Method and system for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence

Abstract

The present invention relates to the field of image enhancement technology, and in particular to a method and system for detecting cardiac occupancy in ultrasonic images based on artificial intelligence, the method comprising: obtaining an ultrasonic cardiac image to be enhanced, mapping each preset direction based on the pixel values ​​corresponding to the central window and each neighborhood window corresponding to the pixel point; determining the artifact change interference factor between the pixel point and each preset direction of the central window and each neighborhood window corresponding to the pixel point; determining the artifact interference influence index corresponding to the pixel point, and clustering the pixel points in the initial image; performing regional segmentation on the initial image, determining the overall artifact interference factor corresponding to each sub-block; adjusting the preset scale parameter, and enhancing through the Retinex algorithm, and performing cardiac occupancy detection on the target image. The present invention realizes the enhancement of the ultrasonic cardiac image to be enhanced, improves the image enhancement effect, and thus improves the accuracy of cardiac occupancy detection.

Description

Technical Field

[0001] The present invention relates to the technical field of image enhancement, and in particular to a method and system for detecting cardiac space-occupying in ultrasonic images based on artificial intelligence. Background Art

[0002] With the development of science and technology, the application of image enhancement technology is becoming more and more extensive. For example, it can be applied to the enhancement of echocardiography in cardiac space-occupying detection. At present, when enhancing an image, the commonly used method is to enhance the image by using grayscale histogram equalization.

[0003] However, when using grayscale histogram equalization to enhance echocardiography, the following technical problems often occur:

[0004] Due to the limitations of the imaging mechanism of echocardiography, artifact interference often exists in echocardiography, and grayscale histogram equalization often performs statistical image enhancement according to the grayscale value distribution of the image, which may cause artifact interference in echocardiography to be over-enhanced, resulting in poor enhancement effect of echocardiography, and further resulting in poor accuracy of cardiac space-occupying detection. Summary of the invention

[0005] In order to solve the technical problem of poor accuracy of cardiac space-occupying detection due to poor enhancement effect of echocardiography, the present invention proposes a method and system for detecting cardiac space-occupying in ultrasound images based on artificial intelligence.

[0006] In a first aspect, the present invention provides a method for detecting cardiac space-occupying in an ultrasound image based on artificial intelligence, the method comprising:

[0007] Acquire an ultrasound cardiac image to be enhanced as an initial image, and perform mapping in each preset direction based on the pixel values ​​corresponding to the pre-acquired central window and each neighborhood window corresponding to the pixel points in the initial image, to obtain a mapping sequence of the pixel points in each preset direction in the corresponding central window and each neighborhood window;

[0008] Determine, according to a mapping sequence of the pixel point in each preset direction of its corresponding central window and each neighborhood window, an artifact change interference factor between the pixel point in each preset direction of its corresponding central window and each neighborhood window;

[0009] According to the artifact change interference factor between the pixel point in all preset directions of its corresponding central window and all neighboring windows, the artifact interference impact index corresponding to the pixel point is determined, and based on the pixel value corresponding to the pixel point and the artifact interference impact index, the pixel points in the initial image are clustered to obtain the target cluster;

[0010] Performing regional segmentation on the initial image to obtain sub-blocks, and determining the overall artifact interference factor corresponding to each sub-block according to the number of pixels belonging to different target clusters in each sub-block and the artifact interference impact index corresponding to the pixels belonging to different target clusters in each sub-block;

[0011] According to the overall artifact interference factor corresponding to each sub-block, the preset scale parameter is adjusted to obtain the target scale parameter corresponding to each sub-block;

[0012] According to the target scale parameters corresponding to different sub-blocks in the initial image, the Retinex algorithm is used to enhance the target image, and the target image is then subjected to cardiac space occupancy detection.

[0013] In combination with the first aspect, in a possible implementation, determining the artifact change interference factor between the pixel point in each preset direction of the corresponding center window and each neighborhood window according to the mapping sequence of the pixel point in each preset direction of the corresponding center window and each neighborhood window includes:

[0014] Determine any pixel point in the initial image as a marked pixel point, determine any neighborhood window corresponding to the marked pixel point as a candidate window, and determine any preset direction as a marked direction;

[0015] Determine the Manhattan distance between the mapping sequence of the marked pixel point in the marking direction of the corresponding central window and the mapping sequence of the marked pixel point in the marking direction of the candidate window as the first change difference between the marked pixel point in the marking direction of the corresponding central window and the candidate window;

[0016] Determine the difference between each adjacent element in the mapping sequence of the marked pixel point in the marking direction of the corresponding central window as a mapping difference, and obtain the mapping difference sequence of the marked pixel point in the marking direction of the corresponding central window;

[0017] Similarly, a mapping difference sequence of the marked pixel point in the marking direction of the candidate window is obtained;

[0018] Determine the absolute value of the difference between the mean of all elements in the mapping difference sequence of the marked pixel point in the marking direction of the corresponding central window and the mean of all elements in the mapping difference sequence of the marked pixel point in the marking direction of the candidate window as the second change difference between the marked pixel point in the marking direction of the corresponding central window and the candidate window;

[0019] Based on the first change difference and the second change difference between the marking direction of the marked pixel point in its corresponding center window and the candidate window, the artifact change interference factor between the marked pixel point in its corresponding center window and the candidate window is determined, wherein the first change difference and the second change difference are positively correlated with the artifact change interference factor.

[0020] In combination with the first aspect above, in a possible implementation, determining the artifact interference impact index corresponding to the pixel point according to the artifact change interference factor between the pixel point and all preset directions of the corresponding central window and all neighborhood windows includes:

[0021] The artifact change interference factors of the pixel point between all preset directions of the corresponding central window and each neighborhood window are used to form an artifact change interference factor sequence of the pixel point in each corresponding neighborhood window;

[0022] According to the artifact change interference factor sequence of the pixel point in all the corresponding neighborhood windows, the artifact interference influence index corresponding to the pixel point is determined.

[0023] In combination with the first aspect above, in a possible implementation manner, the formula corresponding to the artifact interference impact index corresponding to the pixel point is:

[0024] ;in, is the artifact interference impact index corresponding to the i-th pixel in the initial image; i is the sequence number of the pixel in the initial image; is the number of neighborhood windows corresponding to the i-th pixel in the initial image; j and b are the serial numbers of the neighborhood windows corresponding to the i-th pixel in the initial image; It is the mean value of all artifact change interference factors in the artifact change interference factor sequence of the i-th pixel in its corresponding j-th neighborhood window in the initial image; is the mean value of all artifact change interference factors in the artifact change interference factor sequence of the i-th pixel in its corresponding b-th neighborhood window in the initial image; yes and The DTW distance between them; is the artifact change interference factor sequence of the i-th pixel in the initial image under its corresponding j-th neighborhood window; It is the sequence of artifact change interference factors of the i-th pixel in the initial image under its corresponding b-th neighborhood window.

[0025] In combination with the first aspect, in a possible implementation manner, segmenting the initial image into regions to obtain sub-blocks includes:

[0026] Randomly selecting a preset number of pixel points from the arc edge of the initial image as reference pixel points, and determining the vertices of the initial image as target vertices;

[0027] Connecting the target vertex with each reference pixel point to obtain a reference line segment, and dividing each reference line segment into equal parts to obtain a segmentation point sequence corresponding to each reference line segment;

[0028] Perform curve fitting on the segmentation points with the same sequence number in the segmentation point sequence corresponding to all reference line segments to obtain the target fitting curve;

[0029] The initial image is segmented using each target fitting curve as a segmentation line to obtain a plurality of sub-blocks.

[0030] In combination with the first aspect above, in a possible implementation manner, the formula corresponding to the overall artifact interference factor corresponding to the sub-block is:

[0031] ;in, is the overall artifact interference factor corresponding to the x-th sub-block; x is the sub-block number; is the total number of target clusters to which all pixels in the x-th sub-block belong; r is the serial number of the target cluster to which the pixels in the x-th sub-block belong; is the number of pixels in the x-th sub-block; is the number of pixels belonging to the rth target cluster in the xth sub-block; It is the mean value of the artifact interference impact index corresponding to all pixels belonging to the rth target cluster in the xth sub-block.

[0032] In combination with the first aspect above, in a possible implementation manner, a formula corresponding to a target scale parameter corresponding to a sub-block is:

[0033] ;in, is the target scale parameter corresponding to the x-th sub-block; x is the sub-block number; is the preset scale parameter; f is the preset tuning constant; is the overall artifact interference factor corresponding to the x-th sub-block; p is the cumulative value of the overall artifact interference factors corresponding to all sub-blocks.

[0034] In combination with the first aspect, in a possible implementation manner, clustering the pixels in the initial image based on the pixel values ​​corresponding to the pixels and the artifact interference impact index to obtain the target cluster includes:

[0035] The pixel value corresponding to the pixel point and the artifact interference impact index are used to form the target coordinates corresponding to the pixel point;

[0036] According to the target coordinates corresponding to all the pixels in the initial image, the pixels in the initial image are clustered by a density peak clustering algorithm, and each obtained cluster is determined as a target cluster.

[0037] In combination with the first aspect, in a possible implementation manner, obtaining a central window and each neighborhood window corresponding to a pixel point includes:

[0038] Determine any pixel point in the initial image as a marked pixel point, and construct a window with a size equal to a first preset size with the marked pixel point as the center window corresponding to the marked pixel point;

[0039] Taking the marked pixel point as the center, constructing a window with a size equal to a second preset size as the overall neighborhood window corresponding to the marked pixel point, wherein the second preset size is larger than the first preset size;

[0040] A sliding window having a size equal to a first preset size is constructed as a target sliding window. The target sliding window is slid from top to bottom and from left to right within the overall neighborhood window corresponding to the marked pixel point with a preset step size, and the target sliding window after each sliding is determined as a neighborhood window corresponding to the marked pixel point, wherein the union of all neighborhood windows corresponding to the marked pixel point is the overall neighborhood window corresponding to the marked pixel point.

[0041] In a second aspect, the present invention provides a cardiac space-occupying detection system in an ultrasound image based on artificial intelligence, the system comprising:

[0042] An acquisition mapping module is used to acquire the ultrasound cardiac image to be enhanced as an initial image, and perform mapping in each preset direction based on the pixel values ​​corresponding to the pre-acquired central window and each neighborhood window corresponding to the pixel points in the initial image, so as to obtain a mapping sequence of the pixel points in each preset direction in the corresponding central window and each neighborhood window;

[0043] An artifact change interference factor determination module is used to determine an artifact change interference factor between a pixel point in each preset direction of its corresponding central window and each neighborhood window according to a mapping sequence of the pixel point in each preset direction of its corresponding central window and each neighborhood window;

[0044] An index determination clustering module is used to determine the artifact interference impact index corresponding to the pixel point according to the artifact change interference factor between the pixel point's corresponding central window and all preset directions of all neighborhood windows, and cluster the pixels in the initial image based on the pixel value corresponding to the pixel point and the artifact interference impact index to obtain a target cluster;

[0045] The overall artifact interference factor determination module is used to perform regional segmentation on the initial image to obtain sub-blocks, and determine the overall artifact interference factor corresponding to each sub-block according to the number of pixels belonging to different target clusters in each sub-block and the artifact interference impact index corresponding to the pixels belonging to different target clusters in each sub-block;

[0046] A scale parameter adjustment module is used to adjust the preset scale parameter according to the overall artifact interference factor corresponding to each sub-block to obtain a target scale parameter corresponding to each sub-block;

[0047] The image enhancement detection module is used to enhance the target image by using the Retinex algorithm according to the target scale parameters corresponding to different sub-blocks in the initial image, and then perform cardiac occupancy detection on the target image.

[0048] In a third aspect, a server 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.

[0049] 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.

[0050] 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.

[0051] The present invention has the following beneficial effects:

[0052] The method for detecting cardiac space-occupancy in ultrasonic images based on artificial intelligence of the present invention realizes the enhancement of ultrasonic cardiac images to be enhanced, solves the technical problem of poor accuracy of cardiac space-occupancy detection due to poor image enhancement effect, improves the image enhancement effect, and thus improves the accuracy of cardiac space-occupancy detection. Compared with the use of grayscale histogram equalization to perform image enhancement on ultrasonic cardiac images to be enhanced, the present invention comprehensively considers multiple factors related to artifact interference, such as artifact variation interference factor, artifact interference influence index and overall artifact interference factor, so as to adjust the preset scale parameters, and perform image enhancement based on the adjusted scale parameters, so as to realize the enhancement of ultrasonic cardiac images to be enhanced, and reduce the influence of image artifact interference to a certain extent, so as to improve the image enhancement effect, and then improve the accuracy of cardiac space-occupancy detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] 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.

[0054] Figure 1 It is a flow chart of the method for detecting cardiac space-occupying parts in ultrasonic images based on artificial intelligence of the present invention;

[0055] Figure 2 is a schematic diagram of an initial image of the present invention;

[0056] Figure 3 It is a schematic diagram of sub-block segmentation of the present invention;

[0057] Figure 4 It is a schematic diagram of image enhancement comparison of the present invention;

[0058] Figure 5 It is a schematic diagram of the composition structure of the cardiac space-occupying detection system in ultrasound images based on artificial intelligence of the present invention;

[0059] Figure 6 The figure is a schematic diagram of the structure of a computer device of the present invention.

[0060] The reference numerals include: a first region 1 , a second region 2 , a third region 3 , a fourth region 4 , a fifth region 5 , a sixth region 6 , a first image 401 , a second image 402 , and a third image 403 . DETAILED DESCRIPTION

[0061] 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.

[0062] 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.

[0063] refer to Figure 1, showing the process of some embodiments of the method for detecting cardiac space-occupying in ultrasound images based on artificial intelligence of the present invention. The method for detecting cardiac space-occupying in ultrasound images based on artificial intelligence comprises the following steps:

[0064] Step S1, obtain the ultrasound cardiac image to be enhanced as the initial image, and map each preset direction based on the pre-acquired central window corresponding to the pixel points in the initial image and the pixel values ​​corresponding to the pixel points in each neighborhood window, to obtain a mapping sequence of the pixel points in each preset direction in the corresponding central window and each neighborhood window.

[0065] Among them, the ultrasound image to be enhanced may be a denoised ultrasound cardiogram to be enhanced. An ultrasound cardiogram is also called a two-dimensional ultrasound cardiogram or a cross-sectional ultrasound cardiogram. The size of the center window and the neighborhood window may be the same. The pixel point may be located at the center position of its corresponding center window. The neighborhood window corresponding to the pixel point may characterize the area near the pixel point. The preset direction may be a pre-set direction for mapping. For example, there may be 3 preset directions, which may be: 0° direction, 45° direction and 90° direction.

[0066] It should be noted that in cardiac space-occupancy detection, echocardiography is often used, which forms a cross-sectional image by combining the echo signals reflected by the human body in the form of light spots, which often contains the morphology, spatial position and continuous relationship of various cardiac structures. However, due to the limitations of the imaging mechanism of echocardiography, there is often a large amount of speckle noise in the ultrasonic image, and there is often artifact interference in the collected image, which has an adverse effect on the analysis and identification of detailed features in the ultrasonic image. Therefore, the embodiment of the present invention is mainly used to denoise and enhance the collected echocardiogram, which can reduce the influence of noise and artifact interference to a certain extent, thereby improving the image quality for doctors to observe.

[0067] As an example, this step may include the following steps:

[0068] The first step is to obtain an ultrasound cardiac image to be enhanced as an initial image.

[0069] For example, an ultrasound cardiogram of a patient may be obtained, and the ultrasound cardiogram may be denoised using a non-local mean denoising algorithm, and the obtained denoised image may be recorded as an ultrasound cardiogram to be enhanced, that is, an initial image.

[0070] It should be noted that the initial image can represent the captured heart area. Figure 2 The denoised echocardiogram shown is recorded as the initial image. The initial image at this time can be used Figure 2 Secondly, denoising the echocardiogram can reduce the impact of image noise to a certain extent.

[0071] In the second step, any pixel point in the above initial image is determined as a marked pixel point, and a window with a size equal to a first preset size is constructed with the above marked pixel point as the center window corresponding to the above marked pixel point.

[0072] The first preset size may be a preset square size. For example, the first preset size may be a 5×5 size. If the first preset size is a 5×5 size, then the center window may be a 5×5 window.

[0073] The third step is to construct a window with a size equal to the second preset size with the above-mentioned marked pixel point as the center, as the overall neighborhood window corresponding to the above-mentioned marked pixel point.

[0074] The second preset size may be a preset size larger than the first preset size. For example, the second preset size may be a size of 25×25. If the second preset size is a size of 25×25, the neighborhood overall window may be a 25×25 window.

[0075] The fourth step is to construct a sliding window with a size equal to the first preset size as the target sliding window. In the overall neighborhood window corresponding to the above-mentioned marked pixel point, the target sliding window is slid from top to bottom and from left to right with a preset step size, and the target sliding window after each sliding is determined as a neighborhood window corresponding to the above-mentioned marked pixel point.

[0076] The union of all the neighborhood windows corresponding to the above-mentioned marked pixel point may be the neighborhood overall window corresponding to the above-mentioned marked pixel point. The preset step size may be a preset step size for sliding the target sliding window, and its value may be less than or equal to the side length of the target sliding window. For example, the preset step size may be 2.

[0077] For example, if the preset step size is equal to the side length of the target sliding window, it is equivalent to dividing the overall neighborhood window corresponding to the marked pixel point into regions of a size equal to the size of the target sliding window, and determining each region obtained by the equal division as the neighborhood window corresponding to the marked pixel point.

[0078] It should be noted that artifact interference in echocardiography often causes linear changes in echo information. Therefore, constructing the central window and each neighborhood window corresponding to the pixel point in the initial image can facilitate the subsequent analysis of the impact of artifact interference on the linear change of echo information in the local area, and facilitate the subsequent analysis of the pixel distribution and linear change differences in the local area, so as to facilitate the subsequent analysis of artifact interference, wherein the local area can be represented by the neighborhood overall window.

[0079] The fifth step is to map each preset direction based on the central window corresponding to the pixel point in the initial image and the pixel value corresponding to the pixel point in each neighborhood window, and obtain the mapping sequence of the pixel point in each preset direction in its corresponding central window and each neighborhood window, which may include the following sub-steps:

[0080] In the first sub-step, the central window or any neighborhood window corresponding to the pixel point is recorded as the reference window corresponding to the pixel. The reference window is transformed in different preset directions through the Radon algorithm to obtain a set of projection data of the reference window in each preset direction, forming a projection data group of the reference window in each preset direction.

[0081] Among them, the Radon algorithm is also called Radon transform or Radon transform. Radon transform is a mathematical transform that can convert a two-dimensional image into a set of projection data in each direction.

[0082] In the second sub-step, the projection data group of the reference window corresponding to the pixel point in each preset direction is determined as a mapping sequence of the pixel point in each preset direction of the reference window.

[0083] For example, the projection data set of the reference window corresponding to the pixel point in the 0° direction is the mapping sequence of the pixel point in the 0° direction of the reference window.

[0084] It should be noted that the mapping sequence of pixel points in each preset direction of the reference window reflects the pixel distribution characteristics of the reference window in each preset direction.

[0085] Step S2, determining the artifact change interference factor between the pixel point in each preset direction of its corresponding central window and each neighborhood window according to the mapping sequence of the pixel point in each preset direction of its corresponding central window and each neighborhood window.

[0086] As an example, this step may include the following steps:

[0087] In the first step, any pixel point in the above initial image is determined as a marked pixel point, any neighborhood window corresponding to the above marked pixel point is determined as a candidate window, and any preset direction is determined as a marked direction.

[0088] In the second step, the Manhattan distance between the mapping sequence of the above-mentioned marked pixel points in the marking direction of its corresponding central window and the mapping sequence of the above-mentioned marked pixel points in the marking direction of the candidate window is determined as the first change difference between the marking directions of the above-mentioned marked pixel points in its corresponding central window and the candidate window.

[0089] The third step is to determine the difference between each adjacent element in the mapping sequence of the marked pixel point in the marking direction of the corresponding center window as the mapping difference, and obtain the mapping difference sequence of the marked pixel point in the marking direction of the corresponding center window.

[0090] The mapping difference sequence of the marked pixel point in the marking direction of the corresponding central window may include: the difference between all adjacent elements in the mapping sequence of the marked pixel point in the marking direction of the corresponding central window.

[0091] In the fourth step, similarly, a mapping difference sequence of the above-mentioned marked pixel points in the marked direction of the candidate window is obtained.

[0092] The mapping difference sequence of the marked pixel point in the marking direction of the candidate window may include: the difference between all adjacent elements in the mapping sequence of the marked pixel point in the marking direction of the candidate window.

[0093] In the fifth step, the absolute value of the difference between the mean of all elements in the mapping difference sequence of the above-mentioned marked pixel point in the marking direction of its corresponding central window and the mean of all elements in the mapping difference sequence of the above-mentioned marked pixel point in the marking direction of the candidate window is determined as the second change difference between the marking directions of the above-mentioned marked pixel point and the candidate window.

[0094] The sixth step is to determine the artifact change interference factor of the marked pixel point between the corresponding center window and the candidate window in the marking direction according to the first change difference and the second change difference between the marked pixel point and the candidate window in the marking direction.

[0095] The first change difference and the second change difference may both be positively correlated with the artifact change interference factor.

[0096] For example, the formula corresponding to the artifact change interference factor of a pixel point between the preset directions of its corresponding central window and the neighborhood window may be:

[0097] ;in, is the artifact change interference factor between the i-th pixel in the initial image and the a-th preset direction of the j-th neighborhood window. i is the number of the pixel in the initial image. j is the number of the neighborhood window corresponding to the i-th pixel in the initial image. a is the number of the preset direction. It is the absolute value function. It is the mean value of all elements in the mapping difference sequence of the i-th pixel in the a-th preset direction of the corresponding central window in the initial image. It is the mean value of all elements in the mapping difference sequence of the i-th pixel in the a-th preset direction of its corresponding j-th neighborhood window in the initial image. It is the second change difference between the i-th pixel point in the initial image in the a-th preset direction between its corresponding central window and the j-th neighborhood window. yes and The Manhattan distance between them is the first change difference between the i-th pixel point in its corresponding central window and the j-th neighborhood window in the a-th preset direction. It represents the mapping sequence of the i-th pixel point in the initial image in the a-th preset direction of its corresponding central window. It represents the mapping sequence of the i-th pixel point in the initial image in the a-th preset direction of its corresponding j-th neighborhood window.

[0098] It should be noted that artifact interference in echocardiography may cause an increase or decrease in echo information in a local area, resulting in differences in the distribution of echo information in a local area in the echocardiogram. The larger the value is, the greater the linear change difference of pixels between the central window corresponding to the i-th pixel point and the j-th neighborhood window in the a-th preset direction is. The larger the value is, the greater the difference in pixel distribution between the central window corresponding to the i-th pixel and the j-th neighborhood window in the a-th preset direction is. When it is larger, it often indicates that the pixel distribution difference and pixel linear change difference between the central window corresponding to the ith pixel point and the jth neighborhood window in the ath preset direction are larger, which often indicates that the ith pixel point is more likely to be disturbed by artifacts.

[0099] Step S3, according to the artifact change interference factor between the pixel point in all preset directions of its corresponding central window and all neighborhood windows, determine the artifact interference impact index corresponding to the pixel point, and cluster the pixels in the initial image based on the pixel value corresponding to the pixel point and the artifact interference impact index to obtain the target cluster.

[0100] As an example, this step may include the following steps:

[0101] In the first step, the artifact change interference factors of the pixel point between its corresponding central window and all preset directions of each neighborhood window are combined to form an artifact change interference factor sequence of the pixel point in each neighborhood window corresponding to it.

[0102] In the second step, according to the artifact change interference factor sequence of the pixel in all its corresponding neighborhood windows, the formula corresponding to the artifact interference impact index corresponding to the pixel can be determined as follows:

[0103] ;in, is the artifact interference impact index corresponding to the i-th pixel in the initial image. i is the sequence number of the pixel in the initial image. is the number of neighborhood windows corresponding to the ith pixel in the initial image. j and b are both the serial numbers of the neighborhood windows corresponding to the ith pixel in the initial image. It is the mean value of all artifact change interference factors in the artifact change interference factor sequence of the i-th pixel in its corresponding j-th neighborhood window in the initial image. It is the mean value of all artifact change interference factors in the artifact change interference factor sequence of the i-th pixel in its corresponding b-th neighborhood window in the initial image. yes and The DTW (Dynamic Time Warping) distance between them. It is the sequence of artifact change interference factors of the i-th pixel in the initial image under its corresponding j-th neighborhood window. It is the sequence of artifact change interference factors of the i-th pixel in the initial image under its corresponding b-th neighborhood window.

[0104] It should be noted that when When the value is larger, it often indicates that the pixel distribution difference and pixel linear change difference between the central window corresponding to the i-th pixel and the j-th neighborhood window in different preset directions are larger, which often indicates that the i-th pixel is more likely to be disturbed by artifacts. When is larger, it often indicates that the pixel distribution difference and pixel linear change difference between the central window corresponding to the i-th pixel and the b-th neighborhood window in different preset directions are larger, which often indicates that the i-th pixel is more likely to be disturbed by artifacts. The larger the value, the greater the difference in pixel value distribution and linear change characteristics after mapping in different directions in the local area where the i-th pixel is located. When it is larger, it often indicates that the difference between the pixel value distribution characteristics and the pixel linear change characteristics after mapping in different directions of the local area where the i-th pixel is located is greater, which often indicates that the characteristics of the linear change of echo information caused by artifact interference in the local area where the i-th pixel is located are more significant.

[0105] In the third step, the pixel value corresponding to the pixel point and the artifact interference impact index are combined to form the target coordinates corresponding to the pixel point.

[0106] The horizontal coordinate of the target coordinate corresponding to the pixel point may be the pixel value corresponding to the pixel point, and the vertical coordinate may be the artifact interference influence index corresponding to the pixel point.

[0107] The fourth step is to map the pixels in the initial image into a two-dimensional rectangular coordinate system according to the target coordinates corresponding to all the pixels in the initial image. The pixels in the initial image are clustered using the density peak clustering algorithm, and each cluster obtained is determined as the target cluster.

[0108] Step S4, performing regional segmentation on the initial image to obtain sub-blocks, and determining the overall artifact interference factor corresponding to each sub-block according to the number of pixels belonging to different target clusters in each sub-block and the artifact interference impact index corresponding to the pixels belonging to different target clusters in each sub-block.

[0109] As an example, this step may include the following steps:

[0110] In the first step, a preset number of pixels are randomly selected from the arc edge of the initial image as reference pixels, and the vertices of the initial image are determined as target vertices.

[0111] The preset number may be a preset number greater than or equal to 3. For example, the preset number may be 10. Figure 2 represents the initial image, then the target vertex can be Figure 2 The vertex of the fan in the middle, the arc edge of the initial image is Figure 2 The arc of the middle sector.

[0112] The second step is to connect the above target vertex with each reference pixel point to obtain a reference line segment, and divide each reference line segment into equal parts to obtain a segmentation point sequence corresponding to each reference line segment.

[0113] The number of reference line segments may be equal to the number of reference pixel points. The segmentation points in the segmentation point sequence may be equal points obtained by equally dividing the reference line segment. The number of segmentation points in different segmentation point sequences may be the same. The segmentation point sequence corresponding to the reference line segment may be obtained by sorting the segmentation points on the reference line segment to obtain the segmentation point sequence corresponding to the reference line segment, wherein the closer the segmentation point is to the target vertex, the more the segmentation point is arranged forward.

[0114] The third step is to perform curve fitting on the segmentation points with the same sequence number in the segmentation point sequence corresponding to all reference line segments to obtain the target fitting curve.

[0115] For example, curve fitting may be performed on the first segmentation point in the segmentation point sequence corresponding to all reference line segments, and the fitting curve obtained at this time is used as the first target fitting curve.

[0116] The fourth step is to segment the initial image using each target fitting curve as a segmentation line to obtain multiple sub-blocks.

[0117] For example, if Figure 3 The largest sector area shown represents the initial image. At this time, the initial image is segmented into regions to obtain 6 sub-blocks, which may be: a first region 1, a second region 2, a third region 3, a fourth region 4, a fifth region 5 and a sixth region 6; Figure 3 The arc in can represent the target fitting curve.

[0118] It should be noted that in echocardiography, there are often certain differences in image information such as contours and textures in different regions, which results in differences between high-frequency and low-frequency components in different regions. The degree of regional contrast caused by artifact interference is often different. Therefore, regional segmentation of the initial image can facilitate the subsequent adaptive enhancement of each sub-block.

[0119] In the fifth step, the formula for determining the overall artifact interference factor corresponding to the sub-block can be:

[0120] ;in, is the overall artifact interference factor corresponding to the x-th sub-block. x is the sub-block number. is the total number of target clusters to which all pixels in the x-th sub-block belong. r is the serial number of the target cluster to which the pixel in the x-th sub-block belongs. is the number of pixels in the x-th sub-block. is the number of pixels in the x-th sub-block that belong to the r-th target cluster. It is the mean value of the artifact interference impact index corresponding to all pixels belonging to the rth target cluster in the xth sub-block.

[0121] It should be noted that when When the value is larger, it often indicates that the linear change of echo information caused by artifact interference in the local area where all pixels belonging to the rth target cluster in the xth sub-block are located is more significant. Can be used as The weight of When is larger, it often means that the proportion of pixels belonging to the rth target cluster in the xth sub-block is larger, and it often means that the pixels belonging to the rth target cluster in the xth sub-block have a greater influence on the xth sub-block. When it is larger, it often indicates that the linear change characteristics of the echo information of the x-th sub-block caused by artifact interference are more significant, which often indicates that the linear change characteristics of the artifact interference have a greater impact on the detail characteristics of the x-th sub-block, which often indicates that the details of the x-th sub-block need to be enhanced.

[0122] Step S5: adjusting the preset scale parameter according to the overall artifact interference factor corresponding to each sub-block to obtain a target scale parameter corresponding to each sub-block.

[0123] The preset scale parameter may be a scale parameter for the Retinex algorithm set by manual experience, and the influence of artifacts is often not considered during the setting. For example, the preset scale parameter may be 100.

[0124] As an example, the formula for determining the target scale parameter corresponding to the sub-block may be:

[0125] ;in, is the target scale parameter corresponding to the x-th sub-block. x is the sub-block number. is a preset scale parameter. f is a preset tuning constant, that is, a preset constant for tuning. For example, f can be 20. is the overall artifact interference factor corresponding to the x-th sub-block. p is the cumulative value of the overall artifact interference factors corresponding to all sub-blocks.

[0126] It should be noted that when When it is larger, it often means that the characteristics of the linear change of the echo information of the x-th sub-block caused by artifact interference are more significant, which often means that the details of the x-th sub-block need to be enhanced more, and the corresponding scale parameter needs to be reduced. The amount of adjustment reduction of the scale parameter can be characterized. It can represent the scale parameter required for the x-th sub-block.

[0127] Step S6, according to the target scale parameters corresponding to different sub-blocks in the initial image, the target image is enhanced by using the Retinex algorithm to obtain the target image, and the heart occupancy detection is performed on the target image.

[0128] As an example, this step may include the following steps:

[0129] In the first step, according to the target scale parameters corresponding to different sub-blocks in the initial image, the Retinex (Retinex Theory, image enhancement based on the human visual system) algorithm is used to enhance the target image.

[0130] For example, the target scale parameter corresponding to each sub-block can be used as the scale parameter of the Retinex algorithm, and each sub-block can be enhanced through the Retinex algorithm, thereby achieving enhancement of all sub-blocks, and then achieving enhancement of the initial image, and the enhanced initial image is used as the target image.

[0131] For example, Figure 4As shown, if the first image 401 represents the initial image, the second image 402 can represent the image after the first image 401 is enhanced by the Retinex algorithm when the scale parameter of the Retinex algorithm is set to 100; the third image 403 can represent the image after the first image 401 is enhanced by the embodiment of the present invention, that is, the target image.

[0132] The second step is to perform heart occupancy detection on the target image.

[0133] For example, a pre-trained heart occupancy detection network can be used to perform heart occupancy detection on a target image to obtain a heart occupancy result. The heart occupancy result may include: heart occupancy position, heart occupancy size, and heart occupancy morphology. The heart occupancy detection network may be a CNN (Convolutional Neural Networks) for detecting heart occupancy.

[0134] Optionally, the training process of the cardiac occupancy detection network may include the following sub-steps:

[0135] The first sub-step is to construct a CNN as a pre-trained cardiac space-occupancy detection network.

[0136] The second sub-step is to obtain an echocardiogram set, and enhance each echocardiogram in the echocardiogram set by the method for performing image enhancement to obtain a target image recorded in steps S1 to S6, so as to obtain a target image corresponding to each echocardiogram.

[0137] In the third sub-step, the doctor observes the target image corresponding to each echocardiogram and records the heart space-occupying result corresponding to each echocardiogram.

[0138] In the fourth sub-step, the target image corresponding to the echocardiogram in the echocardiogram set is used as a training sample, and the cardiac space-occupying result corresponding to the echocardiogram in the echocardiogram set is used as a training label to train the constructed cardiac space-occupying detection network to obtain a trained cardiac space-occupying detection network, wherein the optimization algorithm in the training process can be an Adam (Adaptive momentum, random optimization of adaptive momentum) algorithm, and the loss function can be a cross entropy loss function.

[0139] Optionally, in order to improve the accuracy of cardiac space-occupying detection, 500 to-be-enhanced ultrasound cardiac images of the same patient may be collected for enhancement to obtain 500 target images, and cardiac space-occupying detection may be performed on the 500 target images so that subsequent doctors can compare and obtain accurate cardiac space-occupying results.

[0140] refer to Figure 5Based on the same inventive concept as the above method embodiment, the present invention provides a cardiac space-occupying detection system in ultrasound images based on artificial intelligence, 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 the cardiac space-occupying detection method in ultrasound images based on artificial intelligence may specifically include:

[0141] The acquisition mapping module 501 is used to acquire the ultrasound cardiac image to be enhanced as the initial image, and perform mapping in each preset direction based on the pixel values ​​corresponding to the pre-acquired central window and each neighborhood window corresponding to the pixel points in the initial image, so as to obtain a mapping sequence of the pixel points in each preset direction in the corresponding central window and each neighborhood window;

[0142] An artifact change interference factor determination module 502 is used to determine an artifact change interference factor between a pixel point in each preset direction of a corresponding central window and each neighboring window according to a mapping sequence of the pixel point in each preset direction of the corresponding central window and each neighboring window;

[0143] The index determination clustering module 503 is used to determine the artifact interference impact index corresponding to the pixel point according to the artifact change interference factor between the pixel point and all preset directions of the central window and all neighboring windows corresponding to the pixel point, and cluster the pixel points in the initial image based on the pixel value corresponding to the pixel point and the artifact interference impact index to obtain the target cluster;

[0144] The overall artifact interference factor determination module 504 is used to perform regional segmentation on the initial image to obtain sub-blocks, and determine the overall artifact interference factor corresponding to each sub-block according to the number of pixels belonging to different target clusters in each sub-block and the artifact interference impact index corresponding to the pixels belonging to different target clusters in each sub-block;

[0145] A scale parameter adjustment module 505 is used to adjust the preset scale parameter according to the overall artifact interference factor corresponding to each sub-block to obtain a target scale parameter corresponding to each sub-block;

[0146] The image enhancement detection module 506 is used to enhance the target image by using the Retinex algorithm according to the target scale parameters corresponding to different sub-blocks in the initial image, obtain the target image, and perform heart occupancy detection on the target image.

[0147] Figure 6 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. Figure 6As shown, the computer device 600 includes: a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and running on the processor 602, wherein when the processor 602 executes the computer program 603, the computer device can execute any one of the artificial intelligence-based cardiac space-occupying detection methods in ultrasound images introduced above.

[0148] Based on the same inventive concept as the above method embodiment, the present invention provides a server, including 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 any one of the above-mentioned methods for detecting cardiac space-occupying in ultrasound images based on artificial intelligence.

[0149] Based on the same inventive concept as the above-mentioned method embodiment, the present invention provides a computer program product, which includes: computer program code, when the computer program code runs on a computer, enables the computer to execute any one of the above-mentioned artificial intelligence-based cardiac space-occupying detection methods in ultrasound images.

[0150] 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 artificial intelligence-based cardiac space-occupying detection methods in ultrasound images.

[0151] In summary, compared with the use of grayscale histogram equalization to perform image enhancement on the ultrasound cardiac image to be enhanced, the present invention comprehensively considers multiple factors related to artifact interference, such as artifact change interference factor, artifact interference influence index and overall artifact interference factor, so as to adjust the preset scale parameters, and perform image enhancement based on the adjusted scale parameters, thereby achieving the enhancement of the ultrasound cardiac image to be enhanced, and reducing the influence of image artifact interference to a certain extent, thereby improving the image enhancement effect, and then improving the accuracy of cardiac space occupancy detection.

[0152] 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 method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence. It is characterized in that The following steps are involved: Acquire an ultrasound cardiac image to be enhanced as an initial image, and perform mapping in each preset direction based on the pixel values ​​corresponding to the pre-acquired central window and each neighborhood window corresponding to the pixel points in the initial image, to obtain a mapping sequence of the pixel points in each preset direction in the corresponding central window and each neighborhood window; Determine, according to a mapping sequence of the pixel point in each preset direction of its corresponding central window and each neighborhood window, an artifact change interference factor between the pixel point in each preset direction of its corresponding central window and each neighborhood window; According to the artifact change interference factor between the pixel point in all preset directions of its corresponding central window and all neighboring windows, the artifact interference impact index corresponding to the pixel point is determined, and based on the pixel value corresponding to the pixel point and the artifact interference impact index, the pixel points in the initial image are clustered to obtain the target cluster; Performing regional segmentation on the initial image to obtain sub-blocks, and determining the overall artifact interference factor corresponding to each sub-block according to the number of pixels belonging to different target clusters in each sub-block and the artifact interference impact index corresponding to the pixels belonging to different target clusters in each sub-block; According to the overall artifact interference factor corresponding to each sub-block, the preset scale parameter is adjusted to obtain the target scale parameter corresponding to each sub-block; According to the target scale parameters corresponding to different sub-blocks in the initial image, the Retinex algorithm is used to enhance the target image, and the target image is then subjected to cardiac occupancy detection. The formula corresponding to the target scale parameter of the sub-block is: ;in, is the target scale parameter corresponding to the x-th sub-block; x is the sub-block number; is the preset scale parameter; f is the preset tuning constant; is the overall artifact interference factor corresponding to the x-th sub-block; p is the cumulative value of the overall artifact interference factors corresponding to all sub-blocks; According to the target scale parameters corresponding to different sub-blocks in the initial image, the target image is enhanced by the Retinex algorithm, including: using the target scale parameter corresponding to each sub-block as the scale parameter of the Retinex algorithm, and enhancing each sub-block by the Retinex algorithm, thereby achieving the enhancement of all sub-blocks, and then achieving the enhancement of the initial image, and using the enhanced initial image as the target image.

2. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 1, It is characterized in that The method of determining the artifact change interference factor between the pixel point in each preset direction of the corresponding central window and each neighborhood window according to the mapping sequence of the pixel point in each preset direction of the corresponding central window and each neighborhood window includes: Determine any pixel point in the initial image as a marked pixel point, determine any neighborhood window corresponding to the marked pixel point as a candidate window, and determine any preset direction as a marked direction; Determine the Manhattan distance between the mapping sequence of the marked pixel point in the marking direction of the corresponding central window and the mapping sequence of the marked pixel point in the marking direction of the candidate window as the first change difference between the marked pixel point in the marking direction of the corresponding central window and the candidate window; Determine the difference between each adjacent element in the mapping sequence of the marked pixel point in the marking direction of the corresponding central window as a mapping difference, and obtain the mapping difference sequence of the marked pixel point in the marking direction of the corresponding central window; Similarly, a mapping difference sequence of the marked pixel point in the marking direction of the candidate window is obtained; Determine the absolute value of the difference between the mean of all elements in the mapping difference sequence of the marked pixel point in the marking direction of the corresponding central window and the mean of all elements in the mapping difference sequence of the marked pixel point in the marking direction of the candidate window as the second change difference between the marked pixel point in the marking direction of the corresponding central window and the candidate window; Based on the first change difference and the second change difference between the marking direction of the marked pixel point in its corresponding center window and the candidate window, the artifact change interference factor between the marked pixel point in its corresponding center window and the candidate window is determined, wherein the first change difference and the second change difference are positively correlated with the artifact change interference factor.

3. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 1, It is characterized in that The method of determining the artifact interference impact index corresponding to the pixel point according to the artifact change interference factor between the pixel point and all preset directions of the corresponding central window and all neighborhood windows includes: The artifact change interference factors of the pixel point between all preset directions of the corresponding central window and each neighborhood window are used to form an artifact change interference factor sequence of the pixel point in each corresponding neighborhood window; According to the artifact change interference factor sequence of the pixel point in all its corresponding neighborhood windows, the artifact interference influence index corresponding to the pixel point is determined.

4. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 3, It is characterized in that The formula corresponding to the artifact interference impact index corresponding to the pixel point is: ;in, is the first i The artifact interference impact index corresponding to each pixel point; i is the serial number of the pixel in the initial image; is the first i The number of neighborhood windows corresponding to the pixel points; j and b The first i The sequence number of the neighborhood window corresponding to the pixel point; is the initial image, i The pixel point corresponds to j The mean value of all artifact change interference factors in the artifact change interference factor sequence under the neighborhood window; is the initial image, i The pixel point corresponds to b The mean value of all artifact change interference factors in the artifact change interference factor sequence under the neighborhood window; yes and The DTW distance between them; is the initial image, i The pixel point corresponds to j Artifact variation interference factor sequence under neighborhood windows; is the initial image, i The pixel point corresponds to b The artifact variation interference factor sequence under the neighborhood window.

5. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 1, It is characterized in that The initial image is segmented into regions to obtain sub-blocks, including: Randomly selecting a preset number of pixel points from the arc edge of the initial image as reference pixel points, and determining the vertices of the initial image as target vertices; Connecting the target vertex with each reference pixel point to obtain a reference line segment, and dividing each reference line segment into equal parts to obtain a segmentation point sequence corresponding to each reference line segment; Perform curve fitting on the segmentation points with the same sequence number in the segmentation point sequence corresponding to all reference line segments to obtain the target fitting curve; The initial image is segmented using each target fitting curve as a segmentation line to obtain a plurality of sub-blocks.

6. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 1, It is characterized in that The formula corresponding to the overall artifact interference factor of the sub-block is: ;in, is the overall artifact interference factor corresponding to the x-th sub-block; x is the sub-block number; is the total number of target clusters to which all pixels in the x-th sub-block belong; r is the serial number of the target cluster to which the pixels in the x-th sub-block belong; is the number of pixels in the x-th sub-block; is the number of pixels belonging to the rth target cluster in the xth sub-block; It is the mean value of the artifact interference impact index corresponding to all pixels belonging to the rth target cluster in the xth sub-block.

7. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 1, It is characterized in that The pixel points in the initial image are clustered based on the pixel values ​​corresponding to the pixel points and the artifact interference impact index to obtain the target cluster, including: The pixel value corresponding to the pixel point and the artifact interference impact index are used to form the target coordinates corresponding to the pixel point; According to the target coordinates corresponding to all the pixels in the initial image, the pixels in the initial image are clustered by a density peak clustering algorithm, and each obtained cluster is determined as a target cluster.

8. The method for detecting cardiac space-occupying parts in ultrasound images based on artificial intelligence according to claim 1, It is characterized in that Get the center window and each neighborhood window corresponding to the pixel point, including: Determine any pixel point in the initial image as a marked pixel point, and construct a window with a size equal to a first preset size with the marked pixel point as the center window corresponding to the marked pixel point; Taking the marked pixel point as the center, constructing a window with a size equal to a second preset size as the overall neighborhood window corresponding to the marked pixel point, wherein the second preset size is larger than the first preset size; A sliding window having a size equal to a first preset size is constructed as a target sliding window. The target sliding window is slid from top to bottom and from left to right within the overall neighborhood window corresponding to the marked pixel point with a preset step size, and the target sliding window after each sliding is determined as a neighborhood window corresponding to the marked pixel point, wherein the union of all neighborhood windows corresponding to the marked pixel point is the overall neighborhood window corresponding to the marked pixel point.

9. An artificial intelligence-based cardiac space-occupying detection system in ultrasound images. It is characterized in that It comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement a method for detecting cardiac space-occupying space in an ultrasound image based on artificial intelligence as described in any one of claims 1 to 8.

Images