] It should be noted that, since different tissue structures are often affected by speckle noise to different degrees, clustering the pixels in cardiovascular images based on the brightness coefficient can often achieve a rough classification of different tissue structures, and different target clusters can often represent different tissue structures.

[0094] As an example, pixels with similar brightness coefficients can be clustered into the same category by using a clustering algorithm according to the brightness coefficients corresponding to the pixels, and the obtained clusters are used as target clusters. Edge detection is performed on each target cluster by using a canny edge detection algorithm, and the detected edges are recorded as target edges. The clustering algorithm can be, but is not limited to, a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, a K-means clustering algorithm, and a hierarchical clustering algorithm.

[0095] Step S3, performing structural texture analysis processing on all target edges in each target cluster to obtain a structural texture coefficient corresponding to each target cluster.

[0096] It should be noted that, since different target clusters can often represent different tissue structures, and the structural texture coefficients corresponding to the quantized target clusters can represent the structural texture characteristics of the target clusters, different structural texture coefficients can represent the structural texture characteristics of different tissue structures.

[0097] As an example, the process of performing structural texture analysis on all target edges in each target cluster to obtain the structural texture coefficient corresponding to each target cluster can be as follows: Figure 2 As shown, the following steps may be specifically included:

[0098] Step 201 , connecting two ends of each target edge to obtain a straight line segment corresponding to each target edge, and determining a straight line direction corresponding to the straight line segment corresponding to each target edge as a reference direction corresponding to each target edge.

[0099] The reference direction corresponding to the target edge may represent the direction of the target edge. The straight line segment corresponding to the target edge may be a line segment obtained by connecting two ends of the target edge. The straight line direction corresponding to the straight line segment may be the straight line direction corresponding to the straight line obtained by extending the straight line segment. The straight line direction corresponding to the straight line may be represented by the slope of the straight line.

[0100] Step 202: Determine the information entropy of the reference direction corresponding to all target edges in each target cluster as the target information entropy corresponding to each target cluster.

[0101] Step 203: Determine the average length of the straight line segments corresponding to all target edges in each target cluster as the target average length corresponding to each target cluster.

[0102] Step 204, determining the structural texture coefficient corresponding to each target cluster according to the minimum distance between target edges in each target cluster, the angle between the straight line segments corresponding to the target edges in each target cluster, the target information entropy corresponding to each target cluster and the target average length.

[0103] The method for obtaining the angle between any two straight line segments may be: respectively extending the two straight line segments to obtain two straight lines, and the angle between the two straight lines obtained is the angle between the two straight line segments.

[0104] For example, according to the minimum distance between the target edges in each target cluster, the angle between the straight line segments corresponding to the target edges in each target cluster, the target information entropy corresponding to each target cluster, and the target average length, the process of determining the structural texture coefficient corresponding to each target cluster can be as follows: Figure 3 As shown, the following steps may be specifically included:

[0105] Step 301 : determine any target cluster as a marked cluster, determine any target edge in the marked cluster as a marked edge, and determine the target edge closest to the marked edge as a reference edge.

[0106] It should be noted that, if there are multiple target edges closest to the marked edge, one target edge may be randomly selected from the multiple target edges as a reference edge.

[0107] Step 302: determine the minimum distance between the marked edge and the reference edge as the target distance corresponding to the marked edge.

[0108] The method for obtaining the minimum distance between the marked edge and the reference edge may be: taking each pixel point on the marked edge as a marked point, taking each pixel point on the reference edge as a candidate point, determining the distance between each marked point and each candidate point as a candidate distance, and taking the minimum candidate distance as the minimum distance between the marked edge and the reference edge. The reference edge is the target edge having the smallest minimum distance to the marked edge.

[0109] Step 303: determine the angle between the straight line segment corresponding to the marked edge and the straight line segment corresponding to the reference edge as the target angle corresponding to the marked edge.

[0110] Step 304: Determine the substructure feature corresponding to the marked edge according to the target distance and target angle corresponding to the marked edge.

[0111] Among them, both the target distance and the target angle can be negatively correlated with the substructure features.

[0112] Step 305 , determining the structural texture coefficient corresponding to the marked cluster according to the target information entropy and target average length corresponding to the marked cluster, and the substructure features corresponding to all target edges in the marked cluster.

[0113] Among them, the target information entropy, target average length and substructure characteristics can be positively correlated with the structural texture coefficient.

[0114] For example, the formula for determining the structural texture coefficient corresponding to the target cluster can be:

[0115] ;

[0116] in, It is The structural texture coefficients corresponding to the target clusters. is the sequence number of the target cluster. It is a hyperbolic function, and is mainly used for normalization in the embodiment of the present invention. It is The target information entropy corresponding to the target clusters. It is The average length of the targets corresponding to the target clusters. It is The number of target edges in a target cluster. It is The ordinal number of the target edge in the target cluster. It is The first target cluster The target angle corresponding to the target edge. It is The first target cluster The target distance corresponding to each target edge. It is a pre-set factor greater than 0, mainly used to prevent the denominator from being 0, such as, It can be 0.001. It is The first target cluster The substructure features corresponding to the target edges.

[0117] It should be noted that when The larger the The degree of disorder in the distribution of reference directions corresponding to the target edges within a target cluster often indicates that The more chaotic the distribution of the direction of the target edge in the target cluster is, the more likely it is that the The texture of the tissue structure represented by the target cluster is relatively denser. The larger the The longer the length of the straight line segment corresponding to the edge of the target cluster, the The longer the edge of the target in the target cluster is, the The more likely the edge of an object in a cluster is to represent the real texture. The smaller the time, the more likely it is that The first target cluster The smaller the target angle and target distance corresponding to the first target edge, the smaller the target angle and target distance corresponding to the first target edge. The closer the distance between the edge of a target and other edges, and the smaller the change in direction, the denser the texture. The larger the The texture of the tissue structure represented by the target cluster is relatively denser.

[0118] Step S4, determining the image information coefficient corresponding to the cardiovascular image according to the differences between the structural texture coefficients corresponding to all target clusters.

[0119] It should be noted that the quantized image information coefficient can characterize the contrast between vascular structure regions when noise exists.

[0120] As an example, this step may include the following steps:

[0121] In the first step, the absolute value of the difference between the structural texture coefficients corresponding to each two target clusters is determined as the target difference between each two target clusters.

[0122] In the second step, the mean of the target differences between all target clusters is determined as the image information coefficient corresponding to the cardiovascular image.

[0123] For example, the formula for determining the image information coefficient corresponding to the cardiovascular image can be:

[0124] ;

[0125] in, is the image information coefficient corresponding to cardiovascular images. is the number of target clusters. and is the serial number of different target clusters, and the pairwise combinations of different target clusters can be indivual. It is the absolute value function. It is The structural texture coefficients corresponding to the target clusters. It is The structural texture coefficients corresponding to the target clusters. It is The target clusters and The target differences between the target clusters.

[0126] It should be noted that when The larger the The texture of the tissue structure represented by the target cluster is relatively denser. The larger the The texture of the tissue structure represented by the target cluster is relatively denser. The larger the The target clusters and The greater the structural texture difference between the target clusters. It can characterize the contrast between vascular structure areas when noise exists. The larger the value, the greater the difference in structural texture between different target clusters. It often means that the contrast between different target clusters is more obvious, and different tissue structure areas can be more clearly distinguished. At this time, the structure and organization of the image can be better preserved and expressed, which often indicates that the impact of noise on cardiovascular images is smaller.

[0127] Step S5, performing morphological closing operation on the cardiovascular image and performing superpixel segmentation to obtain target superpixel blocks, and determining the reference information coefficient according to the edge pixel density corresponding to all target superpixel blocks.

[0128] The edge pixel density can be represented by the ratio of edge pixels in the target superpixel block. The reference information coefficient can represent the contrast between vascular structure regions after noise suppression.

[0129] It should be noted that in order to roughly reflect the situation when the image is not affected by speckle noise, the closing operation of morphological analysis can be used to roughly smooth out some noise textures in the image caused by speckle noise, and make some pixels in the image that originally belong to the same category roughly merged together. Therefore, different target superpixel blocks can roughly represent different tissue structures. Secondly, in cardiovascular images, the edge pixel density of each structural area of ​​the blood vessel is often different. In the presence of noise, the contrast between each structural area is often disturbed. Therefore, the purpose of the closing operation is to close the structural area with a larger edge pixel density, partially close and partially open the structural area with a medium edge pixel density, and keep the structural area with a weaker edge pixel density unchanged. In this way, the influence of noise on each structural area can be suppressed, thereby reducing the difference within each structural area and improving the contrast between each structural area. After the image is closed, the influence of noise is suppressed, and several target superpixel blocks are obtained according to superpixel segmentation. Since the difference within each structural area is reduced and the contrast between each structural area is improved, the contrast between several target superpixel blocks is more significant, that is, the division of each structural area is more significant. Therefore, the larger the quantized reference information coefficient, the greater the contrast between different tissue structural areas displayed.

[0130] As an example, this step may include the following steps:

[0131] In the first step, after performing morphological closing operation on the cardiovascular image, superpixel segmentation is continued, and the segmented superpixel block is used as the target superpixel block.

[0132] In the second step, the absolute value of the difference between the edge pixel densities corresponding to each two target superpixel blocks is determined as the reference difference between each two target superpixel blocks.

[0133] In the third step, the mean of the reference differences between all target superpixel blocks is determined as the reference information coefficient.

[0134] For example, the formula for determining the reference information coefficient may be:

[0135] ;

[0136] in, is the image information coefficient corresponding to cardiovascular images. is the number of target clusters. and is the serial number of different target clusters, and the pairwise combinations of different target clusters can be indivual. It is the absolute value function. It is The structural texture coefficients corresponding to the target clusters. It is The structural texture coefficients corresponding to the target clusters. It is The target clusters and The target differences between the target clusters.

[0137] It should be noted that when The larger the The texture of the tissue structure represented by the target cluster is relatively denser. The larger the The texture of the tissue structure represented by the target cluster is relatively denser. The larger the The target clusters and The greater the structural texture difference between the target clusters. It can characterize the contrast between vascular structural areas when noise exists. The larger the value, the greater the difference in structural texture between different target clusters. It often means that the contrast between different target clusters is more obvious, and different tissue structural areas can be more clearly distinguished. At this time, the structure and organization of the image can be better preserved and expressed, which often means that the impact of noise on cardiovascular images is smaller.

[0138] Step S5, performing morphological closing operation on the cardiovascular image and performing superpixel segmentation to obtain target superpixel blocks, and determining the reference information coefficient according to the edge pixel density corresponding to all target superpixel blocks.

[0139] The edge pixel density can be represented by the ratio of edge pixels in the target superpixel block. The reference information coefficient can represent the contrast between vascular structure regions after noise suppression.

[0140] It should be noted that in order to roughly reflect the situation when the image is not affected by speckle noise, the closing operation of morphological analysis can be used to roughly smooth out some noise textures in the image caused by speckle noise, and make some pixels in the image that originally belong to the same category roughly merged together. Therefore, different target superpixel blocks can roughly represent different tissue structures. Secondly, in cardiovascular images, the edge pixel density of each structural area of ​​the blood vessel is often different. In the presence of noise, the contrast between each structural area is often disturbed. Therefore, the purpose of the closing operation is to close the structural area with a larger edge pixel density, partially close and partially open the structural area with a medium edge pixel density, and keep the structural area with a weaker edge pixel density unchanged. In this way, the influence of noise on each structural area can be suppressed, thereby reducing the difference within each structural area and improving the contrast between each structural area. After the image is closed, the influence of noise is suppressed, and several target superpixel blocks are obtained according to superpixel segmentation. Since the difference within each structural area is reduced and the contrast between each structural area is improved, the contrast between several target superpixel blocks is more significant, that is, the division of each structural area is more significant. Therefore, the larger the quantized reference information coefficient, the greater the contrast between different tissue structural areas displayed.

[0141] As an example, this step may include the following steps:

[0142] In the first step, after performing morphological closing operation on the cardiovascular image, superpixel segmentation is continued, and the segmented superpixel block is used as the target superpixel block.

[0143] In the second step, the absolute value of the difference between the edge pixel densities corresponding to each two target superpixel blocks is determined as the reference difference between each two target superpixel blocks.

[0144] In the third step, the mean of the reference differences between all target superpixel blocks is determined as the reference information coefficient.

[0145] For example, the formula for determining the reference information coefficient may be:

[0146] ;

[0147] in, is the reference information coefficient. is the number of target superpixel patches. and is the serial number of different target superpixel blocks, and the two combinations of different target superpixel blocks can have indivual. It is the absolute value function. It is The edge pixel density corresponding to the target superpixel block, that is, The edge pixel point in the target superpixel block is The proportion of target superpixel blocks. It is The edge pixel density corresponding to the target superpixel block, that is, The edge pixel in the target superpixel block is The proportion of target superpixel blocks. It is The target superpixel patch and The reference difference between the target superpixel patches.

[0148] It should be noted that when The larger the The larger the proportion of edge pixels in the target superpixel block, the larger the The larger the The larger the proportion of edge pixels in the target superpixel block, the larger the The larger the The target superpixel patch and The greater the difference in the proportion of edge pixels between target superpixel blocks, It can characterize the contrast between vascular structure areas after noise suppression. The larger its value is, the greater the difference in the proportion of edge pixels between different target superpixel blocks is, which often indicates that the contrast between different tissue structure areas displayed is greater.

[0149] Step S6, determining the target information difference according to the image information coefficient and the reference information coefficient, taking the target information difference as the scale factor of Wiener filtering, performing Wiener filtering on the cardiovascular image to obtain the target image, and identifying the target image.

[0150] As an example, this step may include the following steps:

[0151] The first step is to determine the formula corresponding to the target information difference:

[0152] ;

[0153] in, It is the target information gap. It is the absolute value function. is the reference information coefficient. is the image information coefficient corresponding to cardiovascular images.

[0154] It should be noted that The contrast between vascular structure regions after noise suppression can be characterized. It can characterize the contrast between vascular structure regions in the presence of noise. It can characterize the interference effect of noise on the vascular structure. The larger its value is, the larger the proportional factor required for subsequent Wiener filtering.

[0155] In the second step, the cardiovascular image is subjected to Wiener filtering, and the cardiovascular image after Wiener filtering is used as the target image. In the process of performing Wiener filtering on the cardiovascular image, the target information difference is used as the scale factor of the Wiener filtering.

[0156] The third step is to identify the target image.

[0157] For example, neural networks can be used to identify tissue structures in target images.

[0158] For another example, the tissue structure in the target image may be segmented by threshold segmentation to achieve recognition of the target image.

[0159] For another example, the tissue structure in the target image can be identified through doctor observation.

[0160] Based on the same inventive concept as the above method embodiment, the present invention provides an OCT-based cardiovascular image recognition device, comprising a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device performs any of the above OCT-based cardiovascular image recognition methods.

[0161] refer to Figure 4 Based on the same inventive concept as the above method embodiment, the present invention provides a cardiovascular image recognition system based on OCT, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the above computer program is executed by the processor, the steps of implementing a cardiovascular image recognition method based on OCT may specifically include:

[0162] An acquisition processing module 401 acquires a cardiovascular image based on OCT, and performs a light and dark distribution analysis process on each pixel in the cardiovascular image to obtain a light and dark coefficient corresponding to each pixel;

[0163] The cluster detection module 402 is used to cluster the pixels in the cardiovascular image according to the brightness coefficients corresponding to the pixels to obtain target clusters, and perform edge detection on each target cluster to obtain target edges;

[0164] The structural texture analysis module 403 is used to perform structural texture analysis on all target edges in each target cluster to obtain a structural texture coefficient corresponding to each target cluster;

[0165] An information coefficient determination module 404 is used to determine the image information coefficient corresponding to the cardiovascular image according to the difference between the structural texture coefficients corresponding to all target clusters;

[0166] The operation segmentation determination module 405 is used to perform morphological closing operation on the cardiovascular image and perform superpixel segmentation to obtain a target superpixel block, and determine a reference information coefficient according to the edge pixel density corresponding to all target superpixel blocks, wherein the edge pixel density represents the proportion of edge pixel points in the target superpixel block;

[0167] The determination filter identification module 406 is used to determine the target information difference according to the image information coefficient and the reference information coefficient, use the target information difference as the scale factor of the Wiener filter, perform Wiener filtering on the cardiovascular image, obtain the target image, and identify the above target image.

[0168] Figure 5 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. Figure 5 As shown, the computer device 500 includes: a memory 501, a processor 502, and a computer program 503 stored in the memory 501 and running on the processor 502, wherein when the processor 502 executes the computer program 503, the computer device can execute any one of the OCT-based cardiovascular image recognition methods introduced above.

[0169] Based on the same inventive concept as the above method embodiment, the present invention provides a computer program product, which includes: computer program code, when the computer program code runs on a computer, the computer executes any one of the above OCT-based cardiovascular image recognition methods.

[0170] Based on the same inventive concept as the above-mentioned method embodiment, the present invention provides a computer-readable storage medium, which stores a computer program code. When the computer program code runs on a computer, the computer executes any one of the above-mentioned OCT-based cardiovascular image recognition methods.

[0171] In summary, compared with the traditional filtering method and the use of the image signal-to-noise ratio to set the scale factor of the Wiener filter, the present invention uses the target information difference as the scale factor of the Wiener filter and performs Wiener filtering on the cardiovascular image, which can improve the filtering effect, thereby improving the effect of denoising the cardiovascular image, and further improving the effect of identifying the cardiovascular image. Among them, the traditional filtering method can be a median filter, a maximum filter or a mean filter. Traditional filtering methods often have certain limitations when processing speckled noise. It is often difficult to correctly distinguish between speckled noise and the texture of the cardiovascular tissue structure, resulting in the non-smooth texture on the cardiovascular tissue structure being filtered out. Speckled noise is also called speckle noise.

[0172] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A cardiovascular image recognition method based on OCT, characterized in that: The following steps are involved: Obtain cardiovascular images based on OCT, and perform light and dark distribution analysis on each pixel in the cardiovascular images to obtain the light and dark coefficient corresponding to each pixel; According to the brightness coefficients corresponding to the pixels, the pixels in the cardiovascular image are clustered to obtain target clusters, and edge detection is performed on each target cluster to obtain the target edge; Perform structural texture analysis on all target edges in each target cluster to obtain the structural texture coefficient corresponding to each target cluster; According to the differences between the structural texture coefficients corresponding to all target clusters, the image information coefficient corresponding to the cardiovascular image is determined; Perform morphological closing operation on cardiovascular images and perform superpixel segmentation to obtain target superpixel blocks. Determine the reference information coefficient based on the edge pixel density corresponding to all target superpixel blocks, where the edge pixel density is represented by the proportion of edge pixels in the target superpixel block. Determine a target information difference according to the image information coefficient and the reference information coefficient, use the target information difference as a scale factor of Wiener filtering, perform Wiener filtering on the cardiovascular image to obtain a target image, and identify the target image; The formula corresponding to the target information difference is: ； in, It is the target information gap; It is the absolute value function; is the reference information coefficient; is the image information coefficient corresponding to cardiovascular images; The light and dark distribution analysis and processing of each pixel in the cardiovascular image is performed to obtain the light and dark coefficient corresponding to each pixel, including: Filtering out target maximum points from the cardiovascular image, and determining the brightness coefficient corresponding to each pixel in the cardiovascular image according to the grayscale brightness distribution between all target maximum points and their corresponding reference points in a second preset neighborhood corresponding to each pixel in the cardiovascular image, wherein the reference point corresponding to the target maximum point is the target maximum point closest to the target maximum point, and the second preset neighborhood is a pre-set neighborhood; Determining the image information coefficient corresponding to the cardiovascular image according to the difference between the structural texture coefficients corresponding to all target clusters includes: The absolute value of the difference between the structural texture coefficients corresponding to each two target clusters is determined as the target difference between each two target clusters; The mean of the target differences between all target clusters is determined as the image information coefficient corresponding to the cardiovascular image; The step of determining the reference information coefficient according to the edge pixel densities corresponding to all target superpixel blocks includes: The absolute value of the difference between the edge pixel densities corresponding to each two target super-pixel blocks is determined as the reference difference between each two target super-pixel blocks; The mean of the reference differences between all target superpixel blocks is determined as the reference information coefficient.

2. The method for cardiovascular image recognition based on OCT according to claim 1, characterized in that: The method of screening out the target maximum point from the cardiovascular image includes: Any pixel point in the cardiovascular image is determined as a marked pixel point. If the grayscale value corresponding to the marked pixel point is greater than the grayscale values ​​corresponding to each pixel point in its first preset neighborhood, the marked pixel point is determined as the target maximum point, wherein the first preset neighborhood is a pre-set neighborhood, and the area of ​​the second preset neighborhood is greater than the area of ​​the first preset neighborhood.

3. The method for cardiovascular image recognition based on OCT according to claim 1, characterized in that: Determining the shading coefficient corresponding to each pixel point in the cardiovascular image according to the grayscale shading distribution between all target maximum points and their corresponding reference points in the second preset neighborhood corresponding to each pixel point in the cardiovascular image includes: Determine any pixel point in the cardiovascular image as a marked pixel point, and determine any target maximum point in a second preset neighborhood corresponding to the marked pixel point as a candidate maximum point; Taking the candidate maximum point and its corresponding reference point as endpoints, a target connecting line segment corresponding to the candidate maximum point is drawn; Determine the mean of the grayscale values ​​corresponding to all non-maximum points on the target line segment corresponding to the candidate maximum point as the dark grayscale factor corresponding to the candidate maximum point, wherein the non-maximum point is a pixel point in the cardiovascular image other than the target maximum point; Determine the average of the grayscale value corresponding to the candidate maximum point and the grayscale value corresponding to the corresponding reference point as the bright grayscale factor corresponding to the candidate maximum point; Determine the difference between the bright grayscale factor and the dark grayscale factor corresponding to the candidate maximum point as the initial bright and dark factor corresponding to the candidate maximum point; Determine the distance between the candidate maximum point and its corresponding reference point as the reference distance corresponding to the candidate maximum point; Determine the brightness coefficient corresponding to the marked pixel point according to the initial brightness factor and the reference distance corresponding to all target maximum points in the second preset neighborhood corresponding to the marked pixel point, wherein the initial brightness factor is positively correlated with the brightness coefficient, and the reference distance is negatively correlated with the brightness coefficient; The formula for the brightness coefficient of the pixel points in cardiovascular images is: ； in, It is the first The brightness coefficient corresponding to each pixel; is the serial number of the pixel in the cardiovascular image; It is the first The number of target maximum points in the second preset neighborhood corresponding to the pixel points; It is the first The sequence number of the target maximum point in the second preset neighborhood corresponding to the pixel point; It is an exponential function with a natural constant as base; It is the first In the second preset neighborhood corresponding to the pixel point, The dark grayscale factor corresponding to the target maximum point; It is the first In the second preset neighborhood corresponding to the pixel point, Brightness grayscale factor corresponding to the maximum point of the target; It is The initial brightness factor corresponding to the target maximum point; It is the first In the second preset neighborhood corresponding to the pixel point, The reference distance corresponding to the target maximum point.

4. The method for cardiovascular image recognition based on OCT according to claim 1, characterized in that: The structural texture analysis is performed on all target edges in each target cluster to obtain the structural texture coefficient corresponding to each target cluster, including: Connecting two ends of each target edge to obtain a straight line segment corresponding to each target edge, and determining a straight line direction corresponding to the straight line segment corresponding to each target edge as a reference direction corresponding to each target edge; The information entropy of the reference direction corresponding to all target edges in each target cluster is determined as the target information entropy corresponding to each target cluster; The average length of the straight line segments corresponding to all target edges in each target cluster is determined as the average target length corresponding to each target cluster; The structural texture coefficient corresponding to each target cluster is determined according to the minimum distance between target edges in each target cluster, the angle between the straight line segments corresponding to the target edges in each target cluster, the target information entropy corresponding to each target cluster and the target average length.

5. The method for cardiovascular image recognition based on OCT according to claim 4, characterized in that: The method of determining the structural texture coefficient corresponding to each target cluster according to the minimum distance between target edges in each target cluster, the angle between the straight line segments corresponding to the target edges in each target cluster, the target information entropy corresponding to each target cluster, and the target average length includes: Determine any target cluster as a marked cluster, determine any target edge in the marked cluster as a marked edge, and determine the target edge closest to the marked edge as a reference edge; Determine the minimum distance between the marked edge and the reference edge as the target distance corresponding to the marked edge; Determine the angle between the straight line segment corresponding to the marked edge and the straight line segment corresponding to the reference edge as the target angle corresponding to the marked edge; Determining a substructure feature corresponding to the marked edge according to a target distance and a target angle corresponding to the marked edge, wherein both the target distance and the target angle are negatively correlated with the substructure feature; Determine the structural texture coefficient corresponding to the marked cluster according to the target information entropy and target average length corresponding to the marked cluster, and the substructure features corresponding to all target edges in the marked cluster, wherein the target information entropy, target average length and substructure features are all positively correlated with the structural texture coefficient; The formula corresponding to the structural texture coefficient of the target cluster is: ； in, It is The structural texture coefficient corresponding to the target clusters; is the serial number of the target cluster; It is a hyperbolic function, mainly used for normalization; It is The target information entropy corresponding to the target clusters; It is The average length of the targets corresponding to the target clusters; It is The number of target edges in a target cluster; It is The serial number of the target edge in the target cluster; It is The first target cluster The target angle corresponding to the target edge; It is The first target cluster The target distance corresponding to the target edge; is a pre-set factor greater than 0; It is The first target cluster The substructure features corresponding to the target edges.

6. A cardiovascular image recognition device based on OCT, characterized in that: It comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement an OCT-based cardiovascular image recognition method according to any one of claims 1 to 5.

7. A cardiovascular image recognition system based on OCT, characterized in that: It is used to implement a cardiovascular image recognition method based on OCT according to any one of claims 1 to 5, specifically comprising: An acquisition processing module acquires cardiovascular images based on OCT, and performs light and dark distribution analysis on each pixel in the cardiovascular images to obtain a light and dark coefficient corresponding to each pixel; A cluster detection module is used to cluster the pixels in the cardiovascular image according to the brightness coefficients corresponding to the pixels to obtain target clusters, and to perform edge detection on each target cluster to obtain target edges; A structural texture analysis module is used to perform structural texture analysis on all target edges in each target cluster to obtain a structural texture coefficient corresponding to each target cluster; An information coefficient determination module is used to determine the image information coefficient corresponding to the cardiovascular image according to the difference between the structural texture coefficients corresponding to all target clusters; An operation segmentation determination module is used to perform morphological closing operations on cardiovascular images and perform superpixel segmentation to obtain target superpixel blocks, and determine a reference information coefficient according to the edge pixel density corresponding to all target superpixel blocks, wherein the edge pixel density represents the proportion of edge pixel points in the target superpixel block; A filter identification module is determined, which is used to determine the target information difference according to the image information coefficient and the reference information coefficient, use the target information difference as the scale factor of the Wiener filter, perform Wiener filtering on the cardiovascular image, obtain the target image, and identify the target image.

Images