Image processing-based ultrasound puncture site anomaly detection method and system

Abstract

This invention relates to the technical field of breast ultrasound image recognition, specifically to a method and system for detecting abnormalities at ultrasound puncture sites based on image processing. In a patient's breast ultrasound image, all similar pixels of a reference pixel are acquired. A spatial coherence coefficient is calculated based on the distance distribution between the reference point and each similar pixel, the number of similar pixels, and regional distribution characteristics. Hyperechoic regions are identified by combining pixel grayscale distribution. For each hyperechoic region, landmark and non-landmark inflection points are detected. An abnormal extension coefficient is calculated based on the positional distribution of landmark and non-landmark points; thus, an abnormality factor is obtained, and abnormal inflection points are screened. Abnormal inflection points are clustered based on the abnormal extension coefficient and abnormality factor to form abnormal clusters. Contour fitting is performed on the abnormal inflection points within each cluster to obtain suspected mass regions. This invention can accurately identify suspected mass regions, thereby improving the detection efficiency and accuracy for relevant personnel.

Description

Technical Field

[0001] This invention relates to the technical field of breast ultrasound image recognition, and specifically to a method and system for detecting abnormalities at ultrasound puncture sites based on image processing. Background Technology

[0002] Mixed cystic-solid masses are a common type of mass, possessing both cystic and solid components, and typically exhibiting complex ultrasound imaging features. These masses are complex in nature; the cystic portion may contain fluid, while the solid portion may be an aggregation of mass cells, displaying different echogenic characteristics. Biopsy is a crucial method for diagnosing mixed cystic-solid masses. Biopsy obtains tissue samples from the mass, and pathological examination can clarify the nature of the mass, helping to determine its benign or malignant nature and specific components, thus providing a scientific basis for further treatment. With the development of image processing technology, combining related techniques to detect mixed cystic-solid mass regions in ultrasound images helps to pinpoint the specific location and morphological characteristics of the mass, guiding the accuracy and effectiveness of biopsies.

[0003] Mixed cystic and solid masses usually have a relatively complex shape, making it difficult to clearly identify the edges of the mass when directly using edge detection algorithms in the global image. Furthermore, the characteristics of the physiological tissues in the breast region result in relatively complex textures in ultrasound images, which also hinders the direct detection of the mass region by edge detection algorithms. Summary of the Invention

[0004] To address the challenges of complex morphologies in mixed cystic-solid breast masses, making it difficult to clearly identify the mass edges using edge detection algorithms in the overall image, and further hindering the direct detection of mass areas by edge detection algorithms due to the complex texture of breast tissue, this invention aims to provide an image processing-based method and system for detecting abnormalities at ultrasound puncture sites. The specific technical solution adopted is as follows:

[0005] A method for detecting abnormalities at ultrasound puncture sites based on image processing, the method comprising:

[0006] Acquire breast ultrasound images of the patient;

[0007] In a breast ultrasound image, a single pixel is selected as a reference pixel. All similar pixels with similar grayscale values ​​to the reference pixel are obtained. Based on the distance distribution between the reference pixel and each similar pixel, the number of similar pixels, and the regional distribution characteristics of all similar pixels, the spatial coherence coefficient of the reference pixel is obtained. Based on the spatial coherence coefficients and grayscale distribution of all pixels, all hyperechoic and hypoechoic regions in the breast ultrasound image are obtained. Landmark inflection points and non-landmark inflection points are obtained within each hyperechoic region. Based on the positional distribution between the landmark inflection points and each non-landmark inflection point, the abnormal extension coefficient of the landmark inflection points is obtained. Based on the magnitude relationship of the abnormal extension coefficients among the landmark inflection points within each hyperechoic region, the abnormality factor of all landmark inflection points within each hyperechoic region is obtained. Based on the abnormal extension coefficients of the landmark inflection points within each hyperechoic region, all landmark inflection points are filtered to obtain all abnormal inflection points.

[0008] All abnormal inflection points are clustered based on their abnormal extension coefficient and abnormal factor to obtain all abnormal clusters; contour fitting is performed on the abnormal inflection points in each abnormal cluster to obtain suspected mass regions.

[0009] Furthermore, the method for obtaining similar pixels includes:

[0010] Set a preset grayscale increase / decrease range for the reference pixel; select all pixels in the breast ultrasound image whose grayscale values ​​fall within the preset grayscale increase / decrease range as similar pixels to the reference pixel.

[0011] Furthermore, the method for obtaining the spatial coherence coefficients includes:

[0012] use The triangulation algorithm obtains the fitted connected region composed of all similar pixels of the reference pixel; the minimum bounding rectangle of the fitted connected region is obtained as the first rectangle;

[0013] The spatial continuity coefficient is obtained according to the formula for calculating the spatial continuity coefficient, which is shown below:

[0014]

[0015] In the formula, Represents the spatial coherence coefficient of the reference pixel; This represents the grayscale value of the reference pixel. This represents the length of the first rectangle; This represents the width of the first rectangle; Indicates the number of similar pixels; Indicates the first Gray values ​​of similar pixels; Indicates the reference pixel and the first Euclidean distance between similar pixels; This represents the absolute value function.

[0016] Furthermore, the methods for obtaining all high-echo and low-echo regions include:

[0017] Using the spatial coherence coefficient of each pixel, all pixels are clustered and the number of clusters is set to 2, thus obtaining two clusters. The mean gray value of the pixels in each cluster is calculated, and the pixels in the cluster with the larger mean gray value are regarded as high echo region pixels, and the pixels in the cluster with the smaller mean gray value are regarded as low echo region pixels.

[0018] All regions composed of pixels in the high-echo region are defined as all high-echo regions, and all regions composed of pixels in the low-echo region are defined as all low-echo regions.

[0019] Furthermore, the method for obtaining the marked inflection points and non-marked inflection points includes:

[0020] Edge pixels within each high-echo region are obtained using an edge detection algorithm; corner detection is used to obtain inflection points within each high-echo region; the two inflection points located at the two ends of the narrow section within each high-echo region are designated as marker inflection points, and the other inflection points are designated as non-marker inflection points.

[0021] Furthermore, the method for obtaining the abnormal extension coefficient includes:

[0022] Establish a Cartesian coordinate system with each inflection point as the origin;

[0023] Sort the non-marking inflection points according to their distance from each marking inflection point from closest to furthest, to obtain the non-marking inflection point sequence for each marking inflection point;

[0024] The abnormal extension coefficient is obtained according to the formula for calculating the abnormal extension coefficient, which is as follows:

[0025]

[0026] In the formula, This represents the abnormal extension coefficient of each inflection point; Indicates the number of non-marked inflection points in the non-marked inflection point sequence; Indicates the first non-marked inflection point sequence The serial number of each non-marking inflection point; Indicates the first non-marked inflection point sequence The ordinate of each non-marking inflection point; Indicates the first non-marked inflection point sequence The x-coordinates of non-marking inflection points; This represents the absolute value function.

[0027] Furthermore, the method for obtaining the abnormal factor includes:

[0028] The ratio between the maximum and minimum abnormal extension coefficients of the two marker inflection points in each hyperechoic region is used as the anomalous factor for all marker inflection points in each hyperechoic region.

[0029] Furthermore, the method for obtaining the abnormal inflection point includes:

[0030] The inflection point with the larger anomalous extension coefficient among the two inflection points in each hyperecho region is retained as the anomalous inflection point in each hyperecho region. All hyperecho regions are traversed to obtain all anomalous inflection points.

[0031] Furthermore, based on the abnormal extension coefficient and abnormal factor of each abnormal inflection point, all abnormal inflection points are clustered to obtain all abnormal clusters; contour fitting is performed on the abnormal inflection points in each abnormal cluster to obtain suspected mass regions, including:

[0032] By utilizing the abnormal extension coefficient and abnormal factor of each abnormal inflection point, the corresponding discrete point of each abnormal inflection point is constructed;

[0033] use The algorithm clusters all discrete points to obtain all clusters.

[0034] Each cluster is used as a reference cluster; the position of the abnormal inflection point corresponding to each discrete point in the reference cluster is obtained; the abnormal inflection point corresponding to each discrete point in the reference cluster is contour fitted to obtain the abnormal region corresponding to the abnormal inflection point of the reference cluster, and the abnormal region is used as the suspected mass region; all clusters are traversed to obtain all suspected mass regions.

[0035] An image processing-based ultrasonic puncture site abnormality detection system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the steps of the method described above.

[0036] The present invention has the following beneficial effects:

[0037] This invention acquires breast ultrasound images of patients. Since the fibrous connective tissue and fibrous strands inherent in the breast organ are long fibers or membranous tissues, they appear as band-like hyperechoic areas in breast ultrasound images, while hypoechoic areas appear as clumps or diffuse background tissue. To distinguish between hyperechoic and hypoechoic areas, the spatial coherence of different pixels with similar gray levels is analyzed. Because the spatial coherence of pixels within hyperechoic areas is relatively good, all hyperechoic and hypoechoic areas in the breast ultrasound image are obtained based on the spatial coherence coefficient and gray-level distribution of all pixels. Since the presence of a mass occupying normal tissue can disrupt the integrity of the hyperechoic area during its extension, the abnormal manifestations of the hyperechoic area during its extension are first analyzed. When the edges of a hyperechoic area are occupied by a lesion area of ​​a mixed cystic-solid mass, the smooth extension and integrity of the normal tissue are affected. When a region is damaged, the occupied area creates a cross-section in the hyperechoic region. This new cross-section leads to the emergence of new edges, which also contain inflection points. The two inflection points at the narrow ends are landmark inflection points. Compared to the landmark inflection points at the ends of normal regions, the non-landmark inflection points arising from the new edges show a significant deviation in their extension angles from the other inflection points. Therefore, landmark and non-landmark inflection points in the hyperechoic region are identified. The abnormal extension of the landmark inflection points is then analyzed to obtain an abnormal extension coefficient. After analyzing the abnormal extension of the landmark inflection points, the hyperechoic regions with abnormal extension are marked, narrowing down the possible location of the mass, thus obtaining all abnormal inflection points. Based on the abnormal extension coefficient and abnormal factor of each abnormal inflection point, all abnormal inflection points are clustered to obtain all abnormal clusters. Contour fitting is performed on the abnormal inflection points in each abnormal cluster to obtain suspected mass regions. This invention can accurately identify suspected mass regions, thereby improving the detection efficiency and accuracy for relevant personnel. Attached Figure Description

[0038] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of an image processing-based ultrasonic puncture site abnormality detection method according to an embodiment of the present invention;

[0040] Figure 2 This is a block diagram of an image processing-based ultrasonic puncture site abnormality detection system provided in one embodiment of the present invention. Detailed Implementation

[0041] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an image processing-based ultrasonic puncture site abnormality detection method and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0043] The following description, in conjunction with the accompanying drawings, details the specific scheme of the image processing-based ultrasonic puncture site abnormality detection method and system provided by the present invention.

[0044] Please see Figure 1 This illustrates an image processing-based method for detecting abnormalities at ultrasonic puncture sites according to an embodiment of the present invention. The method includes:

[0045] Step S1: Acquire breast ultrasound images of the patient.

[0046] The embodiments of the present invention are mainly applied to the scenario of identifying mixed cystic and solid masses in the breast of patients, so the first step is to acquire breast ultrasound images of patients.

[0047] In one embodiment of the present invention, the patient is placed in a supine or lateral decubitus position. After the target location is fully exposed, the physician first applies coupling agent evenly to the surface of the probe, then gently places the high-frequency linear array probe against the skin and performs a continuous sliding scan in a clockwise direction to the periphery. The probe is kept perpendicular to the chest wall and moderate pressure is applied to obtain two-dimensional grayscale images of the breast without omissions in each quadrant, which are then used as breast ultrasound images required for subsequent analysis.

[0048] Step S2: Select any pixel in the breast ultrasound image as a reference pixel; obtain all similar pixels with similar gray values ​​to the reference pixel; obtain the spatial coherence coefficient of the reference pixel based on the distance distribution between the reference pixel and each similar pixel, the number of similar pixels, and the regional distribution characteristics of all similar pixels; obtain all hyperechoic regions in the breast ultrasound image based on the spatial coherence coefficients and gray value distribution of all pixels; obtain the landmark inflection points and non-landmark inflection points in each hyperechoic region; obtain the abnormal extension coefficient of the landmark inflection points based on the positional distribution between the landmark inflection points and each non-landmark inflection point; obtain the abnormality factor of all landmark inflection points in each hyperechoic region based on the magnitude relationship of the abnormal extension coefficients among the landmark inflection points in each hyperechoic region; filter all landmark inflection points based on the abnormal extension coefficients of the landmark inflection points in each hyperechoic region to obtain all abnormal inflection points.

[0049] In reality, mixed cystic-solid masses in the breast appear in breast ultrasound images with the following grayscale characteristics: the cystic portion is hypoechoic, and the solid portion is hyperechoic. It should be noted that hypoechoic means low grayscale value, and hyperechoic means high grayscale value. Furthermore, the cystic cavity occupies a large portion of the volume and surrounds the solid portion, causing the hyperechoic solid portion to remain free within the cystic cavity. In terms of location, mixed cystic-solid masses can occupy areas that originally exhibited hyperechoic characteristics in normal tissues, thus disrupting the grayscale representation of normal breast organs. Since the fibrous connective tissue and fibrous strands inherent in breast organs are long fibers or membranous tissues, they appear as band-like hyperechoic areas in breast ultrasound images, while hypoechoic areas appear as clumps or diffuse background tissue. To distinguish between hyperechoic and hypoechoic areas, this embodiment of the invention analyzes the spatial coherence of different pixels with similar grayscale values.

[0050] Preferably, in one embodiment of the present invention, the method for obtaining similar pixels includes:

[0051] A preset grayscale increase / decrease range is set for the reference pixel; all pixels in the breast ultrasound image whose grayscale values ​​fall within the preset grayscale increase / decrease range are considered as similar pixels to the reference pixel. In one embodiment of the present invention, the grayscale value of the reference pixel is set to... The preset grayscale increase / decrease range is set to It should be noted that the preset grayscale increase / decrease range can be set by the user and is not limited here.

[0052] Preferably, in one embodiment of the present invention, the method for obtaining the spatial coherence coefficient includes:

[0053] use The triangulation algorithm obtains the fitted connected region formed by all similar pixels of the reference pixel; the minimum bounding rectangle of the fitted connected region is obtained as the first rectangle. In one embodiment of the present invention, The triangulation algorithm is a well-known technique in the field and will not be elaborated here.

[0054] The spatial continuity coefficient is obtained according to the formula for calculating the spatial continuity coefficient, which is shown below:

[0055]

[0056] In the formula, Represents the spatial coherence coefficient of the reference pixel; This represents the grayscale value of the reference pixel. This represents the length of the first rectangle; This represents the width of the first rectangle; Indicates the number of similar pixels; Indicates the first Gray values ​​of similar pixels; Indicates the reference pixel and the first Euclidean distance between similar pixels; This represents the absolute value function.

[0057] In the formula for calculating the spatial coherence coefficient, the higher the gray value of the reference pixel, the more likely the reference pixel is to be a pixel in a hyperechoic region, and the larger the spatial coherence coefficient of the reference pixel. Since hyperechoic regions exhibit a strip-like characteristic, the ratio between the length and width of the first rectangle is... The larger the value, the more likely the region composed of pixels with similar gray levels to the reference pixel is to be a hyperechoic region. In this case, the spatial coherence coefficient of the reference pixel is considered to be larger. Since the pixels in the hyperechoic region are more densely clustered and have more similar gray values, the gray level difference between similar pixels and the reference pixel is larger. The smaller, and the distance The smaller the value, the more likely the reference pixel is located in a high-echo region, and the greater the spatial coherence coefficient of the reference pixel.

[0058] Since the spatial coherence of pixels in hyperechoic regions is good, in this embodiment of the invention, all hyperechoic and hypoechoic regions in a breast ultrasound image are obtained based on the spatial coherence coefficient and grayscale distribution of all pixels.

[0059] Preferably, in one embodiment of the present invention, the method for obtaining all high-echo regions and low-echo regions includes:

[0060] Using the spatial coherence coefficient of each pixel, all pixels are clustered, and the number of clusters is set to 2, thus obtaining two clusters. The mean gray value of the pixels in each cluster is calculated, and the pixels in the cluster with the larger mean gray value are regarded as high echo region pixels, and the pixels in the cluster with the smaller mean gray value are regarded as low echo region pixels.

[0061] All regions composed of pixels in the high-echo region are defined as all high-echo regions, and all regions composed of pixels in the low-echo region are defined as all low-echo regions.

[0062] In reality, hyperechoic structures in the breast, such as ligaments and fascia, appear hyperechoic in ultrasound images due to their rich collagen fiber content. These structures naturally integrate into surrounding tissue at their ends, typically exhibiting a naturally extending and gradually tapering shape. Therefore, in breast ultrasound images, the beginning and end of a hyperechoic region containing normal physiological tissue should tend to be naturally closed. When a hypoechoic region resembling a mass appears, the mass's encroachment on normal tissue disrupts the integrity of the hyperechoic region during its extension. Therefore, in this embodiment of the invention, the abnormal manifestations of hyperechoic regions during their extension are first analyzed.

[0063] In a normal hyperechoic region, the difference in extension angle between the inflection points of the region's edge is small, and the transition is smooth. However, when a lesion of mixed cystic and solid mass occupies the area, the smooth extension and integrity of normal tissue are disrupted. At this time, the occupied area creates a cross-section in the hyperechoic region, and the formation of the new cross-section leads to the appearance of new edges. Inflection points also exist on the new edges, and the two inflection points at the two ends of the narrow elongation are landmark inflection points. Compared with the landmark inflection points at the ends of the normal region, the non-landmark inflection points that appear due to the new edges deviate significantly from the extension angles of the other inflection points in terms of the transition. Therefore, in this embodiment of the invention, landmark and non-landmark inflection points in the hyperechoic region are obtained.

[0064] Preferably, in one embodiment of the present invention, the method for obtaining marked inflection points and non-marked inflection points includes:

[0065] Edge pixels within each hyperechoic region are obtained using an edge detection algorithm; inflection points within each hyperechoic region are obtained using a corner detection algorithm. Since normal hyperechoic regions exhibit a naturally extending and gradually tapering shape, the two inflection points at both ends of the narrow region are designated as marker inflection points. Other inflection points may indicate changes in body tissue or the presence of lumps; therefore, these other inflection points are considered non-marker inflection points for easier differentiation. Furthermore, in one embodiment of the invention, [the following is used]... The corner detection algorithm is used, and this algorithm is a well-known technique in the art, so it will not be described in detail here.

[0066] Next, the extension anomalies of the marker inflection points are analyzed. Preferably, in one embodiment of the present invention, the method for obtaining the anomaly extension coefficient includes:

[0067] A Cartesian coordinate system is established with each marker inflection point as the origin. Non-marker inflection points are sorted from near to far according to their distance from each marker inflection point to obtain the non-marker inflection point sequence for each marker inflection point. Since the analysis at both ends is consistent, in this embodiment of the invention, the detachment only analyzes the extension anomaly of the marker inflection point at one end.

[0068] The anomaly extension coefficient is obtained according to the formula for calculating the anomaly extension coefficient, which is shown below:

[0069]

[0070] In the formula, This represents the abnormal extension coefficient of each inflection point; Indicates the number of non-marked inflection points in the non-marked inflection point sequence; Indicates the first non-marked inflection point sequence The serial number of each non-marking inflection point; Indicates the first non-marked inflection point sequence The ordinate of each non-marking inflection point; Indicates the first non-marked inflection point sequence The x-coordinates of non-marking inflection points; This represents the absolute value function.

[0071] In the formula for calculating the abnormal extension coefficient, the first... The smaller the serial number of a non-marker inflection point, the closer it is to a marker inflection point. Non-marker points closer to the marker inflection point exhibit a more pronounced deviation in extension angle, thus their corresponding weights are adjusted accordingly. The larger; Indicates the first The slope of the line connecting each non-mark inflection point to a mark inflection point is considered. A steeper slope indicates a larger deviation angle between the non-mark inflection point and the mark inflection point, corresponding to a more abrupt transition in the magnitude of change. The difference between the deviation angle of a non-mark inflection point and the average deviation angle of all non-mark inflection points is also considered. A larger difference indicates a more pronounced deviation at that inflection point, thus contributing more to the definition of the anomaly coefficient. Since non-mark inflection points have higher weights the closer they are to the mark, and in reality, anomalies tend to occur near mark inflection points, this is used... right We perform weighted averaging to obtain the abnormal extension coefficient of the inflection point.

[0072] After analyzing the abnormal extension of the landmark inflection point, the hyperechoic areas of the abnormal extension were marked, and the possible location of the mass was narrowed down.

[0073] Preferably, in one embodiment of the present invention, the method for obtaining the abnormal factor includes:

[0074] The ratio between the maximum and minimum abnormal extension coefficients at the two landmark inflection points in each hyperechoic region is taken as the abnormality factor of all landmark inflection points in each hyperechoic region. The larger the ratio, the more significant the difference in extension abnormality at both ends of the hyperechoic region. This indicates that there may be a mass in the hyperechoic region. The more abnormal the landmark inflection point in the hyperechoic region, the larger the abnormality factor of the landmark inflection point.

[0075] Preferably, in one embodiment of the present invention, the method for obtaining abnormal inflection points includes:

[0076] In each hyperechoic region, the inflection point with the larger abnormal extension coefficient is retained. The larger the abnormal extension coefficient, the more likely a mass area is to appear near the inflection point. Therefore, the inflection point with the larger abnormal extension coefficient is taken as the abnormal inflection point in each hyperechoic region. All hyperechoic regions are traversed to obtain all abnormal inflection points.

[0077] Step S3: Cluster all abnormal inflection points according to the abnormal extension coefficient and abnormal factor of each abnormal inflection point to obtain all abnormal clusters; perform contour fitting on the abnormal inflection points in each abnormal cluster to obtain the suspected mass region.

[0078] Preferably, in one embodiment of the present invention, the step of obtaining the suspected mass region includes:

[0079] A rectangular coordinate system is established with the anomaly extension coefficient as the horizontal axis and the anomaly factor as the vertical axis. The coordinates of the anomaly inflection point in the rectangular coordinate system are obtained by using the anomaly extension coefficient and the anomaly factor of each anomaly inflection point. Each coordinate point is used as the corresponding discrete point of each anomaly inflection point.

[0080] use The algorithm clusters all discrete points to obtain all clusters. It should be noted that... Algorithms are well-known technical means in the field of science and will not be elaborated here.

[0081] Each cluster is used as a reference cluster; the position of the abnormal inflection point corresponding to each discrete point in the reference cluster is obtained; the abnormal inflection point corresponding to each discrete point in the reference cluster is contour fitted to obtain the abnormal region corresponding to the abnormal inflection point of the reference cluster, and the abnormal region is used as the suspected mass region; all clusters are traversed to obtain all suspected mass regions.

[0082] In summary, the following steps were performed: acquiring breast ultrasound images of the patient; selecting any pixel in the breast ultrasound image as a reference pixel; obtaining all similar pixels with similar grayscale values ​​to the reference pixel; obtaining the spatial coherence coefficient of the reference pixel based on the distance distribution between the reference pixel and each similar pixel, the number of similar pixels, and the regional distribution characteristics of all similar pixels; identifying all hyperechoic and hypoechoic regions in the breast ultrasound image based on the spatial coherence coefficients and grayscale distribution of all pixels; identifying landmark inflection points and non-landmark inflection points in each hyperechoic region; and further analyzing the distance distribution between the reference pixel and each similar pixel, and then analyzing the spatial coherence coefficients and grayscale distribution of all pixels. The abnormal extension coefficient of the landmark inflection point is obtained by analyzing the positional distribution between each non-landmark inflection point; the abnormality factor of all landmark inflection points in each hyperechoic region is obtained based on the magnitude relationship of the abnormal extension coefficients between landmark inflection points in each hyperechoic region; all landmark inflection points are screened based on the abnormal extension coefficients of the landmark inflection points in each hyperechoic region to obtain all abnormal inflection points; all abnormal inflection points are clustered based on the abnormal extension coefficient and abnormality factor of each abnormal inflection point to obtain all abnormal clusters; the abnormal inflection points in each abnormal cluster are contour fitted to obtain suspected mass regions.

[0083] One embodiment of the present invention provides an image processing-based ultrasonic puncture site abnormality detection system. The system includes a memory, a processor, and a computer program. The memory stores the corresponding computer program, and the processor runs the corresponding computer program. When the computer program runs in the processor, it can implement the methods described in steps S1-S3, specifically including:

[0084] Image acquisition module 101 is used to acquire breast ultrasound images of patients;

[0085] The inflection point filtering module 102 is used to: select any pixel in a breast ultrasound image as a reference pixel; acquire all similar pixels with similar gray values ​​to the reference pixel; obtain the spatial coherence coefficient of the reference pixel based on the distance distribution between the reference pixel and each similar pixel, the number of similar pixels, and the regional distribution characteristics of all similar pixels; acquire all hyperechoic and hypoechoic regions in the breast ultrasound image based on the spatial coherence coefficients and gray value distribution of all pixels; acquire landmark inflection points and non-landmark inflection points in each hyperechoic region; obtain the abnormal extension coefficient of landmark inflection points based on the positional distribution between landmark inflection points and each non-landmark inflection point; obtain the abnormality factor of all landmark inflection points in each hyperechoic region based on the magnitude relationship of the abnormal extension coefficients among landmark inflection points in each hyperechoic region; and filter all landmark inflection points based on the abnormal extension coefficients of landmark inflection points in each hyperechoic region to obtain all abnormal inflection points.

[0086] The mass region identification module 103 is used to cluster all abnormal inflection points according to the abnormal extension coefficient and abnormal factor of each abnormal inflection point to obtain all abnormal clusters; and to perform contour fitting on the abnormal inflection points in each abnormal cluster to obtain the suspected mass region.

[0087] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0088] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for detecting abnormalities at ultrasonic puncture sites based on image processing, characterized in that, The method includes: acquiring a patient's breast ultrasound image; randomly selecting a pixel in the breast ultrasound image as a reference pixel; obtaining all similar pixels with similar grayscale values ​​to the reference pixel; and obtaining the spatial coherence coefficient of the reference pixel based on the distance distribution between the reference pixel and each similar pixel, the number of similar pixels, and the regional distribution characteristics of all similar pixels; the method for obtaining the spatial coherence coefficient includes: utilizing... The triangulation algorithm obtains a fitted connected region composed of all similar pixels of the reference pixel; the minimum bounding rectangle of the fitted connected region is obtained as the first rectangle; the spatial continuity coefficient is obtained according to the spatial continuity coefficient calculation formula, which is shown below: ； In the formula, Represents the spatial coherence coefficient of the reference pixel; This represents the grayscale value of the reference pixel. This represents the length of the first rectangle; This represents the width of the first rectangle; Indicates the number of similar pixels; Indicates the first Gray values ​​of similar pixels; Indicates the reference pixel and the first Euclidean distance between similar pixels; The method represents an absolute value function; based on the spatial coherence coefficient and grayscale distribution of all pixels, all hyperechoic regions in the breast ultrasound image are obtained; the landmark inflection points and non-landmark inflection points in each hyperechoic region are obtained; the method for obtaining the landmark inflection points and non-landmark inflection points includes: using an edge detection algorithm to obtain edge pixels in each hyperechoic region; using a corner detection algorithm to obtain inflection points in each hyperechoic region; taking the two inflection points located at the two ends of the narrow section in each hyperechoic region as landmark inflection points, and taking the other inflection points as non-landmark inflection points; obtaining the abnormal extension coefficient of the landmark inflection points based on the positional distribution between the landmark inflection points and each non-landmark inflection point; the method for obtaining the abnormal extension coefficient includes: establishing a Cartesian coordinate system with each landmark inflection point as the origin; sorting the non-landmark inflection points from near to far according to their distance from each landmark inflection point to obtain the non-landmark inflection point sequence of each landmark inflection point; obtaining the abnormal extension coefficient according to the abnormal extension coefficient calculation formula, which is shown below: ； In the formula, This represents the abnormal extension coefficient of each inflection point; Indicates the number of non-marked inflection points in the non-marked inflection point sequence; Indicates the first non-marked inflection point sequence The serial number of each non-marking inflection point; Indicates the first non-marked inflection point sequence The ordinate of each non-marking inflection point; Indicates the first non-marked inflection point sequence The x-coordinates of non-marking inflection points; The function represents the absolute value. Based on the relationship between the abnormal extension coefficients of the landmark inflection points in each hyperechoic region, the abnormality factor of all landmark inflection points in each hyperechoic region is obtained. Based on the abnormal extension coefficient of the landmark inflection points in each hyperechoic region, all landmark inflection points are filtered to obtain all abnormal inflection points. Based on the abnormal extension coefficient and abnormal factor of each abnormal inflection point, all abnormal inflection points are clustered to obtain all abnormal clusters. Contour fitting is performed on the abnormal inflection points in each abnormal cluster to obtain suspected mass regions.

2. The method for detecting abnormalities at ultrasonic puncture sites based on image processing according to claim 1, characterized in that, The method for obtaining similar pixels includes: setting a preset grayscale increase / decrease range for reference pixels; and taking all pixels in the breast ultrasound image whose grayscale values ​​are within the preset grayscale increase / decrease range as similar pixels to the reference pixels.

3. The method for detecting abnormalities at ultrasonic puncture sites based on image processing according to claim 1, characterized in that, The method for obtaining all high-echo and low-echo regions includes: using the spatial coherence coefficient of each pixel to cluster all pixels, and setting the number of clusters to 2, thus obtaining two clusters; calculating the mean gray value of pixels in each cluster, and taking the pixels in the cluster with the larger mean gray value as high-echo region pixels, and taking the pixels in the cluster with the smaller mean gray value as low-echo region pixels; taking all regions composed of high-echo region pixels as all high-echo regions, and taking all regions composed of low-echo region pixels as all low-echo regions.

4. The method for detecting abnormalities at ultrasonic puncture sites based on image processing according to claim 1, characterized in that, The method for obtaining the anomalous factor includes: taking the ratio between the maximum and minimum anomalous extension coefficients of the two marker inflection points in each high-echo region as the anomalous factor of all marker inflection points in each high-echo region.

5. The method for detecting abnormalities at ultrasonic puncture sites based on image processing according to claim 1, characterized in that, The method for obtaining abnormal inflection points includes: retaining the inflection point with the larger abnormal extension coefficient among the two inflection points in each hyperecho region as the abnormal inflection point in each hyperecho region, and traversing all hyperecho regions to obtain all abnormal inflection points.

6. The method for detecting abnormalities at ultrasonic puncture sites based on image processing according to claim 1, characterized in that, Based on the abnormal extension coefficient and abnormal factor of each abnormal inflection point, all abnormal inflection points are clustered to obtain all abnormal clusters. Contour fitting is performed on the abnormal inflection points in each abnormal cluster to obtain suspected tumor regions, including: constructing the corresponding discrete points for each abnormal inflection point using the abnormal extension coefficient and abnormal factor; and utilizing... The algorithm performs clustering on all discrete points to obtain all clusters; each cluster is used as a reference cluster; the position of the corresponding abnormal inflection point of each discrete point in the reference cluster is obtained; contour fitting is performed on the corresponding abnormal inflection point of each discrete point in the reference cluster to obtain the abnormal region corresponding to the abnormal inflection point of the reference cluster, and the abnormal region is used as the suspected mass region; all clusters are traversed to obtain all suspected mass regions.

7. An image processing-based ultrasonic puncture site abnormality detection system, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6.

Images